PDA

View Full Version : Why is the VPR four times slower on a CPU that renders 2.3 times faster?



Mr Rid
11-02-2017, 12:37 AM
Does the VPR use the GPU or not? I cannot find authoritative info on this. One reference says it does, and another says it does not.

I am using a new, faster workstation at work, but the VPR is almost useless when it does not update for 4+ seconds, each time I move anything.

The VPR is slower on this faster CPU:

i7-7820X 8/16core overclocked to 4.5GHz
64GB DDR4 3200MHz PC4-25600
GTX 1080
- 2560 Cuda cores
- 256-bit-wide memory bus
- 320GB/s memory bandwidth
Win 10 Pro


And the VPR is faster on this slower CU:

Xeon E5-1650 v2 12core 3.5GHz
64GB
AMD FirePro D700 (FireGL V)
- 2048 stream processors
- 384-bit-wide memory bus
- 264GB/s memory bandwidth
Win 8.1

What could be slowing the VPR down?

Thanks!

Sensei
11-02-2017, 02:50 AM
Check your gfx card settings, wait for vblank or something like that..

Also check whether one machine has Draft mode on, and second regular mode.

Kaptive
11-02-2017, 02:57 AM
Xeons are quite different chips. I was just reading this on a comparison page... http://www.velocitymicro.com/blog/xeon-vs-i7i5-whats-difference/


Xeon Pros

L3 cache – CPU caches are like small batches of memory that the processor keeps close by to speed up certain applications. Most Xeon processors have 15-30MB of L3 cache depending on the model, close to double their i7 counterparts. This extra cache is one reason why Xeon’s are so much faster at high demand workstation applications than i7.

Support for ECC RAM – Error Checking and Correction (ECC) RAM detects and corrects most common data corruption before it occurs, eliminating the cause of many system crashes and translating to more stable overall performance. Only Xeon processors support ECC RAM.

More cores, multi CPU options – If your applications require as many CPU cores as possible, Xeon is what you need. The new Xeon v4 processors max out at 18 cores (36 after Hyperthreading) whereas even the new Broadwell-E i7-6950X has just ten. Multi-CPU configurations are also only possible with Xeon

So maybe it is something like the L3 cache? Total novice on such things, so just a guess... but I wonder if tracing the direct cause might be difficult without someone familiar with the code and how it works.

rustythe1
11-02-2017, 06:14 AM
are you using an asus motherboard, if so install ai suite and see if turbo is clocking in, there can actually be issues with a slight pause when going from base clock (usually as low as 1100mhz) to turbo, VPR uses CPU so your card should not be an issue,
for the X range of intels you should always use either ROG boards, or Deluxe boards as Asus have a proprietary deal with intel to utilise extra pins on the X range chip, any other board (like msi) and the chip is just a waste of money, ive seen up to 15% performance loss by using cheaper boards and just a guess but that could maybe also cause problems with turbo, it may not be engaging at all so the benchmark could be a lot lower than your xeon, (although i don't think the benchmarks are that far apart for those two chips)
also check your ram set up with the forums and manuals, its very important that they are in the correct order and correct slots, if they are 8gig modules that might also be a problem, its much better to use 4x16 gig modules in asus boards as they tend to not like all the slots filled.

rustythe1
11-02-2017, 08:59 AM
also forgot to say, the i7-7820X is known to throttle if its not water cooled, if you have an air cooler you may find your not getting above 3ghz

Danner
11-02-2017, 01:54 PM
VPR Does not use GPU. I've tested Lightwave on over 20 machines and it is pretty consistent in performance between VPR and regular render. Meaning VPR and render times are always faster on machines with a faster CPU. Have you done any temperature monitoring?

Mr Rid
11-07-2017, 02:08 AM
Thanks for replies!

VPR Q settings are same on both machines. Boost is working. The i7 is liquid cooled. The board mfg is Asustek, but the model does not show under System Summary, or directX Diagnostic.

Why would LW render faster, while the VPR is slower, unless they process differently? I do not see any info from Newtek or in documentation how the VPR works.

gar26lw
11-07-2017, 04:49 AM
I've been finding I'm getting a notable vpr slowdown on Xeon compared to i7 both with ssd and comparable ram. Both dell.