3U Xserve? [Archive] - MacRumors Forums

Laslo Panaflex

Oct 21, 2003, 01:29 PM

Originally posted by pgwalsh
With the PowerMac setup they were concerned with noise so they put on massive heatsinks. This allowed them to avoid putting a fan on the heatsink. That's not the case with the xserve. Granted I'd think they'd like to make it quieter, but I don't think that's a real concern. Most server are very loud. It's all about performance.

Yeah, I understand, but you really think they can put a G5 in a 1U enclosure with your CURRENT knowledge of the G5, I don't think so. That's does not mean they don't have lower heat producing chips from IBM or a better way to cool the things. I just think that it is more feesable for apple to make a 3U xserve, considering if it comes out soon they would have to have been working on the designs for a while and have to design for the then current hardware. Sure will will see a 1U xserve, just not anytime soon.

-end opinion

cubist

Oct 21, 2003, 02:24 PM

The current Xserves use blowers, not fans. They are more than adequate for cooling G5s. As a previous poster noted, they are very, very noisy. The cooling approach taken in the PowerMac G5 is very quiet, and that's why it takes so much space.Unless Apple's gotten some flak about the Xserve's noise, they wouldn't go to 3U just for cooling considerations.

dho

Oct 21, 2003, 02:30 PM

mustang_dvs:

its loud, but I soo want one :)

Sun Baked

Oct 21, 2003, 02:51 PM

Even a 1.0-1.33GHz G5 in the server would be a huge step up.

The increased I/O capacity would be a HUGE boost for the XServe.

A 2-4 PPC 970 IBM blade server with Mac OS X Server running on it, would be an interesting beast. ;)

pgwalsh

Oct 21, 2003, 03:07 PM

Originally posted by Laslo Panaflex
Yeah, I understand, but you really think they can put a G5 in a 1U enclosure with your CURRENT knowledge of the G5, I don't think so. That's does not mean they don't have lower heat producing chips from IBM or a better way to cool the things. I just think that it is more feesable for apple to make a 3U xserve, considering if it comes out soon they would have to have been working on the designs for a while and have to design for the then current hardware. Sure will will see a 1U xserve, just not anytime soon.

-end opinion Yes I think they can put a G5 in a 1U enclosure... The 1 Unit rack space is a huge selling point for the Xserve. That would be something big to give up. However if they put a couple G5's in it, they might want to add a nonstick tephlon top and a graphite spatula.

backspinner

Oct 21, 2003, 03:22 PM

I'm in the market for a new server. I need decent storage, I mean not size but hot-swap and fail save. And I need a real server with a real service contract. The Xserve has that. But I have minor problems with the depth of the system. Maybe a 3U system will be less deep? That would be cool.

peterjhill

Oct 21, 2003, 03:46 PM

Look at dells rack server offerings. They are some of the best rack mounted pc's in the business. They have 1RU servers and larger 2 and 4 RU machines. The 2RU is a sweet spot, allowing 2 processors, a decent amount of disk, and a decent amount of expandability.

I would hope that apple could squeeze two G5s into 2RUs. One RU is asking too much, IMHO. I would rather see a product sooner rather than later. If they are doing 3 RU's I hope they have 4 procs and at least 4 disks. They could potentially mount the disks vertically and fit more than 4, how about 8 disks and 4 G5's, that would be a sweet machine.

Phil Of Mac

Oct 21, 2003, 04:15 PM

Air conditioning.

Enough said.

rsnyder@psu.edu

Oct 21, 2003, 04:17 PM

The story focuses on the Xgrid project mailing list that slipped out. Personally, I am wondering if the xGrid has anything to do with Oracle's 10g (as in "Grid" computing). The word at Oracle World was that Oracle will release a full OS X version with the release of 10g. There was also speculation that this would require the G5.

Since both Oracle and now Apple are talking grid computing . . . . would be sweat to run Oracle on OS X now that Sun is in dumper.

I am holding two servers in my budget waiting for the G5 xServes. My goal is in three years to get off Sun and its overpriced hardware. The Apple reps know that there are more than a few like me holding off spending until the G5 xServe arrives.

szark

Oct 21, 2003, 04:51 PM

Originally posted by rsnyder@psu.edu
Personally, I am wondering if the xGrid has anything to do with Oracle's 10g (as in "Grid" computing).

Possibly. I remember hearing another rumor that it had something to do with Sun's Grid Engine (http://wwws.sun.com/software/gridware/) software.

Personally, I think a 3U is perfect for a blade server design, or a combination Xserve/Xserve RAID (one half RAID, one half server). :D

DharvaBinky

Oct 21, 2003, 04:59 PM

Blade Chassis are typically 3U high. Perhaps we're talking about a Blade system here. with OS X and remote desktop, Blades make perfect sense.

Additionally, we have a bunch of Dell server kit here in my office. Including 1U systems, 2U systems, and 7U systems. The 1U systems offer good density, but only support 3 HD and an optical (so you can barely do RAID 5 on them), if you ditch the optical, you can fit 4 drives in the Chassis (we have some Dell 1U NAS boxes that are done this way). A 3U Xserve would allow 7-8 3.5" hotswap drives with an optical drive and some front mounted ports (if they wanted), however, it seems to me that maybe we're not thinking "far-out" enough... Given Apple's recent prediliction for ATA drives in servers, how about this:

A 4 processor 3U Xserv with:
14!! x 2.5" 80Gb 7200RPM ATA100 Drives (stacked like in an xRAID but slightly shorter because of the 2.5")

at RAID 5 that would give you about a Terrabyte of storage. not bad for a standalone server with no external storage chassis.

I think that Apple would be wise to diversify their server offerings. Keep the Mobo and most of the internals exactly the same, just add 2U and 3U chassis that allow for internal expansion. people like to have storage in their servers w/o having to buy the external chassis.

Also, G5 blades would be just plain sweet!

Dharvabinky

Phil Of Mac

Oct 21, 2003, 05:07 PM

IBM wants to make blade servers with the 970, or at least that's what they said before Apple confirmed that they were using the 970. Perhaps the "blade server" story was a cover.

Rincewind42

Oct 21, 2003, 06:11 PM

I think that Apple will stick with 1U. As has been said, the massive cooling project that is the PowerMac G5 is due primarily to it being a desktop system - it can't be excessively loud. However, I don't think we'll see a G5 Xserve soon - rather I think we'll see them when 90 nm G5 become common. Add to that 4 SATA drives, each on it's own channel and hot plug-able (yes, there is enough room on the hypertransport bus for that).

If Apple did want to make the Xserve bigger, then I would expect no larger than 2U, and then I would also expect quad CPU options at least. Apple's biggest advantage on the Xserve right now is density, if they give up that advantage then there had better be a damn good reason for it.

