Apple's New Mac Pro Begins Showing Up in Benchmarks

[subsonix]

You watch a demo where no other software that exists or other computers are compared and find that it proves something? It's a bar graph with only one bar.

My List said "PCIE 3.0 with a real GPU"

Guess what? the consensus is that another Apple compromise was PCIE 2.0 on the GPUs. Might even be at 8 lanes. So bus speed cut in half or 1/4 so that all those nifty TB controllers have lanes.

Consensus among idiots, who cares? It's either fact or it's not, since no such details has been published, it's not.

It's pretty obvious that your list said that, but your needs are clearly different from actual users of these systems. If you can sell graphics cards your objective is met right?

 

[pscl]

1 & 2,
I don't get all the great comments about it being lighter and smaller? It's a damn desktop. You are not meant to be carrying around in your bag like a laptop. People are suddenly going to start taking a desktop on visiting clients? Packing all the external drives away and taken the as well. That'll look really professional, asking a client to wait 30 minutes while you plug everything. Horsehocky.

3,
Looks better? It's a workstation computer, not a Ferrari. Looks only matter to the sad lot that wont use it to make money.

4, After 4 years to work on this they give a mediocre increase in raw processing power. The GPU's are great but unless you run stuff that is optimised for them they are nothing more than heaters. We should have been seeing motherboards with 4 slots in these machines not a reduction down to 1.

5, Good expandability? HA HA. If you want to double the spend on the initial machine to get a similar set up to what you can already get inside the current systems. As for CopperCripple, don't believe the carp that is said. It was a joke on release, going from optic to copper.

6, Again it's a damn desktop, doesn't need to be carried around. The only time you need to move it is when you take it out of the box and plonk it on the desk.

Or is everyone going to get a plinth and bow down to it every morning?

Totally Agree! Only thing that would lead me to a MacPro is the "Desktop" feeling, and the expandability especialy the GPU power (not a "laptop gpu" like in the imacs). The new macpro to me is a big step back in history of the macpro.

i will go for the new maxed out haswell imac this fall. even if it would be the same price as the entry level macpro, i will get a gorgeous cinema display on top.

 

[thunng8]

When is the last time you looked up information on Power/PowerPC? Your information is outdated. From what I understand they laid off the majority of their design and support people in the power group a week or so ago. They are also shopping around their Server group with Lenovo being the likely purchaser.

Intel's x86 tech isn't maxed out, they have just chosen to focus on power consumption/efficiency/performance per watt over raw power.

What are you talking about? IBM wants to sell their x86 server group. They are not contemplating selling their POWER server group.

In fact, IBM will unveil their latest POWER8 chip at Hotchips later this year:

http://www.pcworld.com/article/2042...new-unix-server-processors-at-conference.html

 

[MacVidCards]

You seem to give more credence to an easily faked bench mark report then a person who has actually used the machine. You also have a lot of faith in Internet consensus.

He said it was faster than anything he had taken out of a box.

Use some critical thought...why did he phrase it that way?

So, accepted benchmark vs. bar graph with one bar and a "best thing ever on a blue sunday" endorsement.

 

[Lesser Evets]

Geekbench is just such a useless benchmark.

Says nothing about x264 performance, 7-zip performance, OpenGL performance, OpenCL performance, LLVM performance and so on.

Partially. Almost.

I discovered how useless it was when I tried a MacPro 1,1 and new Mini with the same benchmark score (approximately). The MacPro was much better, much faster, much more enjoyable to work with in all ways except web browsing.

 

[audioteknika]

People still care about Sh*tassBench ?

Wow... This is 2013 people, wake up.

 

[kemal]

Somewhere deep inside Apple they're machining a triangular heatsink out of copper.

 

[Chupa Chupa]

A 11% speedbump compared to a 2010 machine, that should be way much more.

You have to talk to Intel about that. But it is 11% increase at a about 12% lower clock speed so it's not 100% directly comparable. Likely 20% increase over 2010 if it was clock speed for clock speed. While the 2010 MP tested was the high end, we don't know where this 2.7Ghz 12-core fits in to the new MP lineup. Has Intel publically released a complete spec sheet for its E5 v2 lineup yet?

 

[AnonMac50]

Image

Yum.

 

[TIFrank]

Only a minor boost from last year it seems.

 

[GroundLoop]

Personally, I am hoping to see what a new Mac Pro with the Xeon E5-2687W v2 (8-core, 3.4GHz) will do. I still run a lot of single/minimally threaded applications, so the higher core frequency will be of greater benefit for me. Then when I do need to run highly parallel apps, the 16 threads should be good enough for my needs.

Hope that the little Mac Pro can handle the 150W TDP though.

GL

 

[deconstruct60]

My List said "PCIE 3.0 with a real GPU"

Guess what? the consensus is that another Apple compromise was PCIE 2.0 on the GPUs. Might even be at 8 lanes. So bus speed cut in half or 1/4 so that all those nifty TB controllers have lanes.

