I happen to think Apple would use Itanium2~Shard~ said:Yes, perhaps we will see Xeons in the Xserve - interesting thought... definitely could work...
I happen to think Apple would use Itanium2~Shard~ said:Yes, perhaps we will see Xeons in the Xserve - interesting thought... definitely could work...
EricNau said:I happen to think Apple would use Itanium2
Not yet, though it has been dieing for some years in the opinion of most.ffakr said:Itanium is dead.
ffakr said:You do realize that:
a) EPIC, the Itanium ISA is not related to x86 in any way. It's not even CISC or RISC. It's VLIW (very long instructional word). Porting to Itanium would likely be more difficult than porting from PPC to x86 because more opensource work has centered around ppc to x86.
b) Itanium is dead. The Power5 is more powerful than Itanium. Opteron servers have a price/performance that is SEVERAL times greater than Itanium. Even HP has been panning Itanium lately and they helped create it.
There is ZERO chance that Apple will ship Itanium servers unless there is a sea change in computing over the next DECADE.
ffakr
Not only that, but the early ones ran hotter than the hottest Pentium 4s. I don't know if the later ones corrected that.Skareb said:Not yet, though it has been dieing for some years in the opinion of most.
Still a very powerful chip, and that can't be denied, just over priced and not really supported.
Jordan
~Shard~ said:At times, it helps to be realstic about things - do you really need the dual 30" displays, 16 GB of RAM, quad preocessing and 1 TB of storage? That may help lower your price. 😉 😎
Porco said:Sorry if someone's made the point as I haven't read the whole thread but for those who are disappointed with Apple measly installed amount of RAM I'd just say this - would you prefer they shipped with more RAM at Apple RAM prices?!
I'm very new to the mac world but as a Sound guy I've decided that my next purchase will be a G5, where is this "refurb store" you speak of? Do you think it would be worth while to purchase a refurbished mac?😕cr2sh said:Of the 1.8 G5 machines apple has currently in their refurb store... one comes in a metal case and the other costs $50 more and comes with a 20" display.
Stunning price structure.
🙂
Spankenrear said:I'm very new to the mac world but as a Sound guy I've decided that my next purchase will be a G5, where is this "refurb store" you speak of? Do you think it would be worth while to purchase a refurbished mac?😕
jiggie2g said:Is it me or did Apple find a new way to screw thier loyal users. Let's see.
new power mac goes dual core , as compared to DP so processor price should be down atleast 60% , no more dual socket MB so cut that price in half as well. and throw in a cheapo 4 pipe 6600LE(ROTFL) and 512MB of latency crippled DDR2/ 4200 they still they m,anage to keep the prices up. PCIe should no be a cost factor as this has reached market saturation..like DVD+-RW Drives.
jiggie2g said:Worst of all the 7800GT is going to be delayed till got knows when, and might only be BTO so all u PowerPC nuts who want to upgrade are screwed.
jiggie2g said:DDR2 is a joke until u get to 667-800mhz then it starts to show promise. But even good OCZ DDR3200 can be OC'd to 550-600mhz at a much lower latency for faster performance.
cr2sh said:From the Apple Online Store main page, on the right side, towards the bottom is a red "sales tag" that says someting like "special deals."
That's the refurb store.
If I were to buy a new mac, that's where I'd do it. Immediately after Apple releases a new model they take the old refurb models out of the store. Then about a week later they put the old refurb'd models back up... discounted somewhere between 20 and 40% off the original price.
The 20" 1.8 Imac used to sell for $1800 and its now down to $1150.
I'm not 100% sure what powermacs are up there now.... but I know there's a 1.8 single for $1100, and I believe a dual 2.0 right around $1500, which is marked down from $2000 a couple weeks ago. I think there were dual 2.3's up a while back... not sure.
I've never seen a dual 2.7 on the refurb store... and I'm not sure I'd even consider buying one, but that's another discussion.