ColdZero

Oct 22, 2003, 12:16 AM

Originally posted by Rincewind42
I think that Apple will stick with 1U. As has been said, the massive cooling project that is the PowerMac G5 is due primarily to it being a desktop system - it can't be excessively loud. However, I don't think we'll see a G5 Xserve soon - rather I think we'll see them when 90 nm G5 become common. Add to that 4 SATA drives, each on it's own channel and hot plug-able (yes, there is enough room on the hypertransport bus for that).

If Apple did want to make the Xserve bigger, then I would expect no larger than 2U, and then I would also expect quad CPU options at least. Apple's biggest advantage on the Xserve right now is density, if they give up that advantage then there had better be a damn good reason for it.

There is 3.2gbps availible on the HT link in the G5. Thats gonna get filled up very quickly. You have 2 gigabit ethernet interfaces. 2 PCI-X slots that are going to take 2 gbps together. You're already over 3 gbps (assuming the 2nd ethernet is in a pci-x). 3.2 gbps is fast, but it could get overloaded in a server. As for 4 proc 2u machines, think again. I have never seen a 4 proc 2u machine before. Especially one that needs so much cooling like the g5. I think a 1u dual processor 1.8ghz g5 is where it's at. There will be no quad configs in 2u. Somebody even suggested 10 cpus in 3u, are you crazy? Apple isn't making something that big, even if they could manipulate the very fabric of space itself to fit 10 processors in 3u, they aren't a big enough player in the high end market for that kind of hardware yet. The Xserve and XRaid are nice, but they don't stand up to really big iron, they aren't made for it. When apple has to worry about making big iron, then I'll be happy.

Jeff Harrell

Oct 22, 2003, 01:07 AM

Originally posted by peterjhill
I would hope that apple could squeeze two G5s into 2RUs.Have a look at SGI's Tezro/Origin 350/Onyx 4. (All the same thing, basically.) It squeezes four R16000 CPU's plus disk plus PCI-X into a two-rack-unit enclosure. Apple should, at least in theory, be able to do the same thing. The 970 is big and hot, but it's not that big and hot.

Jeff Harrell

Oct 22, 2003, 01:18 AM

Originally posted by rsnyder@psu.edu
Personally, I am wondering if the xGrid has anything to do with Oracle's 10g (as in "Grid" computing).Okay, little tutorial here. "Grid computing" is a similar idea to cluster computing, but different in an important way. In cluster computing, you take a bunch of machines and arrange them, via hardware and software, so they can act like a single machine in some very limited ways. It usually involves some kind of task-master dolling out jobs to individual members of the cluster and receiving the results. Sometimes, though not always, clusters include special high-speed hardware to reduce inter-node latency.

Grid computing is just like cluster computing, only without the cluster. The term "grid" comes from the notion of an electrical grid. Nobody knows exactly where a given watt of electricity comes from (if you'll pardon the metaphor). We just plug our lamps in and they turn on.

With grid computing, networks of standalone computers form autonomous ad hoc clusters. When you send a compute task out into the grid, you don't know exactly where the CPU cycles to execute that job are coming from, nor do you care. You just send your job off, and then the results come back later.

Grid computing is neat because it maximizes IT ROI. Accounting's database server is idle in the middle of the night; use it to run CFD calculations for engineering. And so on.

The key to grid computing is in the software. It has to be smart enough to do all of that stuff I described by itself, without human intervention.

Guess what? It's already here. It's called Qmaster, and it's a part of both Shake and Compressor. Qmaster uses Rendezvous to assemble standalone computers into an autonomous ad hoc cluster. Because it's based on Rendezvous, it requires no setup or administration. When you take a bunch of machines running Shake and distribute rendering jobs across them with Qmaster, you're not using a cluster. You're using a grid.

Xgrid is almost certainly just going to be an extension and generalization of this basic idea, quite possibly using Qmaster to do it.

(Pardon me if I was pedantic at all. I had to look all this stuff up, see, so I was all excited about being able to share what I'd learned.)

tam

Oct 22, 2003, 01:58 AM

Originally posted by cubist
The current Xserves use blowers, not fans. They are more than adequate for cooling G5s. As a previous poster noted, they are very, very noisy. The cooling approach taken in the PowerMac G5 is very quiet, and that's why it takes so much space.Unless Apple's gotten some flak about the Xserve's noise, they wouldn't go to 3U just for cooling considerations.
I used to think so too, but then we got some hp dl360 g3 running linux without hp's special software. You can hear them through the wall of the server room, the five of them produce even more noise than the three huge cooling towers. Noise isn't that much of an issue in a server room, but they don't have to make it an issue like hp. Currently rack space isn't a problem for us, and I tend to think 1U machines are less sturdy compared to 2U. On the other hand we don't mind being able to mount it with less than four people because of the weight. :)

mproud

Oct 22, 2003, 03:43 AM

Is it just me or did ThinkSecret re-design their site?

manitoubalck

Oct 22, 2003, 07:41 AM

Originally posted by pgwalsh
3U seems big just for heat concerns. I could see a 2U, but with a 3U unit I'd imagine more Processors.. Maybe a 4 to 6 processor unit. I wonder how loud this unit will be?

You guys have appeared to miss the point a bit. It would make sense for the use of a water-cooling solution if the heat were that much of an issue. Plus the water-cooling systems are much smaller and quieter than the air/air units not including the external radiator of corse. This relates to smaller cases, even a 1U since water units can be made of almost any size, and the cooling efficiency remains about the same.

Who could deny the silence and cooling potential, not to mention the performance increase form one of these babies?

Jeff Harrell

Oct 22, 2003, 09:28 AM

Originally posted by manitoubalck
Who could deny the silence and cooling potential, not to mention the performance increase form one of these babies? Me.

Water cooling sounds like a good idea, but the cost is just absurd. Even closed-loop cold plate cooling, which can use tap water instead of Fluorinert, requires a ton of infrastructure that most equipment rooms just don't have. There's the plumbing, first of all, but then there's the heat exchanger, a sizable investment for somebody who just wants to have a render farm for After Effects or whatever.

A few years ago, SGI came up with a plan to use cold-plate liquid cooling to increase the processor density of their Origin 2000 server. This was a machine that already had an entry price in the low six figures, and could easily scale up to millions of dollars, and SGI still rejected the idea because it was going to be too expensive for most of their customers. And SGI isn't known for being a cost-conscious company, you know?

ffakr

Oct 22, 2003, 12:25 PM

Originally posted by Laslo Panaflex
With the current size of the G5 heatsink the G5 xserve would have to be 3U. There is no way that they can make the current G5 in a 1U enclosure. But, they might be working on some sort of new way to cool the G5. . .
Even the 2.0 GHz G5s are quite a bit cooler than Athlons or Pentium 4s and there are hundreds of dual processor, 1U Athlon and Pentium 4 servers to choose from.

Apple could easily put dual G5s in a 1U enclosure.

