Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Hi Chad,

Here's what your 2009 2.66 GHz machine looks like next to a 2006 2.66 GHz just for fun and comparison:

TT_My_XBench.jpg
TT_Chads_XBench.jpg


Yours is the one with the yellow hot-help peeping through. ;) You might have to expand your browser a bit to see both side by side.
 
CINEBENCH R10
****************************************************

Tester : doox00

Processor : E6750
MHz : 3200
Number of CPUs : 2
Operating System : OS X 32 BIT 10.5.7

Graphics Card : NVIDIA GeForce 9600 GT OpenGL Engine
Resolution : <1920x1080>
Color Depth : <fill this out>

****************************************************

Rendering (Single CPU): 3246 CB-CPU
Rendering (Multiple CPU): 6088 CB-CPU

Multiprocessor Speedup: 1.88

Shading (OpenGL Standard) : 5882 CB-GFX


****************************************************


There a way to compare to other cinebench results.. not seeing that option in the software.
 
Hi Chad,

Here's what your 2009 2.66 GHz machine looks like next to a 2006 2.66 GHz just for fun and comparison:


Yours is the one with the yellow hot-help peeping through. ;) You might have to expand your browser a bit to see both side by side.

Thanks Tesselator. What HDD is the '06 model using? The '09 isn't a ton faster though. Do you think these programs take advantage of the new Nehalem processors? Looks like the GTX is doing its job though. :)
 
Thanks Tesselator. What HDD is the '06 model using? The '09 isn't a ton faster though. Do you think these programs take advantage of the new Nehalem processors? Looks like the GTX is doing its job though. :)

The drives used are three HD154UI Samsungs in a RAID0. See: https://forums.macrumors.com/posts/7673878/

And I think all software "takes advantage" of the Nehalem the same as all other Intel CPUs but XBench is fairly poorly written. GeekBench is MUCH worse tho. Processor features are usually fairly transparent to user software. You wouldn't want a benchmarking program to be tuned for a specific processors feature set unless you were only testing different speeds of the same model. :p

The differences between these CPUs should bench about like you see here tho - so kudos to XBench for that. Where you get your speed in that part of the system is in RAM access and cache utilization. Look at those differences. Your RAM details for example are like four times faster than mine. :eek: And that translates into something like a 20% real-world speed increase over mine. Of course against the 2008 2.8 you'd not see 20% - closer to 0% but your clocks are lower too. :p

Don't pay attention to the score column tho. Look instead at the details column. :D
 
I would like to see a ssd boot drive in chads system and run the test again. The hard drive is a bottleneck on his system.

Oh its coming as soon as Intel releases their new SSD's. I would love an extreme 120GB Intel SSD. Especially if the new ones are going to have even faster write/read speeds. I want an Intel SSD for the boot drive/apps and then three 1TB caviar black WD hard drive's in RAID 0 for all my media. :)
 
I am seriously considering one but the prices are just to dam high. 400 bucks for a hard drive is a bit to spend but were all have the need for spped. Probably do my mac book first to see the difference. With your memory test score, bet a ssd boot drive will scream.
 
Well, I'm thinking about 2 Intel extreme's in bay 1 and bay 2 w/ RAID 0. I believe that would boot pretty fast. :) How did the stock 2600 register higher than the GTX285? I'm misunderstanding these tests I think.
 
Hey guys,
I finally got my 2009 Mac Pro in.(Stats in signature) So, after getting everything up and running I decided to run Xbench and thought I would post the results. Now, this is after I installed 6GB of RAM from OWC and the Nvidia GTX285 graphics card from Apple. The GTX285 is using the latest driver off of Nvidia's website. I gotta ask one quick question can anybody look at the picture of my temps and see if those are okay? The temps screenshot was taken while playing watching a youtube video, typing this, and listening to music in iTunes. BTW, the room temperature is 74 degree's Fahrenheit not Celsius. :)

I'll let you know my temps when I get home. I have the same setup except with 3 HD's. For starters, I know your case ambient is lower (mine hovers around 90-92F IIRC).
 
Basically the 2006 (your model) is a 8 core upgraded?

So it means its comparing against a 1st gen 8 core model vs. 8 core (16 thread) Nehalem.

Wow barely any difference!
 
Well for right now. My system is upgraded except for the stock HDD which is making it an even bigger bottle neck. Let me get my Intel SSD in there and it should make for a big difference in numbers, or I hope. :)
 
Tesselator can answer that one. I'm still waiting to see if Intel releases their new SSD's. If they would hurry I would have this thing humming by later next week. But, I need at least 250GB for a boot drive/apps. :eek:
 
Relating to my temp question. I grabbed another screenshot of them in Celsius.
 

