Mac Pro 09 8x2.26GHz 5870 to GTX680/7950 worth it with stock CPUs?

[Mattww]

I normally upgrade after 5 years but surprisingly I'm still pretty happy with my 09 Mac Pro so I'm looking to keep it until there is a bigger gap to the latest models and I can decide which way to go Mac Pro/iMac next time.

I am still considering upgrading the CPUs if I spot a bargain but the stock ones are still doing OK with productivity work so I'm mainly concerned with keeping gaming performance decent with up coming titles. All the benchmarks seem to be done with the highest clocked cpus so has anyone upgraded the humble 2.26GHz model and found it worthwhile?

 

[echoout]

I normally upgrade after 5 years but surprisingly I'm still pretty happy with my 09 Mac Pro so I'm looking to keep it until there is a bigger gap to the latest models and I can decide which way to go Mac Pro/iMac next time.

I am still considering upgrading the CPUs if I spot a bargain but the stock ones are still doing OK with productivity work so I'm mainly concerned with keeping gaming performance decent with up coming titles. All the benchmarks seem to be done with the highest clocked cpus so has anyone upgraded the humble 2.26GHz model and found it worthwhile?

I just upgraded a stock 2x2.26 to 2x3.33 6-cores and my Geekbench results doubled.

 

[pastrychef]

I think this image will give you an idea of how CPU will affect performance. The tests were done on 2.4GHz and GTX 670 then 2.93GHz and GTX 670 then 2.93GHz and GTX 680 everything else was the same.

 

[666sheep]

All depends on apps, acceleration type and kind of usage.
In OpenGL (gaming etc) CPU clock has much bigger impact on graphics performance than in OpenCL and CUDA.

 

[Mattww]

I think this image will give you an idea of how CPU will affect performance. The tests were done on 2.4GHz and GTX 670 then 2.93GHz and GTX 670 then 2.93GHz and GTX 680 everything else was the same.

If I'm reading those correctly there was a little improvement 1% to 20% with the CPU change but wasn't much change going from the GTX670 to the GTX680?

I think the GTX670 is a step above my Radeon 5870 though so maybe that is why the GPU didn't have the effect expected?

 

[pastrychef]

The improvement from upgrading the 2.4GHz CPUs to 2.93GHz were far greater and much more significant than GTX 670 to GTX 680. I knew that the GTX 670s were good performers based on reviews I had read. Plus, the GTX 670 I had was slightly overclocked from the factory. My primary reason for upgrading to the GTX 680 was for the EFI boot screen which I did not have with the GTX 670.

This review of the GTX 680 has benchmarks that have the Radeon 5870 included for comparison so you can see the type of improvement you can expect if you decide to go that route.

 

[Mattww]

The improvement from upgrading the 2.4GHz CPUs to 2.93GHz were far greater and much more significant than GTX 670 to GTX 680. I knew that the GTX 670s were good performers based on reviews I had read. Plus, the GTX 670 I had was slightly overclocked from the factory. My primary reason for upgrading to the GTX 680 was for the EFI boot screen which I did not have with the GTX 670.

This review of the GTX 680 has benchmarks that have the Radeon 5870 included for comparison so you can see the type of improvement you can expect if you decide to go that route.

Based on that review the GTX680 can potentially give around twice the performance of the 5870 which would make it a worthwhile upgrade, however they used an extreme edition CPU that was over clocked which would give much better single thread performance than I have.

 

[pastrychef]

While the 2.26GHz CPUs will probably be a bit of a bottleneck, I think the GTX 680 would still represent a significant upgrade to the 5870.

 

[Mattww]

I really need to restart when next convenient to get the most accurate readings but I have run both of the benchmarks.

With the Valley 1.0 Extreme preset I got:

FPS 27.2, Score 1139, Min FPS 13.8 and Max 45.0

This seems about right given that Barefeats got 29.8 with a 5870 in the 3.33GHz Hex.

With the Heaven 4.0 Extreme preset I got a rather terrible:

FPS 14.7, Score 370, Min 5.6 and Max 44.3

This seems out though as Barefeats show 34.8 with a 5870 in the 3.33GHz Hex?

I'm on 10.9 with 12GB of RAM which would no doubt have an effect.

 

[pastrychef]

For the Unigine Heaven benchmark, Mac OS X 10.9 Mavericks added a functionality called Tesselation to OpenGL that was no present in previous version of OpenGL under OS X. Therefore, to get a better comparison to Barefeats' results, turn off the Tesselation feature.

 

[Mattww]

Thanks with Tesselation off I get:

FPS 33.0, Score 831, Min FPS 7.8 and Max 53.4