This is why we have copper heat sinks and 7000 rpm fans.

Real server rooms have chillers anyway. If you can actually work at a console in your server room, it's too damn warm in there. :-)

pgwalsh

Oct 22, 2003, 12:44 PM

Originally posted by ffakr
Even the 2.0 GHz G5s are quite a bit cooler than Athlons or Pentium 4s and there are hundreds of dual processor, 1U Athlon and Pentium 4 servers to choose from.

Apple could easily put dual G5s in a 1U enclosure.

This is why we have copper heat sinks and 7000 rpm fans.

Real server rooms have chillers anyway. If you can actually work at a console in your server room, it's too damn warm in there. :-) Right... Apple's Massive heatsinks an 9 fans were to minimize the overall heat and to move as much air as possible to keep the noise to a minimum or below 30db's.

You could run them in thin box with small heatsinks and fewer, but more powerful fans. Now if they only come out with 2, 4, 6, and 64 processor versions. hehe not that I'd buy one, but it would cooool.

backspinner

Oct 22, 2003, 02:30 PM

Originally posted by pgwalsh but it would cooool.
No, it would be waaaarm! :p

timothyjoelwrig

Oct 22, 2003, 03:51 PM

http://www.wired.com/news/technology/0,1282,60733,00.html

Apple, AMD, and Intel are already on the bandwagon. I've read their concepts and they seem realistically implementable.

Jeff Harrell

Oct 22, 2003, 03:58 PM

Originally posted by timothyjoelwrig
Apple, AMD, and Intel are already on the bandwagon.That's not what the article says. It says they "cooperated," meaning they probably provided equipment to use for testing.

Blue-sky projects are really neat. But it's a long way from three guys at Stanford to your desk, and a lot can happen along the way to derail the operation.

savar

Oct 22, 2003, 04:22 PM

Originally posted by Jeff Harrell
Grid computing is just like cluster computing, only without the cluster. The term "grid" comes from the notion of an electrical grid. Nobody knows exactly where a given watt of electricity comes from (if you'll pardon the metaphor). We just plug our lamps in and they turn on.

So do you think this represents Apple's attempt to make a stab at corporate environments once again? Tech companies looking to maximize ROI can leverage unused processor cycles from all over the company to run design simulations...meaning that putting an eMac on every person's desk turns out to be a great investment, and a few xServes in the back end organize it all. End-to-end solution sounds pretty compelling. Does Apple have the brawn to pull this off, however?

It also sounds like a great way to reinvigorate university sales. Researchers who already favor macs will certainly pressure IT managers to buy them computers that will be useful beyond the relatively "dumb" features, like e-mail and web surfing and WP, which don't require most of the processing capabilities available on any platform.

Personally, I think that watching Apple position itself under Steve Jobs' direction is fascinating. I can't wait to see where they are headed and what they pull off.

stingerman

Oct 22, 2003, 04:24 PM

Originally posted by pgwalsh
Now if they only come out with 2, 4, 6, and 64 processor versions. hehe not that I'd buy one, but it would cooool.

I believe the G5 has 3 coherent Processor Interconnects. That means an 8-Way SMP is the limit (2^3=8). After that you need to go into a grid formation.

I personally feel that Apple will not enter an already saturated server market in the 2U space and higher. Initially I thought a 2U G5 would be cool, but open further examination of Apple's previous practices, it doesn't make sense. Apple needs to innovate when it comes out with a new line. I believe the xServe RAID is the model Apple will follow.

Can Apple fit G5's in a blade configuration in a 3U form-factor. I think they can, but it will require some significant engineering changes. Cooling is a factor of course, but the G5 generates less heat than it's XEON Blade competitors. A fan right on the processor will cool it and allow it to be in a smaller form-factor. Since the servers are in a headless environment without the need for expansion slots a lot of economy can be achieved. This goes for storage as well, Apple Firewire technology can be used to share a common disk subsystem that would serve each blade its OS X image for start up. Connectivity to an xServe RAID box over a SAN switch would provide all the storage any blade would need and leveraging SAN switching would allow for unlimited dynamic storage expansion. Communications between the blades can be facilitated over Apple's proprietary low latency FireWire Extreme (800, 1600, 3200) IP connectivity, going to a 1GB to 10 GB Ethernet port to the rest of the Ethernet world.

It would make for a truly innovative, super high performance Server system at Apple's normally reasonable prices. And, it would be truly an awesome server system.

timothyjoelwrig

Oct 22, 2003, 04:30 PM

Originally posted by Jeff Harrell
That's not what the article says. It says they "cooperated," meaning they probably provided equipment to use for testing.

Blue-sky projects are really neat. But it's a long way from three guys at Stanford to your desk, and a lot can happen along the way to derail the operation.

I didn't specify that my information came from that article alone. Cooligy has made it clear that they will be supplying qualification systems to participating computer systems developers and manufacturers by Q4. Obviously this isn't set in stone, but the concept and development have been underway for quite sometime, and it seems plausible that this is possible solution.

Also, of similar note, Hitachi has water-cooled notebook computers already on the retail market, and a standard water-cooling kit can be purchased for about $99 retail. Granted, that's a bit more than the $.30 all the cooling fans would cost OEM, but I would think it would be worth it for the noise factor.

bankshot

Oct 22, 2003, 04:52 PM

Well, crud.

We're looking at getting an Xserve RAID and accompanying server in the next couple of weeks. I was really hoping that G5 Xserves would be right around the corner -- in that case, the choice of server machine is obvious. But this article makes it sound like Apple is still in the design/prototype phase, which means a G5 Xserve is probably still several months out.

Right now, I would have to choose between a G5 desktop or a G4 Xserve. The G5 would be cool, but since it'll be controlling the Xserve RAID, it can't exactly sit in someone's office as their workstation. Which means it'll probably go largely unused except for remote command-line processing of data. Seems like a bit of a waste.

The Xserve of course is more of a server and geared for this application, plus it comes with Panther Server, which may turn out to be a nice bonus. But it's only G4, with the lower system bus speeds... Darn! What to do, what to do?? :confused:

I suppose it's not the worst dilemma to have, though... :rolleyes:

ddtlm

Oct 22, 2003, 06:23 PM

People worrying about G5's being too hot for a 1U are just not thinking this through very well. Its not a laptop, nor a desktop, its a server. If it's loud, ugly, and hot to the touch noone really cares. (Not to mention, the G5 really isn't breaking any new ground as far as massive heat production goes.)

ffakr:

Even the 2.0 GHz G5s are quite a bit cooler than Athlons or Pentium 4s
Its not clear if this is true at all. To my knowledge, neither IBM nor Apple has ever said what the maximum heat output of a G5 is. They've stated "typical" numbers, but thats stupid because who knows what "typical" use is. Intel/AMD usually declare the max output of all their products, and its these figures that invariably end up being compared to Moto/IBM's vague "typical" figures. So at this point it is really not known if the G5 uses less power than, for example, an Opteron.

stingerman:

I believe the G5 has 3 coherent Processor Interconnects.
You seem to be thinking on the Opteron, which scales more smoothly than the G5 does. Each G5 connects to the system controller, not to other G5's. An 8-way G5 would require an insane system controller, or a series of them linked together. Neither option is cheap.

manitoubalck

Oct 22, 2003, 07:16 PM

He's right; AMD Opteons can easily be configured into 8-way systems because of the Hyper-Transport technology between each Processor and each processor's individual RAM. The G5 architecture requires that each processor must first communicate with the NorthBridge then to the communal system memory or another processor. Not so in the AMD architecture, since many of the traditional NorthBridge functions are undertaken on the CPU itself. Resulting in less bottlenecking at the NorthBridge itself.