You can also buy Applecare for refurb machines, so there's no real risk involved, from my point of view. The savings is considerable and they're still fantastic machines.. buying from refurb keeps you up to date but off the bleeding edge.
🙂
Where did you get your information?Frobozz said:The G5 does not suffer the same latency problems, and as such, is actually one of the first computers to not suffer latency with DDR2.
anandtech said:http://www.anandtech.com/mac/showdoc.aspx?i=2436&p=2
A real complaint can be lodged against the latency to the memory. Apple's own webpage talks about 135 ns access time to the RAM. Now, compare this to the 60 ns access time that the Opteron needs to access the RAM, and about 100-115 ns in the case of the Pentium 4 (with 875 chipset).
...
(see link for table comparing G5/Xeon/Opteron memory benchmarks)
...
Memory latency is definitely a problem on the G5.
DDR2 does have more latency than DDR1...that's an inevitable part of the DDR2 design. Unlike a typical Windows PC bus, the G5 has the bus speed to fully take advantage of DDR2 memory. However, like AidenShaw stated, the G5 has inherent memory latency issues. This most likely limits the gains from DDR2 to some extent.Frobozz said:Processor price won't be 60% less. The cost of a processor involves a lot of factors, but I would be very surprised if the cost was any lower. It's twice the silicon needed for a single core processor because they're both at 90nm. Should be very similar.
DDR2 / 4200 RAM being "latency crippled" is a nasty rumor that just won't die. DDR2 RAM is only latency crippled in PC Hardware because of their bus structure. The G5 does not suffer the same latency problems, and as such, is actually one of the first computers to not suffer latency with DDR2.
The PowerMac is the first PCIe-only motherboard I have ever seen. Anywhere. Maybe there's another one somewhere, but it's certainly not a mainstream computer it's being used in. I wouldn't count any of your points as being a price reduction.
Agreed. It sucks. It's probably more for show that real world workstation usage, but it still sucks.
Again, this is a myth propogated by the review of PC motherboards when using DDR2 memory. The new PowerMac memory controller can push 8.5 GB/s. The RAM runs at a lower ratio (1.066GHz effective clock) to the processor (2.5 GHz) than the old memory did (0.8 GHz effetive clock) when it was on a slower 6.5 GB/s memory controller. Long story short-- the RAM works much better now because it can push more per cycle with more total cycles.
This is where my interpretation, and not paraphrasing from industry reviewers, takes over. Feel free to correct me if I'm wrong ...
Most modern PC Motherboards run at 4 passes of 200 MHz (800 MHz FSB implementation.) So if the RAM is an even divisor of 200, you don't get latency. The problem is that the numbers you're dealing with are so slow, that any hit on latency is significant.
400 MHz Ram would work well on a quad pumped 800 MHz MB. 533 would obviously leave wiggle room and not be very effective versus 400.
On a PowerMac, the RAM runs at 533 x 2 and the path to the processor gives it half the processor speed. Let's say that's 1.25 GHz for a 2.5 GHz G5. That's 625 MHz bidirectionally, which is MORE per clock than the RAM. Not so in a PC, where it's a mere 1/3rd the speed.
wrldwzrd89 said:DDR2 does have more latency than DDR1...that's an inevitable part of the DDR2 design. Unlike a typical Windows PC bus, the G5 has the bus speed to fully take advantage of DDR2 memory. However, like AidenShaw stated, the G5 has inherent memory latency issues. This most likely limits the gains from DDR2 to some extent.
The 1MB L2 cache does help mask the latency quite a bit, but it can't eliminate its effects entirely. A larger L2 cache would help even more. As far as large amounts of data go, that's true of all memory improvements, not just DDR2. Other than those minor niggles, you are correct.andiwm2003 said:i thought the 1mb level 2 cache masks the latency problem to an extent where it doesn't matter in real life anymore.
i also thought when uploading large amounts of data (video stuff) the ddr2 is faster because of the faster clock speed and latency isn't an issue there.
therefore apple got the faster memory for large amounts of data while overcoming most of the latency with the cache.
correct me if i'm wrong.