Attachments

  • Picture 2.png
    Picture 2.png
    55.4 KB · Views: 307
Thats surprising. Is the Apple drive a single platter?

Tesselator can answer that one. I'm still waiting to see if Intel releases their new SSD's. If they would hurry I would have this thing humming by later next week. But, I need at least 250GB for a boot drive/apps. :eek:

No idea. Find the model of the drive in question and look it up on-line. I would be a little surprised if it was but that doesn't mean much. - I'm a little surprised every morning when I awaken to find the planet is still here or that I'm still on it. :D

Also 250GB of boot, OS, and Home data is about right. If it's an SSD this may not apply (I dunno) but for rotational media you typically don't want to fill it over about 50% ~ 60% full. So ideally, with rotational media, if you had 250GB of data you would want a 640GB drive. After that performance begins to degrade and some people even claim fuller drives are more prone to data loss and breakage - which makes sense when you think about it.
 
No idea. Find the model of the drive in question and look it up on-line. I would be a little surprised if it was but that doesn't mean much. - I'm a little surprised every morning when I awaken to find the planet is still here or that I'm still on it. :D

Also 250GB of boot, OS, and Home data is about right. If it's an SSD this may not apply (I dunno) but for rotational media you typically don't want to fill it over about 50% ~ 60% full. So ideally, with rotational media, if you had 250GB of data you would want a 640GB drive. After that performance begins to degrade and some people even claim fuller drives are more prone to data loss and breakage - which makes sense when you think about it.

I agree Tesselator. But would rotational media apply to a SSD? Technically couldn't it fill up and never loose speed?
 
I agree Tesselator. But would rotational media apply to a SSD? Technically couldn't it fill up and never loose speed?

I don't have one so I haven't tested anything but using logic from my understanding of how they work, you wouldn't lose any speed at all from them being 99% full. You would however (maybe) kill the empty portion of the drive MUCH faster tho. If I understand it correctly, Wear-Leveling slash Dynamic wear-leveling uses the available free space and the incoming data to equally wear each of the physical blocks. This helps to reduce Block Retention time errors, Raw Bit Error Rates (RBER), and Uncorrectable Bit Error Rates. (UBER). Static data is not included in the available pool of wear-leveled blocks, leaving a portion of the drive with no wear and a portion of the drive (the free space) hyper-worn. :p I read about a new feature called "Static Wear-Leveling" - which considers all physical blocks in the SSD, regardless of content, and maintains an even level of wear across the entire drive but I don't think it's been implemented into any currently available products yet. I could be wrong tho. :D

So with current SSDs you still want to keep them as empty as possible. Emptier they have a good UBER tho I forget exactly what it is. The manufacturer site will list it. Half full the UBER spec would logically be cut in half. And at 90% full the UBER would only be 10% of the manufacturer's listed specification. :p

There's also two kinds and I dunno what products use which. There are Multi-Level Cell (MLC) and Single-Level Cell (SLC) NAND devices. MLC has two bits of data stored per memory cell and four states [erased (11), two thirds (10), one third (01) or programmed (00)] and needs logic level correction via EDC (Error Detection and Correction) whereas SLC is only 2 states and one data bit. EDC, Ware-leveling routines, and a number of other factors contribute to longevity or device durability but SLC fairs much better. The program and erase operations of SLC NAND Flash last 100,000 cycles - ten times longer than MLC NAND Flash, whose program and erase operations only last 10,000 cycles. After 10,000 and 100,000 cycles respectively, the reliability of NAND’s program and erase operations are not guaranteed. :D Because of it's more simplistic design Samsung is also claiming that their SLC based products will be (or are?) six times faster. Six time is massive and so is a 10x UBER, so make sure you select the right type. They're still too expensive for me so I've not looked into the situation at the product slash model level yet - meaning I dunno what's actually available nor who's using which technology.
 