He's an actual picture of a 4-way Opteron system, showcasing the individual RAM of each processor

Scottgfx

Oct 23, 2003, 02:07 AM

Originally posted by manitoubalck
You guys have appeared to miss the point a bit. It would make sense for the use of a water-cooling solution if the heat were that much of an issue. Plus the water-cooling systems are much smaller and quieter than the air/air units not including the external radiator of corse. This relates to smaller cases, even a 1U since water units can be made of almost any size, and the cooling efficiency remains about the same.

Who could deny the silence and cooling potential, not to mention the performance increase form one of these babies?

What the heck is that thing, a Jarvik-7? :)

manitoubalck

Oct 23, 2003, 06:45 AM

Originally posted by Scottgfx
What the heck is that thing, a Jarvik-7? :)

That thing as you put it is a water cooler for a CPU. it's quiet, it's small, it costs a little more, and cools a hell of a lot better than an air/air system. Used mainly in 'extereme' overclocking situation, but since you probabley own a mac you would know about such things:rolleyes:

Rincewind42

Oct 23, 2003, 10:43 AM

Originally posted by ColdZero
There is 3.2gbps availible on the HT link in the G5. Thats gonna get filled up very quickly. You have 2 gigabit ethernet interfaces. 2 PCI-X slots that are going to take 2 gbps together. You're already over 3 gbps (assuming the 2nd ethernet is in a pci-x). 3.2 gbps is fast, but it could get overloaded in a server.

Unless Apple's G5 technote is wrong, the HT bus is 3.2 GBYTES/s with the lower half being 1.6 GBYTES/s. So multiply by 8 before comparing to other technologies on the bus. Therefore, you have 25.6 Gbps to the PCI-X slots, leaving 12.8Gbps (PCI-X is 1GBps, not 1Gbps, for 133Mhz slots. For 100Mhz slots each card can use up to 800MBps, not certain if that is aggregate or if both can pull that much, my numbers are for each pulling that much). 12.8 Gbps happens to be just as much as we get from the lower half of the HT bus (1.6GBps). This is distributed as 1.5Gbps for each SATA drive, 1Gbps for each Ethernet, 800Mbps for each FW, and 240Mbps for each USB. Assuming 2 each of everything and 4 each of SATA that gives us 6+2+1.6+.48, which is about 10.1 Gbps - well under the 12.6 Gbps limit of the lower half of the HT bus.

As for 4 proc 2u machines, think again. I have never seen a 4 proc 2u machine before. Especially one that needs so much cooling like the g5. I think a 1u dual processor 1.8ghz g5 is where it's at. There will be no quad configs in 2u.

First, the G5 doesn't need as much cooling as the PowerMac G5 implies. Two of them in the Dual 2 config consume only about 25% more power than a single high-end Pentium 4. (95ish watts in the G5 vs 80ish watts in the P4). do you see any P4 systems with those massive heatsinks? No! Why? Because those systems don't give a damn about how loud their fans are (remember - Apple's last G4 machine was nicknamed Windtunnel - I don't think they want to see another nick name like that...). So It is not cooling that would be keeping the G5s out of an Xserve, although it is probably an engineering issue in a 1U form factor. A 2U factor allows them to alleviate the cooling issue (they can use larger heatsinks) and gives them more room for CPUs too.

Now, I don't actually expect to see a Quad Xserve (although you could push me over with a feather if Apple made one). But I also don't see a 2U Xserve coming. At worse, I see the 1U G5 Xserve being pushed back until the 90nm 970s are available. But if Apple did decide to go 2U, then I think that they would have to go 4xCPU - it's one of the biggest advantages the Xserve has - density of CPUs. And with the VT cluster scoring 8.1 Tflops on 1056 nodes (no idea why they are missing 44...) for Apple to reduce their CPU density would be madness.

ColdZero

Oct 23, 2003, 11:48 AM

Sorry, but there is not the physical room for 4 processors in 2u. Servers that have 4 processors have memory riser cards and are in larger cases for a reason. In 2u there is not the room for one of these riser cards vertically. You would greatly limit the amount of ram you could fit in them. It would be a choice of RAM, Processors or Expansion. You can't have all 3 in a 2u enclosure. Thats why you don't see any quad 2u's. Apple 1u servers are just as processor dense as any other manufacturer who makes dual 1u boxes. More so, the Xserve, in its current form, is not comparable to the horsepower in 1u servers on the x86 side. Its a very attractive bundle, and I use one at work, but if you need a lot of proc power, you look elsewhere.

The VT cluster achieved 7.41TF with 1100 nodes. And holds the #4 position in the world, when the lists get updated the next time around.

ddtlm

Oct 23, 2003, 12:42 PM

Rincewind42:

Two of them in the Dual 2 config consume only about 25% more power than a single high-end Pentium 4. (95ish watts in the G5 vs 80ish watts in the P4).
I'm pretty sure if you look into your source for that info you'll find that it is unofficial. The version I've heard suggests that 90W is the max output for each 2.0ghz G5. That's perhaps not unreasonable if the "typical" output for a 1.8ghz model is 47W. It may also be the case that 90W is the maximum for any G5, and is what Apple designed to in order to handle faster speeds. Who knows... clearly noone here does.

ColdZero:

Sorry, but there is not the physical room for 4 processors in 2u.
There's definately room, just little market for it. It would be an over $10k 2U box with few mounts for hard disks or anything else, cause it would be all full of heatsink and power supply, not to mention the mobo having to cover a large amount of the floorspace.

Servers that have 4 processors have memory riser cards and are in larger cases for a reason.
Larger cases yes, riser cards no. There are 4-CPU Opteron rackmount boards without riser cards. What was that company... Newisys? Someone was making such a system, anyway.

