So, I called an apple business rep on the phone, and he suggested to me that the comparisons on the site were purely clock-speed, and did not reflect the real-world advantages of the systems in multicore apps (ie, logic, my main concern). He said that the 2.26 would "definitely" be faster in real-world situations than the old 3.2. He seemed very confident, and said that they've "been waiting for over a year for intel's new architecture, and it was not for no reason." So, don't know if it's true or not, but got my hopes up all over again...
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
From the Horse's Mouth: 2.26 faster than 3.2 (not confirmed)
- Thread starter cybeross
- Start date
- Sort by reaction score
I just called apple customer support and was transferred to the desktop division.
I spoke to a fairly knowledgeable guy named Kevin. I asked him verbatim, "Is the march 2009 mac pro eight core 2.26 faster in terms of overall performance than the 2008 eight core 3.2 system?"
He said, "The answer to that question is yes."
I said, "Yes? Really? Faster than the 3.2?"
Again, "Yes," he said. He continued, "You'll see a good amount of a speed advantage in the new ones, and even the lowest eight core of the new models should be faster."
I spoke to a fairly knowledgeable guy named Kevin. I asked him verbatim, "Is the march 2009 mac pro eight core 2.26 faster in terms of overall performance than the 2008 eight core 3.2 system?"
He said, "The answer to that question is yes."
I said, "Yes? Really? Faster than the 3.2?"
Again, "Yes," he said. He continued, "You'll see a good amount of a speed advantage in the new ones, and even the lowest eight core of the new models should be faster."
Yep which is why there is a lot of confusion, but a 40% increase in performance across all applications in benchmarks isn't something that is being found on the single quad core Nehalems that are already out. Let alone real world performance.
In memory intensive apps this may be true, but generally this sounds like a lie plain and simple. It would be interesting to underclock a 920 to 2.26 and see how it's numbers compare to to a 3.2Ghz eight core (nanofrog, I demand you do this immediately). All Windows or all OSX, of course. Not an entirely accurate comparison, but should give people at least some idea of how they stack up.
This article lists some interesting result from the server side real-world applications:
We were quite amazed, even slightly suspicious, when HP and Fujitsu-Siemens Published their SAP numbers. These numbers showed that the newest Xeon X5570 (Nehalem EP) series offer an enormous performance boost over the Xeon X5470 (Harpertown). After all, an almost 100% improvement at a slightly lower speed (2.93 GHz vs 3.3 GHz) is nothing short of amazing. Turns out that the real clockspeed is 3.2 GHz (2.93 GHz + 266 MHz turbo) but that does not alter the fact that these are truly incredible performance numbers.
Why is everyone so surprised? Did NONE of you read any of the Core i7 reviews on Windows? It's the same story, 40% or more performance increase. It was a given that the new Xeons would be the same.
SAP is perfect for showing off the improvements in the Nehalem/Tylersburg architectures. It is in the server space where it will really shine. That doesn't necessarily translate to the workflow of someone who spends their days creating digital content though.
40% was banded about early on, then it was more 20%-40% depending on Application. Then when the processors went in to the wild things like this appeared:
I know Nehalem has big improvements, but I am skeptical that a 2.26GHz processor is "faster" than a Penryn 3.2GHz. I'm happy to be proved wrong
. I would love it if through Snow Leopard and encouracing the bigger application developers to imrpove code that the architecture changes and more specifically multi-threading capabilities could really enhance performance.
I know Nehalem has big improvements, but I am skeptical that a 2.26GHz processor is "faster" than a Penryn 3.2GHz. I'm happy to be proved wrong
Here is my test results.
I tested MPEG4 to Divx (v6.6) Export Encoding by EyeTV 3.1 on Leopard 10.5.6. (by EFI-X, OSx86 machines)
Source video file : 20min (511MB) MPEG4 format on FW400 Ext HDD
Q6600 3.2GHz (2GB RAM) : 345sec to finish exporting by Divx Encoding
i7 920 2.66GHz (3GB RAM) : 317sec (8.1% faster from Q6600 3.2GHz)
i7 920 2.92GHz (3GB RAM) : 293sec (15.1% faster from Q6600 3.2GHz)
i7 920 3.2GHz (3GB RAM) : 271sec (21.5% faster from Q6600 3.2GHz)
All of this results are based on quad-core, but we can guess by 8-core.
I think if you had previous 8-core 3.2GHz MP, you should get 2.93GHz model (or at least 2.66GHz). Otherwise, won't be impressed...
So a guy at Apple told you that their newest Apple product was faster than their older, less expensive Apple product?
WOW!
Those numbers are exactly sort of thing I had in mind when discussing things in the "Gainestown: What we know" thread.
Being that 6 of the benchmarks on Apple's site put the 2.93GHz as 1.2x faster than the 3.2Ghz that means the 2.26GHz is 93% as fast as a 3.2GHz in those benchmarks. (Purely based on strict numbers of course)
i know, i know. but i read the OP and wanted to see if i'd get the same answer out of an apple rep. since i called, i figured i'd post. note that i didn't call a business rep, but a desktop tech rep.
Do you have an '08 MP 3.2 Octo lying around?
I don't,
so I can't actually compare them. If we can get someone with one and is willing to help, we could pull it off.
MacRumors Testing Labs Presents...
How about someone post a 300 meg file (test file) to encode with handbrake and all of us with mac pros do this and post results. Doesnt have to be exactly 300 but the same file size everyone can use. We start this now and by the time folks receive the new mac pros, we will have a good data set to compare. We would need a place for everyone to download this from though. I think the handbrake test would give us a good idea of raw power.
Do you have an '08 MP 3.2 Octo lying around?
I don't,
If we can get someone with one and is willing to help, we could pull it off.
MacRumors Testing Labs Presents...
I meant you could supply the i7 numbers. I'm sure someone around here has a 3.2Ghz Mac Pro.
Just to get something up quickly, download this video of me playing Halo. Do whatever you want to it, just make sure you say what you did so people can use an identical method. It's only about 60MB, but as long as the file is the same the data will be fine.
http://files.filefront.com/cobrentwmv/;13389160;/fileinfo.html
Woo, double post. I wish I'd pay more attention.
I know what you meant.I meant you could supply the i7 numbers. I'm sure someone around here has a 3.2Ghz Mac Pro.
Any particular benchmarks (preferably free trials)?
What I've already downloaded:
CPU-Z
Sisoft Sandra
Prime95
Everest Ultimate
Geekbench
The Q6600 is built on the 65nm process while the Xeons in the Early 2008 Mac Pro are built on the 45nm process. The Xeons should be faster than the Q6600.
Well I don't have the 3.2 but my Octo MP is base 2.8ghz and I have underclocked my Hackintosh i7 920 to 2.26 (by lowering the multipler).
Geekbench Score stock Mac Pro 2.8 Octo-core= 9143
Geekbench Score underclock 920 @ 2.26ghz quad-core = 6841
Now if we get an octo 3.2 benchmark..
Geekbench Score 920 @ 3.2ghz quad-core = 9679
it's important to remember that benchmarking single-socket Core i7 vs. Core 2 Quad does not reflect the performance difference between a dual Harpertown and a dual Gainestown.
The new Intel QPI (Quickpath Interconnect) is similar to AMD's HyperTransport and allows the CPUs to talk directly to each other without having to go through the north bridge. The integrated memory controllers allow each CPU to control its own memory, again without being gated by the north bridge. In fact, there is no north bridge anymore, the X58 is merely an I/O hub.
What I'm saying is that there is much more of a performance difference between dual Harpertowns and dual Gainestowns than there is between single CPUs of similar architecture.
On top of that, we have hugely better memory latency and throughput. This is not merely a comparison of DDR2-800 to DDR3-1066. The Harpertowns used FB-DIMMs, and the buffering adds significantly to latency. Depending on load, an FB-DIMM can add 25% latency vs. an ordinary DDR2-800 setup.
The new Intel QPI (Quickpath Interconnect) is similar to AMD's HyperTransport and allows the CPUs to talk directly to each other without having to go through the north bridge. The integrated memory controllers allow each CPU to control its own memory, again without being gated by the north bridge. In fact, there is no north bridge anymore, the X58 is merely an I/O hub.
What I'm saying is that there is much more of a performance difference between dual Harpertowns and dual Gainestowns than there is between single CPUs of similar architecture.
On top of that, we have hugely better memory latency and throughput. This is not merely a comparison of DDR2-800 to DDR3-1066. The Harpertowns used FB-DIMMs, and the buffering adds significantly to latency. Depending on load, an FB-DIMM can add 25% latency vs. an ordinary DDR2-800 setup.
It's obvious that these benchmarks have more holes than Swiss cheese, but we want to get an idea of how the new single quad compares to the dual harpertown model. Not sure why I thought of underclocking (tipsy post).
Instead of a canned benchmark, try converting that file I posted from the current wmv to MPEG4 with HB and see how long of takes. Quick and dirty, but close enough.
Instead of a canned benchmark, try converting that file I posted from the current wmv to MPEG4 with HB and see how long of takes. Quick and dirty, but close enough.