Consensus by who????

It is easy to hook up the GPUs , Thunderbolt, PCI-e SSD without the GPUs on PCI-e v2.0

Xeon E5 (40 lanes at v3.0 )

x16 GPU + PCI-SSD switched v3.0 ( effectively a x14 and x2 split of bandwidth )
x16 GPU
x4 Thunderbolt controller ( limited to v2.0 )
x4 Thunderbolt controller ( iimited to v2.0 )

C602 chipset ( 8 lanes at v2.0 )

x4 Thunderbolt
x1 Ethernet
x1 Ethernet
x1 Wifi/Bluetooth
x1 USB 3.0


The GPU and SSD share PCI-e lanes above likely largely because that is where the PCI-e card is mounted. With a v3.0 PCI-e interface the SSD only really needs x2 for what would be (x4 lanes of v2.0 traffic.). Even if SSD is v2.0 (x4 ) a compensating switch would make it a 12 v3.0 for GPU card ( which is still equivalent to 24 PCI-e v2.0 worth of bandwidth.). If there is a problem with the switching due to the SSD being v2.0 that is the SSD's fault not TB.

In the above design, one GPU is full blast v3.0 no issues. The other is a trade-off in part driven by where to physically place the SSD so that it can be removed and serviced. Whether TB was there or not really isn't relevant to that. ( there could be one less TB controller and it would still be there is this size & format factor was a design objective. )

 

[haravikk]

How well does Geekbench measure OpenCL performance as a factor of overall system performance? I think that for the pro market that's the area that is going to make the biggest difference in the long run as two GPUs means you effectively have a triple processor machine, for the right applications.

Even so, it's a little disappointing to see only a small increase for the processor, as while it's the same core count overall I would have expected the other advances to make a much bigger difference. The smaller multi-core gains are likely to be a result of the slower overall clock speed (single-thread will be better thanks to a good turbo boost speed), and possibly having only four RAM slots rather than being able to feed data from eight at once.

If they do offer a version swapping a GPU for a CPU board then the machine would make massive gains, but I'm not sure how likely that is to be seen this generation (maybe the next model will have the choice of 2x CPU or 2x GPU).

 

[JarJarThomas]

Personally, I am hoping to see what a new Mac Pro with the Xeon E5-2687W v2 (8-core, 3.4GHz) will do. I still run a lot of single/minimally threaded applications, so the higher core frequency will be of greater benefit for me. Then when I do need to run highly parallel apps, the 16 threads should be good enough for my needs.

Hope that the little Mac Pro can handle the 150W TDP though.

GL

Well if you look at the cooling system it should easily handle that.
It has a far bigger surface than the traditional coolers, so for sure it can handle it. (Even if some here think that is wrong ... empty space in a desktop doesn't cool)

 

[ElderBrE]

The news should tell that Geekbench does not check the GPUs, which is a huge part on the design of this Mac Pro. They expect those GPUs to be used for computing power.

 

[757015]

1 & 2,
I don't get all the great comments about it being lighter and smaller? It's a damn desktop. You are not meant to be carrying around in your bag like a laptop. People are suddenly going to start taking a desktop on visiting clients? Packing all the external drives away and taken the as well. That'll look really professional, asking a client to wait 30 minutes while you plug everything. Horsehocky.

3,
Looks better? It's a workstation computer, not a Ferrari. Looks only matter to the sad lot that wont use it to make money.

4, After 4 years to work on this they give a mediocre increase in raw processing power. The GPU's are great but unless you run stuff that is optimised for them they are nothing more than heaters. We should have been seeing motherboards with 4 slots in these machines not a reduction down to 1.

5, Good expandability? HA HA. If you want to double the spend on the initial machine to get a similar set up to what you can already get inside the current systems. As for CopperCripple, don't believe the carp that is said. It was a joke on release, going from optic to copper.

6, Again it's a damn desktop, doesn't need to be carried around. The only time you need to move it is when you take it out of the box and plonk it on the desk.

Or is everyone going to get a plinth and bow down to it every morning?

You don't speak for everybody. Portability of a high power workstation is a major convenience for lots of people who travel in between offices several times a week. Looks are not very important, but can be appreciated. I'm not too concerned about appearance.

The performance of the machine in its current state (early beta, months from release, not final in any way shape or form) according to a test that was run on a free "32 bit" version of a benchmarking software that has never had its results accurately gauge performance considering how much more of your computer ANY app uses... Well I wouldn't make any financial decisions based on it.

OpenCL is used in many popular apps like Photoshop, Final Cut, Premier, even Handbrake... GPUs are a big part of this equation and you're foolish to dismiss that. That won't become any LESS true. This is free and often untapped processing power that has been recognized as a significant contributor and will continue to become more relied upon.