ColdZero

Oct 23, 2003, 01:29 PM

I'm not saying there isn't enough room, I'm saying that there isn't enough room for all the components of a general use server like the Xserve. The 4 way opterons you speak of need to have their drives stacked on top of eachother to fit 4 in there. The motherboard is physically too big to have in a 1u chassis with a powersupply, 4 drives, pci slots and such. I don't have measurements, but from pictures I've seen of a 4way opteron board, I don' think there is enough width in a 1u enclosure to fit even a powersupply besides the mobo and processors.

4 Way Opteron (http://colfax-intl.com/jlrid/SpotLight_more.asp?L=71&S=1&B=624)

Also the opteron has hypertransport links to each other processors memory. The g5 does not share memory like this. If that board had 4dimm slots per processor, a g5 could only access 4gb of memory using 1gb dimms. Where on the opteron each processor could access the whole 16gb of the server through the hypertransport links.

Rower_CPU

Oct 23, 2003, 01:32 PM

Originally posted by manitoubalck
That thing as you put it is a water cooler for a CPU. it's quiet, it's small, it costs a little more, and cools a hell of a lot better than an air/air system. Used mainly in 'extereme' overclocking situation, but since you probabley own a mac you would know about such things:rolleyes:

Nor would most PC users, since they just buy systems from Dell/HP/etc and don't get all "extereme".

:rolleyes:

Rincewind42

Oct 23, 2003, 03:24 PM

Originally posted by ColdZero
The VT cluster achieved 7.41TF with 1100 nodes. And holds the #4 position in the world, when the lists get updated the next time around.

Check again, the numbers have been updated. NYT had preliminary numbers. The latest is here (http://www.netlib.org/benchmark/performance.pdf) and show 8.164 Tflops.

manitoubalck

Oct 23, 2003, 06:13 PM

Originally posted by Rower_CPU
Nor would most PC users, since they just buy systems from Dell/HP/etc and don't get all "extereme".

:rolleyes:

Hello Rower_CPU, If you've every looked passed the pictures in my posts you would know that I dislike DELL/HP/etc... for the same reasons I don't see apple as an economic computing choice. I built my PC from the parts I selected because they were the best for the I was able to spend($2200AUD.) I have had problems yes, but the defective part was swaped on the spot no questions asked. I could have bought a mac or a prebuilt PC for DELL/HP but I would have had to pay an extra $600 dollers in order to get a similar machine.

And more on the topic a water cooling solution all be it expencive would be the ideal solution, even you have to agree to that:)

Sun Baked

Oct 23, 2003, 06:24 PM

Originally posted by manitoubalck
And more on the topic a water cooling solution all be it expencive would be the ideal solution, even you have to agree to that:) The ideal solutions are ones that don't require trained monkeys to accomplish, something that an intern couldn't even screw up.

Water cooling and/or large heat sinks that require a screwdriver and thermal paste are less than ideal.

Rower_CPU

Oct 23, 2003, 06:29 PM

Originally posted by manitoubalck
Hello Rower_CPU, If you've every looked passed the pictures in my posts you would know that I dislike DELL/HP/etc... for the same reasons I don't see apple as an economic computing choice. I built my PC from the parts I selected because they were the best for the I was able to spend($2200AUD.) I have had problems yes, but the defective part was swaped on the spot no questions asked. I could have bought a mac or a prebuilt PC for DELL/HP but I would have had to pay an extra $600 dollers in order to get a similar machine.

And more on the topic a water cooling solution all be it expencive would be the ideal solution, even you have to agree to that:)

That still doesn't invalidate my point that it's ridiculous to say a "typical Mac user" doesn't know about these things because typical computer users don't know about these things.

Would your average PC user be able to identify that picture as a water cooler? Don't think so.

manitoubalck

Oct 23, 2003, 06:33 PM

Originally posted by ColdZero
Also the opteron has hypertransport links to each other processors memory. The g5 does not share memory like this. If that board had 4dimm slots per processor, a g5 could only access 4gb of memory using 1gb dimms. Where on the opteron each processor could access the whole 16gb of the server through the hypertransport links.

I like the point you make and agree. The pic of the 4-way opteron board I posted earlier looks too large for a 1u inclusure but it was posted to show 1 example of a 4-way opteron system.

manitoubalck

Oct 23, 2003, 06:44 PM

Originally posted by Sun Baked
The ideal solutions are ones that don't require trained monkeys to accomplish, something that an intern couldn't even screw up.

Water cooling and/or large heat sinks that require a screwdriver and thermal paste are less than ideal.

1. People who build servers know what they are doing, especially mac technicians.
2. Thermal paste is now commonplace on all aftermarket and most stock CPU coolers today.
3. AMD recommend the use of a screwdriver to attach the stock AMD 2500+ heat sink to the CPU and motherboard. (I know because I installed one on the weekend)
4. I have no formal training and have built several PC including my own with no trouble.
5. Any company considering buying expensive servers could probably afford to keep a computer tech on the pay roll.

Sun Baked

Oct 23, 2003, 07:13 PM

OK, you've just named some things that would be in a market that the XServe isn't aiming for.

Sure it may be powerful enough for a corporate cluster, but it supposed to be simple enough for a novice to use.

Which has been one of the selling points of Mac OS Server for years.

The XServe was the first piece of hardware that brought ease of repair to the novice, before it would have taken that trained Mac tech to do the repair -- or sending the unit out for service.

Now most every part should take less than 10 minutes to replace, including showing the intern how to do it.

And no thermal paste.

manitoubalck

Oct 23, 2003, 07:41 PM

You make a strong point, but anyone with the "need" for a server is going to have some substansial computing knowlege.

And what do you have against thermal paste. AMD Heatsinks come standard with it because it results in better cooling

ddtlm

Oct 23, 2003, 07:55 PM

ColdZero:

If 2 Opterons and everything fit fine in 1U, I fully expect that 4 Opterons and everything can fit in 2U, just not with many disks in there, and with only one or two PCI slots. For example, here's Newisys's quad 4U Opteron:

http://www.newisys.com/products/4300_specifications.html

In that pic, you can see that a lot of volume is lost to the 6 full length PCI slots and the additional half-length PCI slot. The case also holds 5 or 6 hot-swap SCSI disks. I think you'll agree that if they cut this down to perhaps one PCI slot and 3 or 4 disks they'd be in 3U territory at least. I still think 2U, but I don't think much can be settled on that matter.

Also the opteron has hypertransport links to each other processors memory. The g5 does not share memory like this. If that board had 4dimm slots per processor, a g5 could only access 4gb of memory using 1gb dimms. Where on the opteron each processor could access the whole 16gb of the server through the hypertransport links.
I think your a bit confused here. Yes Opterons use hypertransport to link chips to other chips, so that they can share memory. However, a quad G5 would also need to share memory, or else its not a quad, its separate machines. It would do this by either having one mammoth system controller with 4 FSB's and 4 or more channels of RAM, or it would link multiple smaller system controllers together with something like hypertransport.

