Just upgraded to 10.9.3 to test the speed.
The BruceX test stay the same at 23-24 seconds with two HD7970s.
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 5.1 owner, make it your king of Mac with 2 D700s
- Thread starter Prince134
- Start date
- Sort by reaction score
Just upgraded to 10.9.3 to test the speed.
The BruceX test stay the same at 23-24 seconds with two HD7970s.
Try Luxmark. BruceX isn't pure OpenCL. When I upgraded from 10.9.1 to .2, I noticed ~10-15% performance drop in OpenCL benchmarks along with "XT Prototype engine" instead of "D700". It was never back to 10.9.1 level, hence my conclusion. On the other hand, 10.9.3 is lousy in OpenGL. About 20% performance drop compared to 10.9.2.
I use geeks3D GPUTest 0.7 JuliaFP64 and FP32 to measure this, because it has smallest CPU impact.
In general I don't like the direction in which 10.9.x AMD drivers go.
Last night, I used handbrake to convert one of my favourite movie "Prestige 2006" from MKV to MP4, then import it to Final cut pro 10.1.1. It was so fast. I know this was solely CPUs job.
While opening the activity monitor, the Final Cut Pro rendering the imported MP4 at background, I saw the system utilised my 128gb ram up to 108GB. It's like 2 hr and half long footage(not quite since it's cut), every single frame was rendered and stored in the whole 108GB, you can feel by browsing the "footage" as smooth as hot knife with butter.
And this is in 1080p, I'd like try some 4K footage soon. Since nMP claims to be smoothly rendering as that, I am excited to try. I don't think it will disappoint me.
That's what more ram can do for you. Who said 128GB is over kill?
And the dual HD7970, never forget....Final cut cMP user really should try upgrade these cards, it's so inexpensive now.
While opening the activity monitor, the Final Cut Pro rendering the imported MP4 at background, I saw the system utilised my 128gb ram up to 108GB. It's like 2 hr and half long footage(not quite since it's cut), every single frame was rendered and stored in the whole 108GB, you can feel by browsing the "footage" as smooth as hot knife with butter.
And this is in 1080p, I'd like try some 4K footage soon. Since nMP claims to be smoothly rendering as that, I am excited to try. I don't think it will disappoint me.
That's what more ram can do for you. Who said 128GB is over kill?
And the dual HD7970, never forget....Final cut cMP user really should try upgrade these cards, it's so inexpensive now.
Last edited:
I agree 10.9.1 is faster. OpenCL wise in 10.9.3 Luxmark drop within 100 points so 1-2% drop is the most from my test. I don't see any difference in BruceX test of Final Cut PRO.
This is to update to you guys that two HD7970 are in my machine and without external power. And it's silent! You only spend a fraction of money comparing to buy new MacPRO.
Meet the new King of Mac
Dual X5690 3.46GHZ, dual HD7970, 128GB DDR3 1333mhz (8X16gb), two 500gb SAMSUNG EVO at 1.0TB raid0.
3000$ for processor upgrade, 3500$ for the Mac Pro (the cheapest dual socket 2010 Mac Pro price) and 600$ for dual 7970. Total price 7100$.
A nMP with 12 processors and Dual D700 costs 7500$. How is 7100$ a fraction of 7500$?
You are off market.
I bought $2000 for the two X5690s. $2000 for Dual CPU 2010 Mac Pro model. $400 for Dual HD7970s. You can also sell original CPUs, video cards, and save up to $500 to $800 depending on what you have. For one can get a Mac Pro fully compatible to nMP (except TB) at half of price of $3500. Isn't 1/2 a fraction?
What about dual X5680s, only $1300. I think they can perform just on par with nMPs. That's $700 less.
Last edited:
So you bought a second hand Mac Pro and comparing the price of that to a new one. Obviously you gotta compare retail to retail.
And how do you buy two X5690's for 2000$ when the retail price for one is 1500$ on amazon.com right now?
http://www.amazon.com/gp/product/B004O0CXAW/ref=s9_simh_gw_p147_d0_i1?pf_rd_m=ATVPDKIKX0DER&pf_rd_s=center-3&pf_rd_r=0FJMPK07T7ZECB415ERQ&pf_rd_t=101&pf_rd_p=1688200422&pf_rd_i=507846
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
nMP is great, is as fast as max mac pro tower. So what's bothering you here? Maybe someday I will worship nMP, when I get one. You never know.
I have never against people from buying nMP. The calculation is just to give you an idea for what you can get with how much money. And the thread is to share with people how you can down clock and volts with dual D700s and put in their tower mac pro with internal power. Thus having their King of Mac.
If you don't plan to use mac pro tower, any opinions offered became off topic.
(side note: try other sources other than Amazon, you will believe my number)
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Care to link them? Newegg didn't have them in stock so I checked google. The only "in stock" link was amazon.com.
----------
What's bothering me is the advice. oMP maxed out is a great machine. But it's not worth buying today. If you have a dual socket 2010, maxed out, you can keep using it for a while since the new one isn't a big upgrade. But EOL machines have very low resale value. So any upgrades people do on their oMP's today are gonna be expenses they won't get back when they sell the machine.