The amount that the larger cache offsets the latency is quite dependent on the application.wrldwzrd89 said:The 1MB L2 cache does help mask the latency quite a bit, but it can't eliminate its effects entirely. A larger L2 cache would help even more. As far as large amounts of data go, that's true of all memory improvements, not just DDR2. Other than those minor niggles, you are correct.
Frobozz said:DDR2 / 4200 RAM being "latency crippled" is a nasty rumor that just won't die. DDR2 RAM is only latency crippled in PC Hardware because of their bus structure. The G5 does not suffer the same latency problems, and as such, is actually one of the first computers to not suffer latency with DDR2.
On a PowerMac, the RAM runs at 533 x 2 and the path to the processor gives it half the processor speed. Let's say that's 1.25 GHz for a 2.5 GHz G5. That's 625 MHz bidirectionally, which is MORE per clock than the RAM. Not so in a PC, where it's a mere 1/3rd the speed.
andiwm2003 said:i haven't read all of this thread but quite a bit. it seems some sort of consensus comes up along the line of what you said: most people are happy with the dual dual monster. therefore the positive ratings.
many people are happy with the dual core 2.3 ghz but think it's way to expensive. therefore the negative ratings.
so when people think it's a great update then they usually go for a dual dual monster. when people complain then they were in the market for a midrange model.
i wonder how that will play out.
are people not going to buy the midrange models?
and will apple react at some point and make a mini update or a price drop?
or will most people buy the midrange models anyway?
numediaman said:No displays. If you add displays the final cost is $23,420. It's the RAM that gets you to $17K.
To use 16GB RAM on the PowerMac you need to install eight 2GB DIMMs. The price of 2GB DIMMs is still high, but you can find them for about $350 or $400, so a good price would be around $3000. Still much less than Apple's prices.Photorun said:I went to dealram.com, you can get 16 GB for as low as $1600. Hmmm, I dunno, $1600 vs. $12000... they're close in price... NOT!!! If you want more certified worthy RAM you'll get up to around $2000 and save yourself, ready for this, $10,000!!! For $10,000 you can buy yourself three 30" Cinema Displays with video cards.
robc2323 said:RE: Quote:
Originally Posted by BGil
....................................................
Bottom line: The top-end PowerMac is very nicely priced but the two other models aren't even worth half the price Apple wants for them considering dual core PC's start at about $800 with 1GB ram, dual layer DVD burner, PCIe, and a 200-250GB drive.
ksz said:Originally Posted by Photorun
I went to dealram.com, you can get 16 GB for as low as $1600. Hmmm, I dunno, $1600 vs. $12000... they're close in price... NOT!!! If you want more certified worthy RAM you'll get up to around $2000 and save yourself, ready for this, $10,000!!! For $10,000 you can buy yourself three 30" Cinema Displays with video cards.
To use 16GB RAM on the PowerMac you need to install eight 2GB DIMMs. The price of 2GB DIMMs is still high, but you can find them for about $350 or $400, so a good price would be around $3000. Still much less than Apple's prices.
Were you looking at 2GB DIMMs?
That explanation makes a lot of sense...thanks AidenShaw, I guess that's cleared up now.AidenShaw said:The amount that the larger cache offsets the latency is quite dependent on the application.
Remember that the cache does nothing to help latency until you make the second read from the same cache block (128 bytes). On the first read, it does not help you - the 2nd and subsequent reads are much faster because they come from the cache.
Programs that make widely scattered references will get misses on virtually every read, and will see the latency each time.
Most programs have some measure of locality, and benefit from the cache. As the cache size is increased, more programs meet the "some measure of locality" point, and benefit.
So, larger cache is definitely good - but it is overstating the goodness to say that it means that high latency "doesn't matter in real life anymore." Any application that spills out of cache will still see the high latency.