ColdZero

Oct 23, 2003, 08:45 PM

Originally posted by ddtlm
I think your a bit confused here. Yes Opterons use hypertransport to link chips to other chips, so that they can share memory. However, a quad G5 would also need to share memory, or else its not a quad, its separate machines. It would do this by either having one mammoth system controller with 4 FSB's and 4 or more channels of RAM, or it would link multiple smaller system controllers together with something like hypertransport. [/B]

G5s do not share memory. They have seperate interfaces to their individual bank of memory. Thats why you need to install it in pairs (in dual machines) rather than dimm by dimm. Just because they don't share memory doesn't mean they are seperate machines. They each have a FSB link to the chipset, from there they connect to the I/O devices. The dual G5 does this same thing, just look at the diagram on Apple's site. Memory is not shared between processors. All a quad g5 is is two more processors off the chipset. The processors then interface with the memory directly, opteron style. They don't share their memory, you need them in each bank.

Sun Baked

Oct 23, 2003, 08:52 PM

Originally posted by ColdZero
G5s do not share memory. They have seperate interfaces to their individual bank of memory. Thats why you need to install it in pairs (in dual machines) rather than dimm by dimm. Just because they don't share memory doesn't mean they are seperate machines. They each have a FSB link to the chipset, from there they connect to the I/O devices. The dual G5 does this same thing, just look at the diagram on Apple's site. Memory is not shared between processors. All a quad g5 is is two more processors off the chipset. The processors then interface with the memory directly, opteron style. They don't share their memory, you need them in each bank. LOL...

The memory is shared.

The memory controller grabs the memory in 128-bit chunks to build the block of memory it's fetching, since each DIMM is 64-bits. It's grabbing memory chunks across two DIMMs at once.

G5 developer notes (http://developer.apple.com/documentation/Hardware/Developer_Notes/Macintosh_CPUs-G5/PowerMacG5/index.html)

Phil Of Mac

Oct 23, 2003, 08:57 PM

manitoubalck

Oct 23, 2003, 09:17 PM

Originally posted by ColdZero
G5s do not share memory. They have separate interfaces to their individual bank of memory. That’s why you need to install it in pairs (in dual machines) rather than dimm by dimm.

You might want to read these threads I started "A shift to Serial" and "G5 memory: Dual Channel or not " The G5 runs what is knows as 'Dual-channel' DDR. It was developed by Nvidia for use with the AMD Athlon XP but has seen wide adoption throughout the computing world due to the fall of RAMBUS. Unlike windows based systems (where it is not essential to use both channels,) the G5 must use both channels, hence installing RAM in Pairs.
The DDR memory modules in the Dual 2.0GHz G5 run @ 400MHz, but the dual channel system allows a total of 800MHz, 400MHz per channel. Hence the memory is shared between the 2 processors

The Opetrons however each have their own bank of Dual Channel DDR. (See the pic on page 2 of this post)

ffakr

Oct 23, 2003, 11:34 PM

Originally posted by manitoubalck
You might want to read these threads I started "A shift to Serial" and "G5 memory: Dual Channel or not " The G5 runs what is knows as 'Dual-channel' DDR. It was developed by Nvidia for use with the AMD Athlon XP but has seen wide adoption throughout the computing world due to the fall of RAMBUS. Unlike windows based systems (where it is not essential to use both channels,) the G5 must use both channels, hence installing RAM in Pairs.
The DDR memory modules in the Dual 2.0GHz G5 run @ 400MHz, but the dual channel system allows a total of 800MHz, 400MHz per channel. Hence the memory is shared between the 2 processors

The Opetrons however each have their own bank of Dual Channel DDR. (See the pic on page 2 of this post)
You don't quite have that right.
Dual channel wasn't invented by Nvidia. Machines especially server have used dual bank, quad bank, octal bank memory for ages. Pentiums and early PowerPCs needed SIMMs in pairs because a SIMM is 32 bits wide and the memory controller was 64bits wide. Nvidia was first to market with Dual-channel DDR, but that's hardly an 'invention', just a copy of what everyone else has done for every other ram technology used (RAMBUS, SDRAM, EDO, FPM...)
Also dual channel DDR 400 does not cause the system to run at 800MHz, it causes the 400MHz memory to operate as a 128bit wide memory bank. This is very different.

manitoubalck

Oct 24, 2003, 12:15 AM

Originally posted by ffakr
Nvidia was first to market with Dual-channel DDR, but that's hardly an 'invention', just a copy of what everyone else has done for every other ram technology used (RAMBUS, SDRAM, EDO, FPM...)
Also dual channel DDR 400 does not cause the system to run at 800MHz, it causes the 400MHz memory to operate as a 128bit wide memory bank. This is very different.

'This effectively doubles the bandwidth, enabling the Power Mac G5 to reach a maximum memory throughput of up to 6.4GB per second' for www.apple.com

OK, DDR400 opperates @3.2GB/s, while what you said is true about allowing the memory to operate as a 128bit wide bank, the fact that the memory throughput is 6.4BG/s shows that the dual channel system operates at an effective 800MHz

Rocketman

Oct 24, 2003, 12:39 PM

Originally posted by rsnyder@psu.edu
The story focuses on the Xgrid project mailing list that slipped out. Personally, I am wondering if the xGrid has anything to do with Oracle's 10g (as in "Grid" computing). The word at Oracle World was that Oracle will release a full OS X version with the release of 10g. There was also speculation that this would require the G5.

Since both Oracle and now Apple are talking grid computing . . . . would be sweat to run Oracle on OS X now that Sun is in dumper.

I am holding two servers in my budget waiting for the G5 xServes. My goal is in three years to get off Sun and its overpriced hardware. The Apple reps know that there are more than a few like me holding off spending until the G5 xServe arrives.

The xgrid mailing list has alot of drama right now. People are being booted from it, and one guy is posting to it claiming to be a G5XS beta tester.

I just wanted to get some ideas for topologies for existing/any CPU's and that was not forthcoming on that list. The fiolks there are basicly trying to get advance info on product announcements, not adding to the discourse.

Even questions I answered resulted in punishment.

Caution.

Rocketman

ffakr

Oct 24, 2003, 04:08 PM

Originally posted by manitoubalck
'This effectively doubles the bandwidth, enabling the Power Mac G5 to reach a maximum memory throughput of up to 6.4GB per second' for www.apple.com

OK, DDR400 opperates @3.2GB/s, while what you said is true about allowing the memory to operate as a 128bit wide bank, the fact that the memory throughput is 6.4BG/s shows that the dual channel system operates at an effective 800MHz

Actually no, it still doesn't run at an effective 800MHz. You can't access dual channel DDR 400 at 800Million times a second. 800MHz would likely lower the latency as there would be more cycles per time unit.
It makes the memory effectively 128bit wide which is very different.

