Render Speed (CPU Cores VS GHz)

[moifa]

I currently have two workdstations:

1) HP Xeon E5440@2.83Ghz (2x Duo Core, 8 threads), 4GB RAM (Window Experience Index 5.9)

2) Dell Xeon E5630@2.53Ghz (2x Quad Core, 16 threads), 12GB RAM, SSD running at SATA2 (WEI 7.1)

 

I am rendering the exact same scene with VRAY 2.3 in Maya 2013. The scene use the same irradiance map and light cache map saved locally in the machine. The VRAY settings on both machine is the same except the dynamic memory limits are set to 3000 (for HP) and 10000 (for Dell). Max render threads are set to 0 (use all threads available).

 

Which machine do you think it will render faster?

 

Most people would think Dell should kick HP ass, but in fact HP render about 11% faster.

 

Render result:

1) 7m8.4s

2) 8m4.9s

 

What am I missing here? Does the CPU Ghz play such an important role? or in general, HP is a better machine than Dell? The Dell machine is much newer than the HP one. I just don't understand why HP is faster than Dell with 4 less cores .. :banghead:

[Scott Schroeder]

Of course GHZ plays an important role. That is like the age old question, what is faster. 8 1-GHZ machines or 1 8-GHZ machine? It could also be a better motherboard, better cooling, and so on. If the Dell gets too hot (which Dell's are notorious for poor airflow), then the chip will auto throttle back to stop it from burning out.

 

There could be a host of extra factors going into your findings. But just from the surface, I'm not surprised at all that the less core but faster GHZ machine won.

[Matt McDonald]

Your HP has the same number of real cores as the Dell, higher clock speed and faster memory. So it is not surprising to me that it's faster. The processors in the Dell are hyper-threaded which can provide a performance benefit in certain situations, rendering typically being one of them, but its something like a possible 10% increase in speed. So it's not like you are really comparing an 8 core machine against a 16 core machine.

[RyderSK]

I currently have two workdstations:

1) HP Xeon E5440@2.83Ghz (2x Duo Core, 8 threads), 4GB RAM (Window Experience Index 5.9)

2) Dell Xeon E5630@2.53Ghz (2x Quad Core, 16 threads), 12GB RAM, SSD running at SATA2 (WEI 7.1)

 

than Dell with 4 less cores

 

 

The highlighted numbers are wrong. As Matt above me said, they are in fact all 4 quad-cores, so it's 2x4 vs 2x4(2) with the additional virtual threads being due to Hyper-threading, which might contribute less than the higher clock. It's very close though.

 

Neither machine is particulary fast though.

[moifa]

Thanks for correcting me. I always assume they both have hyper-threading. Yeah, these two workstation I bought the Dell second hand and HP is free gift from my friend, so they are more than 4 years old.

 

I did few test batch renders and have some interesting result.

 

I change the render region division under Vray settings. Here are the results:

 

HP

region 64x64(default) > 7m 19.9s

region 32x32 > 4m 50.8s

region 16x16 > 5m 3.5s

 

Dell

region 64x64(default) > 8m 4.9s

region 16x16 > 3m 29.1s

 

The division size effects the render time significantly depends on how many threads you have. Smaller not necessary better when you can see HP is actually slower when render at 16x16.