http://www.barefeats.com/wst10c3.html
Those numbers don't sound right though. It makes no sense that 2.4GHz 8-core is faster than 2.66GHz 8-core.
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.
New mac pro purchase advice ASAP
- Thread starter jb502
- Start date
- Sort by reaction score
http://www.barefeats.com/wst10c3.html
Those numbers don't sound right though. It makes no sense that 2.4GHz 8-core is faster than 2.66GHz 8-core.
I closed the window already, but if you barefeats.com its only a little bit down on the frontpage. I am not sure if they included the 2008 mac pros in the test as I wasn't looking for that model in specific, but they listed quite a few. The take home message is that if you cant afford a 12 core, the next best machine is the 6 core, pretty much across the board.
Architecture matters. I am not stunned that a 2010 2.4 could slightly outperform a 2009 2.66 in the real world.
Both use the same Nehalem microarchitecture. Westmere was only a die shrink and brought a couple of improved features such as new AES instruction sets and better virtualization support. Those should not affect the encoding performance.
doesn't westmere also have more L3 cache? And don't the 2010 westmere's have faster memmory 1333 vs. 1066? Does this not come into play?
If it doesn't have some positive effect, why would they do it?
Yeah, Westmeres have 12MB of L3 cache while Gainestowns have 8MB. More cache should not affect encoding speeds though. Only some Westmeres support 1333MHz DDR3 and E5620 (2.4GHz) is not one of them. In fact, X5550 (2.66GHz) supports 1333MHz DDR3 but Apple blocked this in EFI somehow.
Westmere brought 6-core CPUs, that is why Intel did it. Those are extremely expensive so Apple used quad cores in low-end as they are cheaper.
Barefeats is a great and reliable site in my opinion and the results speak for themselves regarding how these systems play out in the real world. There are some other differences as touched on in the previous post as well aside from a die shrink that contribute to speed.
I'm not saying that Barefeats is not a reliable source of information but there is always a possibility of an error or a simple typo. Especially when considering that the guy who runs the site does not own or test all those machines AFAIK.
If you compare the 2.93GHz 8-core to the 2.66GHz 8-core, something does not match. 2.93GHz is 10.2% faster (2.93 divided by 2.66) but in that test, the 2.93GHz is 18.1% faster than the 2.66GHz. This time, both are Nehalem so you can't even say Westmere makes a difference.
I have to say that I find Barefeats to be quite unreliable. The benchmarks presented on this site aren't explained well (test cases, environments etc.) and often fairly misleading. Personally, I wouldn't bet a penny on those results.
well, i just went to my local apple store. They had a 2.4 8 core sitting next to a 2.8 4 core.
I opened up imovie on both. Loaded up one of the pre loaded projects on both. and hit export to 1080p on both at the same time.
the 2.4 8 core got done first.
This surprised me. Before i started i though it would be the other way around because everyone said imovie does not use multiple cores well, and the 2.8 has a higher clock speed. But the fact is the slower clocked 8 core beat the higher clocked 4 core by about 10 seconds.
I opened up imovie on both. Loaded up one of the pre loaded projects on both. and hit export to 1080p on both at the same time.
the 2.4 8 core got done first.
This surprised me. Before i started i though it would be the other way around because everyone said imovie does not use multiple cores well, and the 2.8 has a higher clock speed. But the fact is the slower clocked 8 core beat the higher clocked 4 core by about 10 seconds.
yeah but did they have the same ram did they have the same hdd did they have the same dvd player
well, they were both 100 percent stock, so the 8 core had more ram, everything else should be exactly the same.
The X5550 Nehalem 8-core (16 thread) is faster in any operation that is heavily threaded than a W3680. The W3680 has a higher TDP and less aggressive turbo modes, as well as more cache. Once your application's thread count goes lower, the W3680 will catch up rapidly in most situations.
Thats exactly what I have noticed with 6 core
Big hype about speed and test programs, but in real life
8 core crushed 6 core in lightroom (a true multicore program for exports) resize from raw and export to half size jpg and resize 2000 images from 26 Mb each to 500kb each
50 gig of folders took just over 62 min where 8 core westmere took less than 47 min
OK 6 core had 24 gig of ram where 8 core had 48
but still very impressive for 8 core
Both systems where asked to run Time Machine new complete back from zero at the same time
both 1 gig drives had 400 gig of files on them
Like I sad maybe in a day to day aps like aperture or photoshop may be faster, maybe even in regular edits in lightroom
but when asked to perform to max core thread output
more cores meen more power
BTW I have not noticed difference between the two 6 or 8 core just working in Aperture or Photoshop
Even using NIK or NX2 software was not a big deal they both worked with the same speed
If one was slower not by much
Since I work with very large amount of files export import and work on files at the same time more cores is much better
I know 12 or soon 16 will be better but for now 8 will do the job
for day to day use my 3.06 imac was as fast as the 8 core MP.