The expandability is excellent. It will be expensive but that isn't exactly an issue for studios buying several of these things and helping drive the price down for those with smaller budgets. I paid 600 dollars for a CD ROM drive for a Packard Bell once. Now I have 10 of them in a shoe box in a storage facility I haven't visited in months. And I haven't bought a machine with an optical drive in 2 years. And we don't use the Mac Pro's. Things change, and Apple doesn't usually swing and miss when it comes to core functionality. External Expansion is likely the future. Hopefully it can somehow look prettier than I'm imagining it.

Totally Agree! Only thing that would lead me to a MacPro is the "Desktop" feeling, and the expandability. The new macpro to me is a big step back in history of the macpro.

i will go for the new maxed out haswell imac this fall.

How will that solve any of the shortcomings? You gain a monitor but lose significant power. Doesn't sound like you were in the market for a Mac Pro to begin with.

 

[deconstruct60]

How well does Geekbench measure OpenCL performance as a factor of overall system performance?

Does not measure it even in the slightest. It is a CPU only benchmark that measures the CPU's computational 'speed' and the memory throughput to the CPU.

I suppose it is an indirect bechnmark of OpenCL code that could be run on the CPU, but it does not measure it directly.

So yes, Geekbench on this 2013 Mac Pro completely avoided measuring the computational abilities of several hundred cores residing inside the system. And yet somehow is implicitly being used to accurately measure the computational capability of the whole system.

 

[ikir]

I was expecting overall geekbench scores to be much higher...

On a prerelease machine, with a beta software? compared to one of the most powerfull workstation which exist today? The results are already very good and i'm quite sure it will change a lot with final hardware.
Also the new Mac Pro will destroy the old one in gfx performance, you will not see this thing in geekbench since it counts only cpu and ram speed.
Also think about 1000MB/s SSD speed compared to standard 80-100MB/s HD or even raid :-D
it will be a GIANT LEAP

 

[z3r0]

I'd like to see benchmarks with the current highend Mac Pro with OWC 6G SSD drives and the fastest compatible NVIDIA GPU versus the new Mac Pro.

That would be real interesting.

Too bad SLI isn't possible on OS X.

Oh puh-leaze.

A 12-core chip that's 10% faster than 12-cores of 3 year old chips isn't "kicking the crap" out of anything.

And what are these magical missing optimizations? (links, please)

Apple has killed the power mac, and finally produced the xMac. Too bad, though, for the people who needed Power Macs.

 

[Chupa Chupa]

Totally Agree! Only thing that would lead me to a MacPro is the "Desktop" feeling, and the expandability especialy the GPU power (not a "laptop gpu" like in the imacs). The new macpro to me is a big step back in history of the macpro.

i will go for the new maxed out haswell imac this fall. even if it would be the same price as the entry level macpro, i will get a gorgeous cinema display on top.

Your post confuses me. The key reason anyone buys a MP is expandability, esp GPUs. But if that is your interest how does the iMac solve that?

The iMac is great for someone who just wants to plug in their new computer and use it, but won't be consistently doing high end graphics work.

You mock the "laptop GPU" in the iMac then talk about how you are interested in one with a "gorgeous cinema display"

 

[iBug2]

- "still has good expandability with thunderbolt 2"? Expensive expandability maybe. "Good expandability" is internal SATA3 and more USB3.

SATA 3? Are you kidding me? Even USB 3.0 is more practical than SATA 3, let alone TB.

----------

Too bad SLI isn't possible on OS X.

Well, obviously it'll be possible in 10.9 since they are putting 2 GPU's in the new Mac Pro. Or did you mean Nvidia's double GPU implementation only? If they go the AMD way and don't support double nvidia GPU's, well then they will never use Nvidia in Mac Pro again.

 

[Digitalclips]

Since the performance is only incrementally more and the size is substantially less and it is a single chip machine, I suspect this will be under $2k price point. A Mac-Mini x2 of sorts.

The question becomes, what about dual chip versions?

Sub 2K ... God I hope so but that would kill high end iMac sales.

I wonder if they daisy chain/ cluster if you want more power?

 

[WilliamLondon]

Who? I don't wanna connect with anyone on Facebook. Did you just refer to yourself in third person?

I'm pretty sure he meant the gay porn star who shares his name.

 

[Chupa Chupa]

You also have a lot of faith in Internet consensus.

"Internet consensus" -- now there is a 21st Century oxymoron

 

[auero]

I'll be a little disappointed if these numbers are true. I was expecting a little higher but we'll have to wait and see. Apple doesn't have control over what processors Intel makes so it's not really their fault. All these months of waiting weren't because Apple didn't feel like releasing a machine but because Intel didn't have anything to really offer. I feel Intel drags their feet when it comes to the Xeon line.