I'm sorry to be a stickler, but this isn't apples and oranges, it's apples and puppies (quite different)

MacWhispers

Oct 24, 2003, 05:26 PM

Nearly every substantive decision coming from Cupertino these days is marketing based, not engineering based. It must be that way, as Apple is in high gear growth mode, and is playing some of the most aggressive hardball ever seen in the computer industry. My point?

There is zero chance that Apple will walk away from the slow but steady success they have had in the 1RU server space. In fact, if anything is being tinkered with in the Apple R&D labs, other than a 1RU form factor, it would more like be smaller/denser, not larger.

Steve Jobs and Team are motivated these days by the idea of pushing the envelope, and just knocking the socks off of people with any new product released. So, what would do that in the next Xserve release?

I expect at least dual G5s in the next 1RU Xserve. I would just smile to see a quad G5 1RU Xserve announced. Not possible, someone says? Tell that to the guys being paid the big bucks at Apple to work 20-hour days in the lab chasing ideas that are "not possible." Don't bet against them.

Phil Of Mac

Oct 24, 2003, 06:06 PM

Originally posted by MacWhispers
Not possible, someone says? Tell that to the guys being paid the big bucks at Apple to work 20-hour days in the lab chasing ideas that are "not possible." Don't bet against them.

20 hours a day leaves them just 3 hours a day to sleep, accounting for 1 hour commuting to and from home and eating. The human body cannot endure that. A 20 hour workday would be less productive than a shorter workday for that reason.

But, what do you expect from this guy?

ffakr

Oct 25, 2003, 12:32 AM

Originally posted by Phil Of Mac
20 hours a day leaves them just 3 hours a day to sleep, accounting for 1 hour commuting to and from home and eating. The human body cannot endure that. A 20 hour workday would be less productive than a shorter workday for that reason.

But, what do you expect from this guy?
I think this is just the extension of the old 'Steve Jobs the Tyrant' days of Apple stories.
I guarantee you that people don't make a habit of working 20 hour days at Apple just because SJ is in charge again. I'm sure it happens when there's a deadline though... I've put in a few longer than that (in IT positions).

Phil Of Mac

Oct 25, 2003, 12:34 AM

Originally posted by ffakr
I think this is just the extension of the old 'Steve Jobs the Tyrant' days of Apple stories.
I guarantee you that people don't make a habit of working 20 hour days at Apple just because SJ is in charge again. I'm sure it happens when there's a deadline though... I've put in a few longer than that (in IT positions).

Sure, but not on a regular basis!

manitoubalck

Oct 26, 2003, 05:41 AM

Originally posted by ffakr
Actually no, it still doesn't run at an effective 800MHz. You can't access dual channel DDR 400 at 800Million times a second. 800MHz would likely lower the latency as there would be more cycles per time unit.
It makes the memory effectively 128bit wide which is very different.

I'm sorry to be a stickler, but this isn't apples and oranges, it's apples and puppies (quite different)

Ok sorry for flogging the dead horse but this is a direct quote from www.apple.com/powermac/architecture.html

"400MHz dual-channel memory
The new Power Mac G5’s memory controller supports fast 400MHz, 128-bit DDR SDRAM, and a dual-channel interface enables main memory to address two banks of SDRAM at a time, reading and writing on both the rising and falling edge of each clock cycle. This effectively doubles the bandwidth, enabling the Power Mac G5 to reach a maximum memory throughput of up to 6.4GB per second"

So in essence were both right, DDR400 has a bandwidth of 3.2GB's (hense its other name DDR3200,) the max memory thoughput is double that (6.4GB/s.) Therefore it is also 128-bit DDR.

I would still like to see an apple chipset that used rambus which which is up to a frequency of 1333MHz, effective thoughput upto 10.7GB/s. (I know that the cost and RAMBUS's inablity to see eye to eye with JDec have been it's downfall.) These specs are taken from www.rambus.com

ffakr

Oct 26, 2003, 12:13 PM

Originally posted by manitoubalck
So in essence were both right, DDR400 has a bandwidth of 3.2GB's (hense its other name DDR3200,) the max memory thoughput is double that (6.4GB/s.) Therefore it is also 128-bit DDR. I'm still not sure why that makes us both right... just because the other poster came to the same 6.4GB by miss-understanding how the memory controller works. ;-)

I would still like to see an apple chipset that used rambus which which is up to a frequency of 1333MHz, effective thoughput upto 10.7GB/s. (I know that the cost and RAMBUS's inablity to see eye to eye with JDec have been it's downfall.) These specs are taken from www.rambus.com
No one sells (or really makes) RDRAM over 1066MHz right now. Apple would be pushing the envelope if they went 1066MHz, let alone 1333MHz.
Not only this but since Rambus is even more rare these days, it's suffering worse economies of scale. Newegg.com has 512 MB of RDRAM 1066 for $235. The same ammount of DDR 400 runs around $80, 3x less.
At this markup, if you wanted to fill a G5 with 8GB of Mushkin RAM from newegg, it would cost around $1850. At 3x more, it would cost around $5550 to do the same. This is a huge issue.
Newegg doesn't even offer 1GB RIMMs. Given the size of newegg, I'd say this indicates a potential availability problems too.
RAMBUS is being produced in 256Mbit densities while DDR is being made in 512Mbit densities and companies are preparing 1Gbit DDR-II chips. You'd need 4x as many physical chips to get the same amount of memory on RDRAM as you'd need for upcomming DDR-II. Even with economies of scale, RDRAM would never cost as much as DDR unless the manufacturers could make up a lot of lost research time and bridge that density gap.
RDRAM also has a higher latency than DDR. The Opteron performs so well partially because it has an on-die memory controller that lowers the latency of memory access. I'm not sure it's a good thing to give Opteron more of an advantage in this area.
Also, don't forget that RIMMs are 16bit wide. One PC800 RIMM has as much bandwidth as one DDR 400 DIMM. RDRAM 1066 may sound way faster than DDR 400, but 2 RDRAM 1066 RIMMS actually provide the same bandwidth as dual channel DDR 533 (Apple's using dual DDR 400).

RDRAM has one advantage over DDR. It has a higher frequency.
It has several disadvantages: cost, density, latency, availability, width of data pipes.

The other thing I really really hate about rambus is their marketing. They love to talk up unreleased parts and make them sound like they are actually available.
The link off the main page says RDRAM is available from 800MHz to 1600MHz, but i you click on it, you get a press release stating that speeds from 800 to 1200MHz are available with 256Mbit densities. Now go into the real world and try to find anyone selling anything other than 800 and 1066MHz RDRAM.
Look into the site more and you find tons of stories about super high speed RDRAM, overclocked RDRAM, higher density chips... all presented as if this technology were available today. The majority of the site is dedicated to vaporware.

Imagine if you went to Apple's web site and the G5 page boasted about PPC 970s available from 1.6GHz to 3GHz.