Xbench
Results . . . . . . 209.75
System Info
Xbench Version . . . . . . 1.3
System Version . . . . . . 10.5.7 (9J61)
Physical RAM . . . . . . . . 16384 MB
Model . . . . . . . . . . . . . MacPro4,1
Drive Type . . . . . . . . . . Hitachi HDE721064SLA360​
CPU Test . . . . . . 214.61
GCD Loop . . . . . . . . . . . . 348.36 . . . . . . 18.36 Mops/sec
Floating Point Basic . . . . . . 201.72 . . . . . . 4.79 Gflop/sec
vecLib FFT . . . . . . . . . . . . 134.31 . . . . . . 4.43 Gflop/sec
Floating Point Library . . . . . 297.15 . . . . . . 51.74 Mops/sec​
Thread Test . . . . . . 845.82
Computation . . . . . . . 1009.74 . . . . . . 20.46 Mops/sec, 4 threads
Lock Contention . . . . . . 727.69 . . . . . . 31.31 Mlocks/sec, 4 threads​
Memory Test . . . . . . 340.63
System . . . . . . 326.37
Allocate . . . . . . 344.35 . . . . . . 1.26 Malloc/sec
Fill . . . . . . . . . . 221.82 . . . . . . 10785.35 MB/sec
Copy . . . . . . . . . 561.86 . . . . . . 11604.96 MB/sec​
Stream . . . . . . 356.19
Copy . . . . . . 353.95 . . . . . . 7310.73 MB/sec
Scale . . . . . . 350.60 . . . . . . 7243.27 MB/sec
Add . . . . . . . 362.38 . . . . . . 7719.58 MB/sec
Triad . . . . . . 358.04 . . . . . . 7659.37 MB/sec​
Quartz Graphics Test . . . . . . 279.86
Line . . . . . . . . . . 274.55 . . . . . . 18.28 Klines/sec [50% alpha]
Rectangle . . . . . . 344.24 . . . . . . 102.77 Krects/sec [50% alpha]
Circle . . . . . . . . . 284.16 . . . . . . 23.16 Kcircles/sec [50% alpha]
Bezier . . . . . . . . . 264.64 . . . . . . 6.67 Kbeziers/sec [50% alpha]
Text . . . . . . . . . . 248.70 . . . . . . 15.56 Kchars/sec​
OpenGL Graphics Test . . . . . . 248.74
Spinning Squares . . . . . . 248.74 . . . . . . 315.55 frames/sec​
User Interface Test . . . . . . . . 548.58
Elements . . . . . . . . . . . . 548.58 . . . . . . 2.52 Krefresh/sec​
Disk Test . . . . . . 65.88
Sequential . . . . . . 130.74
Uncached Write . . . . . . 150.90 . . . . . . 92.65 MB/sec [4K blocks]
Uncached Write . . . . . . . 152.79 . . . . . . 86.45 MB/sec [256K blocks]
Uncached Read . . . . . . . 86.95 . . . . . . 25.45 MB/sec [4K blocks]
Uncached Read . . . . . . . 168.85 . . . . . . 84.86 MB/sec [256K blocks]​
Random . . . . . . 44.03
Uncached Write . . . . . . 14.96 . . . . . . 1.58 MB/sec [4K blocks]
Uncached Write . . . . . . . 144.31 . . . . . . 46.20 MB/sec [256K blocks]
Uncached Read . . . . . . . 94.56 . . . . . . 0.67 MB/sec [4K blocks]
Uncached Read . . . . . . . 154.16 . . . . . . 28.61 MB/sec [256K blocks]​

CINEBENCH R10
****************************************************

MHz : 2.93
Number of CPUs : 16
Operating System : OS X 32 BIT 10.5.7

Graphics Card : NVIDIA GeForce GTX 285 OpenGL Engine
Resolution : 2560x1600
Color Depth : 24bit

****************************************************

Rendering (Single CPU): 4065 CB-CPU
Rendering (Multiple CPU): 25939 CB-CPU

Multiprocessor Speedup: 6.38

Shading (OpenGL Standard) : 5816 CB-GFX


****************************************************


Comments
Seems the Xbench 'Text' test runs rather slow on the GTX 285. Weird.
Xbench crashed a few times just after the text test. Needed a restart or it would crash at the same spot again.

The stock 640GB drive is no speed demon either.
 
How's your gaming experience with it? I have only played L4D and it just pegs at 60fps at max settings, which is the highest it can register I believe. Also, the 640GB HDD is rather slow compared to many others. I'm still waiting on Intel. I might just go ahead and order an extreme drive next week but I know the minute I do they will release the updated drives. Remember Nvidia is still trying to get the drivers just right. The new ones are alot better but I figure another release and snow leopard and we will see the GTX285 really be a great card! :)
 
Oddly enough I just ran Cinebench again(just the open gl test) and after a fresh reboot and no applications open it pegged at 6800. Its being run on a 22" Dell monitor thats resolution is set to 1920x1080.
 
I don't really do any gaming that requires high frame rates. So can't really answer that.
(Got the card mostly for CUDA programming and Maya.)

Cinebench renders its tests at 800x600. Your screen size will likely not have any impact if:
a) it is higher than 800x600.
b) free VRAM leaves the OpenGL test enough room to breathe, which I'd guess all modern graphics cards do.

Re HDD:
I think I'll wait for a good 512GB SSD and then swap to that and use the 640GB as extra backup. (I've got a 2TB disk for Time Machine plus two external FireWire HDs for secondary off-site backups.)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.