Resale value is not important. The majority of people use computer until they break down and then throw them away or they give them to other people when they don't perform enough.
For some application/use case an omp is a better choice than the nmp and vise-versa. The only thing hindering the viability of the omp is the fact that Apple has the final say on when software support is to end. If the os was independent of the hardware people would keep their omp for many more years.
For example my 2010 iMac will probably not be supported by the next version of OSX if one comes out in 2015 since Apple tend to drop 5+ years old models. But that iMac still runs great and there is no real tech barrier that can be used to exclude it from using the next version.
No, i don,t care, since you are pretending yourself not knowing existence of Ebay. And you are 100% off topic.
I checked eBay. 1550$ was the cheapest price for it on eBay as well. And I'm not off topic. The topic is how much this machine costs. And I'm calculating it.
I guess that you only considered the new CPU. We are here to compare the performance only. And try to lower the cost as much as possible. Some of us are more than happy to use 2nd hand components with a reasonable price. If you can get a 2nd hand nMP which is cheaper (less than $3500) and faster than this dual X5690 + dual HD7970, please let us know. I am sure lots of us are interested in it. I really want to know the 2nd hand price and performance comparison.
On the other hand, why you insist to make your own calculation? Prince134 already told us his figure. If you fail to beat that price, it's your own problem, but doesn't mean that no one can build the same machine with that cost (or even lower). I just did my own search as well. Luckily I can read Chinese, and I can easily find a X5690 in China with about $850 USD ($5200 CNY, in case you want this link). So, I can further lower the cost to $3200 for building that machine. Should I say that the nMP is even more over price?
I just want to point out that there is no point to argue about the building cost. You can always find something cheaper or more expensive. And performance won't care about the machine is new or old, they are just the numbers telling us how fast it can run.
Anyway, I agree with you that it's not worth to buy the old Mac Pro in your calculation ($7100). If anybody can't find a good price like Prince134 did. They may be better to buy the nMP. However, if someone manage to do that, I personally think it's worth to build a performance matched machine for only half of that price.
If you are in the second hand market then your price doesn't apply to everyone. Maybe you managed to find some cheap component but it doesn't guarantee others can find it at the same price. That's why second hand pricing isn't a good indicator of how much a machine would cost to "everyone".
And again, comparing second hand price of the old one to the retail of nMP is meaningless. Second hand items being cheaper than new ones isn't a big surprise really.
----------
Yes but an old machine is more likely to break than a new one. Digital components certainly have wear and tear. So buying something old always comes with risks. So one should take that into account when buying a second hand computer.
That's the point! Your 1st hand eBay price doesn't apply to everyone either. You get a new X5690 with $1500, I can find one with just $850 (yes, that's new!). If you argue that my supplier may not be trustable, the same rule apply on the eBay seller. The market outside Apple is free, all price can going up and down. And no price can guarantee applicable to everyone. Even though Apple sell their products at difference price (in terms of USD) in different country. No price can apply to everyone.
Anyway, I agree that the 2nd stuff most likely will die sooner. And a self build Mac Pro like this most likely will have more problems than the stock nMP. And this is why there is a price difference. The cMP can't win in every aspect. It's cheaper, but more risk. If I want a new computer with than performance, I think I will go for the nMP as well. However, if I want something relatively cheap but the same performance, I will go for the 5,1.
I bought 2 @ used 5690s for $750/each. Took a little time and bidding, but that was more than 6 months ago.
The 3690 I got for quite a bit less.
Have had no trouble with any of them.
All used, all working just fine.
The 3690 I got for quite a bit less.
Have had no trouble with any of them.
All used, all working just fine.
That's not too bad. My GTX680 only scores 670s.
Can you also run the CPU + GPU? ( Mine is 5146 you can see at post #22). Because that might give us whole picture of the compute power for OpenCL including CPU involved in your nMP.
Also BruceX when you can, since I saw from somewhere that BruceX test spent only 21 seconds for nMP(not sure the configuration though, think is maxed. And forget about Hackintosh that it maybe crazily maxed to get 16sec ). This test in FCP 10.1.1 is a focus point that majority feel Final Cut Pro is hugely improved along with the debut of nMP. My GTX680 gave me poor result. It spent over 1 minutes and a half.
Last edited:
What I dont understand is...
ATI in Luxmark: ATI high win vs. Nvidia
Real Apps: ATI and Nvidia at the same level.
http://www.barefeats.com/tube05.html
Conclusion = Luxmark is a very bad bench to test the real world performance.
ATI in Luxmark: ATI high win vs. Nvidia
Real Apps: ATI and Nvidia at the same level.
http://www.barefeats.com/tube05.html
Conclusion = Luxmark is a very bad bench to test the real world performance.
Last edited:
I just did an update with more details with pictures and one important change in arranging both card's power plugs.
ALL OTHER READERS, PLEASE READ THE FIRST POST, AS IT"S BEEN UPDATED WITH PICTURES.