CMillerERAU

Oct 26, 2003, 07:15 PM

Originally posted by manitoubalck
And what do you have against thermal paste. AMD Heatsinks come standard with it because it results in better cooling

and it goes great on bagels!

Sun Baked

Oct 26, 2003, 07:29 PM

Originally posted by CMillerERAU
and it goes great on bagels! And it'll void your warranty if Apple catches you rubbing your bagels on the heatsink.

Only the Apple Certified Techs are allowed to do that. ;)

manitoubalck

Oct 26, 2003, 07:47 PM

I one year when all the warrenty's are over it maybe interesting to see what the overclocking potential of the G5 is, or what preformance gains can be aquired by upgrading the cooling solution. Also potential case upgrades with the upcomming BTX form factor(maybe a second optical bay or a 3.1/4" external bay for a zip drive, etc...)

The comment about RAMBUS is is just me wanting to see how the G5 would run on serial RAM. I am well aware of the companies downfalls:)

ffakr

Oct 26, 2003, 09:39 PM

Originally posted by manitoubalck
I one year when all the warrenty's are over it maybe interesting to see what the overclocking potential of the G5 is, or what preformance gains can be aquired by upgrading the cooling solution. Also potential case upgrades with the upcomming BTX form factor(maybe a second optical bay or a 3.1/4" external bay for a zip drive, etc...)

The comment about RAMBUS is is just me wanting to see how the G5 would run on serial RAM. I am well aware of the companies downfalls:)

I'd prefer to see what type of performace we would see with an on-die memory controller and good old DDR. I wonder what types of problems the engineers would run into if they had on-die memory controllers with shared memory in SMP configurations. I'd hate to trade an on board mem controller down the line for the benefits of having one large memory space. :-)

Anyone post info on how to adjust the clock speed on these things yet?

manitoubalck

Oct 27, 2003, 04:05 AM

Originally posted by ffakr
I'd prefer to see what type of performace we would see with an on-die memory controller and good old DDR. I wonder what types of problems the engineers would run into if they had on-die memory controllers.

On die memory controlers already exist, check out the line of AMD 64's and Opterons.

ddtlm

Oct 27, 2003, 01:29 PM

manitoubalck:

...and various Sun products had on-die controllers before AMD. (Not sure if anyone beat Sun to it.)

Sun Baked

Oct 27, 2003, 02:30 PM

Originally posted by manitoubalck
On die memory controlers already exist, check out the line of AMD 64's and Opterons. There are also some PowerPCs with on die memory controllers, RIO ports, TCPIP acceleration engines, etc.

The tech is there for the PPC line, it just hasn't been added to anything Apple would use as a CPU.

rjstanford

Oct 29, 2003, 05:18 PM

Originally posted by ColdZero
Sorry, but there is not the physical room for 4 processors in 2u. ... Apple 1u servers are just as processor dense as any other manufacturer who makes dual 1u boxes.Hmm. So in 2U you can have two separate dual-processor machines (with redundant power supplies, four network ports, duplicate fiber channel cards, two graphic cards, et cetera) but if you try to remove the middle piece of metal and delete the extra network and graphics cards, it won't fit? Or would get too hot? 'Scuse me?

Personally, I'd love to see a 2U 4-way that did have redundant power supplies. Dual graphics boards I could live without, though... Still, I wouldn't expect this to replace the 1U two-way box, just add to it as another machine option. That would make a fantastic midrange server, if it could come in somewhere around the $7,500 mark. In my field, there's substantial need for boxen around the 4-modern-processor size, with rapidly diminishing need above and below that point (the exceptions tending to cluster around a 32-way level).

That would put a good-sized database server with, say 1TB of usable space on a dedicated XServe RAID available well under $20,000 - even after AppleCare and lots of extra memory. And that's a very appealing price point. Using embedded licensing, that's still under $30,000 including necessary Oracle costs. That's a good deal no matter how you slice it.

-Richard

pgwalsh

Oct 29, 2003, 07:20 PM

Maybe the 4U Rack is going to house 1 or 2 Power 5's... ;)

Opteron

Feb 27, 2004, 03:57 AM

Bump.

It's interesting to see how opinions on water cooling change over time.

ffakr

Feb 27, 2004, 12:07 PM

Originally posted by manitoubalck
On die memory controlers already exist, check out the line of AMD 64's and Opterons.

Yes, but I was referring specifically to the PPC 970, as in 'I wonder exactly how much a 970 would benefit from an on die mem controller with current memory technology'.

I'm Back

Jun 24, 2004, 10:01 AM

While this thead's title topic has been answered long long ago. The second topic that imerged about water cooling has only just reciently been answered by apple.

Just found it interesting reading over posts from page two onward, in light of recient developments.

ddtlm

Jun 25, 2004, 11:42 PM

Now I must say I feel a bit dirty replying to a thread this old, not to mention a thread where the last poster appears to have been banned, but I think that this quad Opteron in 1U is interesting:

http://www.appro.com/product/server_1142h.asp

Certainly settles the arguement about how much can be stuffed into what space.

u07ch

Jun 26, 2004, 08:27 AM

Lets piece all the pictures together; and then add some views from the enterprise front line. People are focussing on render farms and video on this multiprocessing dreamfest. The majority of high end servers of this world arent doing video manipulation; they are running the world; high end database servers cost companies 50k at the entry level. Now if only there were a large scalable high end database for os/x - oh wait oracle are likely to launch one at wwdc; maybe we have a high end server coming for running ERP and other large databae tools. A 1u server will not be able to cope with that demand; and clustering is not as efficient as smp (at least not in sql worlds.) The performance benefit of the processor should make apps run quicker especially if oracle do 64bit) the kit is a nice alternative to the IBM black i see day in day out too ... What sort of kit am i talking about though

The Wintel server I specced for my server room was

IBM X365 4 x xeon 3.2ghz, 16gb ram, fibre channel, 4 x 36 15k scsi raid 10 for operating system

IBM Fastt SAN + 120 x 76.4 gb 15k disks raid 10

which was built as a failover cluster ( so two of these less the san) running sql server enterprise. The total cost was over 100k before i looked at app servers (and paying for my pieces of paper to run the server - damned licensing.

I am slightly locked into sql server but not greatly an oracle system running on apple kit would be of great interest if it were cheaper than wintel. I have a concern of the performance of the IBM san's as they are on sata disks not scsi; this is partly fear of the unknown; but also knowledge that SCSI disk lives are pretty much written in stone; and the disks themselves are way quicker, have more cache.

The caveat is for my data centre is this: is the support is there? Support is something IBM/ compaq et al are fantastic at; i have had disks arrive in a taxi at 2am (followed not too distantly by an engineer) to meet their support agreement; and it only costs 7-800 pounds for 3 years. Apple support is europe is pathetic (mail everything back to europe type of support); which immediately turns me off the idea.

Maybe an apple database server is a bad idea on reflection ...