I had 6 core for less than a week and it went back to the store as it almost was a fast as my new 27" imac running with 16 gig of ram
When talking to apple they recommended a min of 8 core to notice a difference
If you ever had to wait for some exports to be completed before you can do more edits you will know
and i'm not talking about single file edits
I'm talking about editing 50 files or more at once
Last edited:
Basically you are making a case for your extra 4 RAM slots. It had TWICE the memory. And it must be exports because it sucks on imports.
http://macperformanceguide.com/Reviews-MacProWestmere-RAW-Lightroom3.html
All general users not needing 48GB of memory to deal with their pictures will see the 2.4 as much slower (the slowest) of Apples offerings.
Imports use single core where export use multicore
and not just operate on on raw MHz
export and processing use multithreads and thats where it shines
same with encode of true HD, that takes full advantage of multicore/thread
yes 3.33 is fast but at the same time limited for pro usage do to amount of ram slots available for multithreads aps
Care to retract? Only because you said "any" operation.
http://www.barefeats.com/nehal03.html
Look at Cinebench r10 Multi.
My W3680.
Mac: 25113
Win7: 27618
Thats about as multi threaded as it gets.
So you're trying to claim that your six 3.33GHz cores are faster than eight 2.93GHz cores? In heavily threaded operations, I'm not buying it.
If we talk about actual clocks (including turbo) you are at running at 3.5GHz effective. The X5570 is running at 3.2GHz effective. The reason the hexcore has a larger cache is because it has two more processors to feed with the shared L3. 2MB per core, which is the same as the X5570 (8MB, four cores). So that is 8 cores @ 3.2GHz with 2MB of L3 each. vs 6 cores @ 3.5GHz with 2MB of L3 each. 300MHz per core extra isn't going to make up for two more physical cores.
(To be fair, 12MB of shared L3 between six cores is probably better than 8MB between four cores simply because of Nehalem's architecture. It isn't magically going to make the chip insanely faster, though.)
Why don't you try Cinebench R11.5 and see what numbers you get. Note that Cinebench is a case I am very familiar with because I work professionally in Cinema 4D. I can tell you flat out that one W3680 is slower than twin X5550s because when I went to buy a Mac Pro I tested them extensively with C4D R12. The render engine in R11.5+ is different than in R10.
Buy the hexcore. It is a faster all-around performer, and a fabulous machine. It is difficult to find Mac Pros with X5550 or X5570s in them, and when you do they command a huge premium in the aftermarket.
Last edited:
Unless you will actually require more than 32 gigs of ram, the 6 core will be faster than any other mac pro, short of a 12 core model. The previous examples of an 8 core exporting libraries faster than a 6 core was because it had TWICE the ram for a ram intensive task. If both machines had the same amount of ram (and they both can unless you actually need more than 32 gigs) the 6 core will be as fast or faster at basically everything. There really isn't much room for debate anymore except from those looking to justify the purchase of their 8 core machines, or those that actually need and use 32+ gigs of ram.
I was talking 2.66GHz not the 2.93GHz. I am really not claiming anything. Those are the numbers my Mac posted. I am a dork and keep a testing spreadsheet
Major difference could be Rob tested with a 10.5 flavor (not sure) and my tests are with 10.6. But Win 7 is the test champion. Since you asked I also have Cinebench r11 numbers (again, spreadsheet)
OS X:
35.01 OpenGL, 8.94 MultiCPU, 1.19 Single
Win 7:
69.14 OpenGL, 8.82 MultiCPU, 1.20 Single
Also I know Geekbench isn't that great but it shows similar "general" results with the 2.66 @ 14458, 2.93 @ 14906, 3.33 6-core @ 14183. Looks very similar. Gains not worth worrying about anyway. I don't know, blame the software.
Last edited:
Yes. However, in the chart you linked your machine is at parity with an X5570 machine as well.
Because of the score your machine was able to pull in R10, I deeply suspect that it is not properly parallelized. Clock seems to play a much larger role than it should.Those are the numbers my Mac posted. I am a dork and keep a testing spreadsheet
I just ran my machine through R11 as a comparison. It came around roughly where I'd have expected it to.
Multi-CPU: 9.87
Single-CPU: 1.01
Somewhat ahead in multithreaded performance and somewhat behind in single threaded. There may be a small amount of additional performance on the table here, I only closed the major programs to run Cinebench.
I'm not seeing those numbers at all in Geekbench. My 32-bit score is 15,063. Ran it twice for confirmation and it's reporting basically the same.
Regardless, I agree with where you're going. The single hexcore machine is a great piece of hardware. It will satisfy the average multithreaded software, do very well with single threaded. Only the user with a heavily parallelized workload need consider more cores. Even then the Nehalem dual socket machines are not a great deal faster. Most applications will have limited parallelization, so the hexcore is faster more often than not. As can be seen here: http://www.barefeats.com/wst10c2.html
Aperture Test: Max of six cores addressable, hexcore takes it thanks to good core count and lots of clock speed.
Color: Not well threaded, hexcore takes it due to clock.
Handbrake: Limited amount of parallelization in the H.264 encode process. Several models are the same here. Unsure why the twin X5670 box is faster (error?).
Photozoom: Highly threaded. More cores win.
Last edited: