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.
Mac Pro (Early 2008) Geekbench Benchmarks
- Thread starter MacRumors
- Start date
- Sort by reaction score
Just checked and they only have the older Mac Pro. For the old one, a 4 GB RAM stick is $332 and change. I checked how much it would cost to upgrade the new Mac Pro to 4 4GB sticks (I know different kinds of RAM, but close enough). Just to upgrade to 4 4GB from the standard config (2 x 1 GB) is $4300. If I were to get 4 4GB sticks from dealram.com, it would be only about $1333. What a rip off.
I took a look at that result and my guess is that someone's running Mac OS X as a virtual machine. Geekbench has a couple of known issues when running under virtual machines that can cause scores to be overinflated. I've marked that result as invalid so it won't show up in the Top Geekbench 2 Scores listing.
FWIW the result in question isn't from a Hackintosh with a Mac Pro motherboard. Hackintoshes rarely report the hardware they're running on accurately. I guess people want to pretend their $800 PC is equivalent to a $2700 Mac.
Reminds me of the Vector Engine days, when people used to ohh ahh over numbers, yet only a few application taken full advantage of its capabilities.
What a waste?
SSE4 is great, however if developers do not allow they applications to utilize it, this becomes another Vector Engine scenario. History repeats itself.
10.5.2 might be SSE4 compliant, however how much faster is the OS going to operate if its still loading from a HDD rather than flash memory.
What next numbers of SSE4 with 10.5.2 loading from flash memory?
What a waste?
SSE4 is great, however if developers do not allow they applications to utilize it, this becomes another Vector Engine scenario. History repeats itself.
10.5.2 might be SSE4 compliant, however how much faster is the OS going to operate if its still loading from a HDD rather than flash memory.
What next numbers of SSE4 with 10.5.2 loading from flash memory?
I suspect that the extra 1GB of ram in the new model is giving it a hugh boost, SSE4 might have a slight edge as well, however not by that much.
your women gave you the new mac pro? WTF. You must of been good in bed or listen to her BS.
your women gave you the new mac pro? WTF. You must of been good in bed or listen to her BS.
His split personality is a woman.
So does this mean it will run Myspace faster then my 1.6 G5? *Runs away*
Safari will feel snappier.
Geekbench FP benchmarks seems strange
Apple has posted SPECfp_rate_base2006 data of:
8* 2.8 GHz 68.5
8* 3.2 GHz 72.6
8* 3.0 GHz(old) 55.9
which suggest that the new "harpertown" processors offer a substantial improvement in floating point (FP) performance as compared to the older processors--even at the lower clock speeds. The Geekbench data seems strange in that it suggests the old 3.0 GHz processors are 12% faster than the new 2.8 GHz harpertown processors, which is about DOUBLE what one would guess purely on the clock difference--normally performance scales sub-linearly with clock speed (consider, for example, Apple's benchmarks that show the 3.2 GHz processors as ~6% faster than the 2.8 GHz processors as compared to a raw clock difference of 12%). Additionally, one would expect that a refined processor with 50% more L2 cache, faster memory, and a faster bus would provide a significant improvement in FP performance as Apple's data suggests. The Geekbench FP performance seems suspect, if it is right this is a very disappointing result....
Apple has posted SPECfp_rate_base2006 data of:
8* 2.8 GHz 68.5
8* 3.2 GHz 72.6
8* 3.0 GHz(old) 55.9
which suggest that the new "harpertown" processors offer a substantial improvement in floating point (FP) performance as compared to the older processors--even at the lower clock speeds. The Geekbench data seems strange in that it suggests the old 3.0 GHz processors are 12% faster than the new 2.8 GHz harpertown processors, which is about DOUBLE what one would guess purely on the clock difference--normally performance scales sub-linearly with clock speed (consider, for example, Apple's benchmarks that show the 3.2 GHz processors as ~6% faster than the 2.8 GHz processors as compared to a raw clock difference of 12%). Additionally, one would expect that a refined processor with 50% more L2 cache, faster memory, and a faster bus would provide a significant improvement in FP performance as Apple's data suggests. The Geekbench FP performance seems suspect, if it is right this is a very disappointing result....
SPECfp is a memory benchmark
SPECfp to a great degree measures memory bandwidth at least as much as floating point arithmetic speed.
One would expect the Stoakley platform to be much better than the older chipset at memory bandwidth, and therefore to perform much better on SPECfp than expected from simple comparisons of CPU frequency.
It's not at all surprising that the 2.8 Stoakley easily bests the 3.0 5000X at SPECfp, and not at all surprising that other benchmarks would show something different.
SPECfp to a great degree measures memory bandwidth at least as much as floating point arithmetic speed.
One would expect the Stoakley platform to be much better than the older chipset at memory bandwidth, and therefore to perform much better on SPECfp than expected from simple comparisons of CPU frequency.
It's not at all surprising that the 2.8 Stoakley easily bests the 3.0 5000X at SPECfp, and not at all surprising that other benchmarks would show something different.
Something different maybe, but to see that when normalizing the results to the same clock, the old xeons are supposedly somewhat 8% faster in floating point calculations than the new ones in the geekbench tests makes one wonder at how unreliable the geekbench tests must be... (I am very surprised they published such results !)
phjo
I got my real-world benchmark, as I had ordered my Dual 2.8Ghz Quad Core Mac Pro on Tuesday. It came and I set it all up yesterday.
Here is my Real-World benchmark... you ready?
F**king FAST!!! So much so, I got bugs in my teeth.
But seriously, I still think it is a bit early to tell. I mean, we have to receive updated software that has been optimized for SSE4. But so far, on my end, it's all good. I am so glad I toughed it out and waited for this batch of Mac Pros to come out. I was seriously struggling by the end of this year, aching to make my purchase. My poor little G4 is sitting in the corner, retired... kind of feel bad for the little inanimate object that served me so well.
I think it would work at 16% slower clock speed. I don't know if it's still true, but Intel tends to lock the clock multiplier so that the CPU runs at a certain multiple faster than the bus. If you put the chip in a slower bus, you're going to get a lower speed.
For best performance, you want to have matched sets of four, so it should be 1GBx4.
Can you back this up with some benchmarks? Sorry, I don't question the statement and you are not the first person to say it. But I would like to see some test results before I go spending a lot of money on memory. I have searched and find conflicting information on every page I read. The new Mac Pro configuration guide does not address the question of "Best Performance" only that DIMMs must be installed in pairs of 1GB, 2GB or 4GB.
I want 6GB of memory now and more later. If I believe you are correct and buy 4x1GB DIMMs I will lock myself into a max of 8GB without replacing all the memory with 2 or 4GB DIMMs. So I need to know what the performance hit is if I use a mix of 1 and 2GB DIMM pairs. I had to chuck the 2x512MB DIMMs to upgrade my iMac to 2GB. Will I now have to chuck the 2x1GB DIMMs to upgrade the Mac Pro? Anyone, Help please.
Maybe I will do some test myself. 2x1GB, 2x2GB and 2x1GB + 2x2GB. Anyone know of a FREE memory benchmark?
I also find it odd that the 2GB machine ships with one DIMM on the top board and one on the bottom. When adding memory you must first move the DIMM from the bottom board to the top board and then add the new DIMM pair to the bottom board.
If there's a performance difference, its not big enough to be noticable. Don't bother with it too much.
This is for the old Mac Pro, but here it is:
http://www.anandtech.com/mac/showdoc.aspx?i=2816&p=12
I haven't seen a memory test that actually uses a program that can use four or more cores simultaneously, I think that's where the speed boost is. All or most of those programs use only one or two cores, and I think that's why the apparent boost is so low, it's being limited by the processor FSB and the program is only on one processor at a time.
That's news to me, and the docs do confirm that.
Thanks for the link JeffDM. I'm starting to understand. If I go with 2x1GB and 2x2GB then there will be 2GB, half of the 2GB DIMMs, accessible by only 2 FBD channels.
Do you think Apple places a DIMM on each board to ensure memory on each branch of the North Bridge?
I really don't know. I think it might be for more even cooling. I'm not sure if the northbridge even cares, but I don't feel like reading the Intel documents.