Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Yup, next time I'm in the attic I'll dig around and Photo it.


Anyway for now this is all about you bro! So far your benchmarks are proving my initial hunches all wrong. Had I been on my mark your scores would be coming in at around 14000 to 15000. At 17665 I feel like I have to go back through all my posts and apologize for bad guesses. :eek:

Although the Cinebench10 scores play out right. So maybe I'm safe. :D

yep and with the cheapest octad coming in at over 13100 things are looking pretty good for those of us that need multi- threading power
 
With 17k for the 8 core 2.92 numbers being so high, it doesn't take a lot to extrapolate some results from those to guess what the four-core nehalem is going to be like:

2.92 x 8 = 17665 geekbench

If that scales perfectly with 100% utilization of all cores, the base model 4-core would be:

17665/2.92/2*2.66= 8046 geekbench

This puts it just under the 2008 2.8 eight-core which isn't bad for a four core entry level Mac Pro.

More realistically, the second chip on the 2.92 is likely to be underutilized compared to the first, with diminishing returns as shown by the 6x results from the cinebench test. I'm guessing the four-core machine pulls at least an additional 15%, putting the GB around 9200, pretty much a dead heat with the 8-core 2008 model.

This creates a bit of a quandary for the new buyer. 8 core xeon 54xx on liquidation or four core nehalem? Applications that don't take advantage of the second cpu may prefer the nehalem system, and the GT120 is a better video card that the stock 2600 that comes with the 2008 model.

I think it will all come down to RAM and specific application performance. Personally, I'd love to see some Handbrake encode numbers from the 2.66 Nehalem....


In reference to Cinebench R10 at least, my 8 Core 2.8 early 2008 with 16GB RAM and a 8800GT gets:

3186 1 CPU
17464 xCPU
5.48X Multiprocessor speedup

6002 on Open GL

The 2.93 Nehalem isnt THAT much faster.

In fact, taking the roughly 33% speed over the stock 8 core 2.2 down, the Nehalem Mac Pro will actually be slower than the 8 core 2.8 Penryn Xeon by a tad.

That number actually shocked me. I'm just not seeing the benefit over the 2008 models, especially at the inflated cost. And I was about to buy one of these today; I'll just get the Radeon card instead.
 
Exactly. I'm really surprised by the single threaded scores. Most consumer grade apps struggle to use more than one core so it's interesting to see my 3.2GHz 2008 mac pro beat the 2.26GHz 2009 mac pro in every single theaded benchmark by quite a margin.

2008 3.2Ghz 8 Core Mac Pro
Blowfish single threaded scalar: 2341
Text Compress single-threaded scalar: 2858
Image Compress single-threaded scalar: 2579
Lua single-threaded scalar: 5441

2009 2.26GHz 8 Core Mac Pro
Blowfish single threaded scalar: 1353
Text Compress single-threaded scalar: 1659
Image Compress single-threaded scalar: 1473
Lua single-threaded scalar: 2388

Basically. If you use multi-core apps the 2009 2.26 Mac Pro is on par with the 2008 3.2Ghz Mac Pro. If you use apps that only use a single core, like most consumer based apps as well as pro apps like Adobe Illustrator the 2008 3.2GHz Mac Pro whips its ass (in my opinion) ;)

Still LOVING my 2008 Mac Pro :D

Given the changes in processor architecture, this is not a surprise at all.

Nehalem brings back hyperthreading ensuring all 8 cores are fully utilized with 16 threads in the pipeline whereas the Core 2 architecture can only execute a single thread per core.

Single threaded apps are most affected by raw clock speeds... so it's not surprising that a 40% increase in clock speeds nets a significant gain.

It's another reminder to people that unless your apps are massively threaded, a higher speed quad core is a better investment than a slower octo-core.
 
I'm sad to see that the QPI is lower in the octo 2,26 (5,8 GT/s) vs 2,93 (6,4 GT/s).
If so I believe that the Quad 2,66/2,93 will be 4,8 GT/s :'(

2,26
Memory Memory performance 3283
Stream Memory bandwidth performance 3588
Bus Frequency 5.87 GHz

2,93
Memory Memory performance 5075
Stream Memory bandwidth performance 4636
Bus Frequency 6.40 GHz
 
Isn't snow leopard going to pave the way for multi-threaded apps? I anticipate my 2.26GHz 8 core to be more useful (to a point) as time goes on due to this since more apps will be moving towards multi-thread. Am I mistaken in this assumption?
 
Basically there's no advantage to owning a 2009 2.26GHz mac pro over owning a 3.2GHz 2008 mac pro. None what-so-ever. The poor single threaded performance of the 2.26GHz 2009 mac pro in comparison to the 3.2GHz 2008 is a deal beaker in my opinion.

Might be jumping the gun a little but if you look at the bigger picture the 3.2GHz 2008 mac pro is either equal or much faster than the 2.26GHz 2009 Mac Pro. Since the majority of apps rely on the power of the single thread, the 2008 3.2GHz mac pro is essentially faster (in my opinion) and therefore a fantastic purchase over the 2009 2.26GHz for less money if you can get a second hand unit or a refurb.

Er, with a new OS about to be released which is likely to be optimized for the new CPUs (e.g., Intel HT support, etc.), this seems more than a little early to call IMHO.

While it certainly seems likely that prices for the Nehalem units will fall in the coming months, in any event, the OS and Mac apps will be built to support the new technology over the next several years, and real world performance will likely increase substantially over time.
 
Basically there's no advantage to owning a 2009 2.26GHz mac pro over owning a 3.2GHz 2008 mac pro. None what-so-ever. The poor single threaded performance of the 2.26GHz 2009 mac pro in comparison to the 3.2GHz 2008 is a deal breaker in my opinion.

Might be jumping the gun a little but if you look at the bigger picture the 3.2GHz 2008 mac pro is either equal or much faster than the 2.26GHz 2009 Mac Pro. Since the majority of apps rely on the power of the single thread, the 2008 3.2GHz mac pro is essentially faster (in my opinion) and therefore a fantastic purchase over the 2009 2.26GHz for less money if you can get a second hand unit or a refurb.

My MBP circa 2006 always has two CPUs going. More apps take advantage of multiple CPU's than you'd think. Did anyone really expect the 2.2GHz Nehalem to not be outdone on a single thread by a 3.2GHz ?

Lets all wait for more real world benchmarks. I have yet to see anything that would convince me that there is a machine other than the 2.26 Nehalem Octo in my future.
 
Two very good points above but I'm sticking with my analysis. I don't think Snow Leopard will magically start giving the likes of Illustrator CS4 8 cores of power when it's only designed to use the one. We'll see.

I think my comparison is more of a feel good note for existing owners of the 2008 mac pros and something for those to consider when purchasing the 2.26 8 core 2009 model. :)
 
photoshop

yea its a bit of a bummer since I'm a photoshop jockey, but I expect snow leopard/CS4 or CS5 will finally recognize multiple cores. I blame Adobe.

I also have no idea if Apple's nVidia strategy will apply when latest nVidia boards come online for Mac Pro, so in theory Photoshop can scream with transparent utilization of all available chips in parallel, if designed right. But Adobe has a lot of legacy code to rewrite. I used to program massively parallel processing supercomputers - it's no easy task and highly algorithm-dependent.
 
Nothing optimized, came out to 47 seconds on 2.26/6GB on CS3

2.8/2GB Early 2008 Mac Pro (with stock HD) on CS3: 29 seconds.

Would you like to retest? That looks suprisingly slow to me.

Bloody ell! I just got 19.2 seconds in Photoshop CS4 on my 2008 3.2GHz Mac Pro. For the money I'm soo glad I went for the 3.2Ghz back in January and didn't wait! I'd hate to pay the same for the 2.26 model and get this kind of performance. Ouch!
 
Photoshop Speed Test

Well I just bothered to read the instructions this time, set the PS settings and rebooted (reboot took 30 secs to login screen exactly, btw) and now I get 33 secs.

Ive got to get some serious RAM and a Velociraptor in there and see about overclocking it.
 
Nothing optimized, came out to 47 seconds on 2.26/6GB on CS3



Bloody ell! I just got 19.2 seconds in Photoshop CS4 on my 2008 3.2GHz Mac Pro. For the money I'm soo glad I went for the 3.2Ghz back in January and didn't wait! I'd hate to pay the same for the 2.26 model and get this kind of performance. Ouch!

Wait until Snow Leopard comes out with QT-X then transcode a HD video and the 2.26 will increase performance dramatically.;)

That's why I got the 2.26. I know what Apple's gonna do with this newer technology. One of the advantages of being an ADC member. hehe
 
Don't know what's up with Geekbench, but here are the 64-bit 2.26 results:

Does anyone have a thought why the 8x2.26 would show a lot more performance than the 8x2.8 on the Mandelbrot, Image Sharpen, and Image Blur floating point tests (single- and multi-threaded) yet worse performance for most of the other floating point tests?

Could you run Cinebench on your 2.26 please?
 
Wait until Snow Leopard comes out with QT-X then transcode a HD video and the 2.26 will increase performance dramatically.;)

That's why I got the 2.26. I know what Apple's gonna do with this newer technology. One of the advantages of being an ADC member. hehe

You got it! I'm really hoping that the event on the 24th will have a mess of software to complement the processors. Even the Snow Leopard being released then would be nice...
 
Wait until Snow Leopard comes out with QT-X then transcode a HD video and the 2.26 will increase performance dramatically.;)

That's why I got the 2.26. I know what Apple's gonna do with this newer technology. One of the advantages of being an ADC member. hehe

A very fair point. Looking forward to see what Snow Leopard brings but I'm not sure it'll make that much difference. We'll see ;)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.