AMD Threadripper 2 & V-Ray - Has anyone tried yet?

[mitchgates]

Has anyone tested the 2990WX or 2950X with V-Ray yet?

[Magnum_design]

Here´s a chart from a youtubevideo I just watched... It´s a beast fror rendering and multitasking.

[mitchgates]

I saw that too. I'm wondering if this is the same benchmark used on the official VRay Bench site. The 7980XE clocks in at just under 27 seconds on that chart.

[Alex_M]

The 7980XE clocks in at just under 27 seconds on that chart.

Well, you're looking at the top result which means it was likely overclocked pedal to the metal, very likely with exotic cooling solution and to the point of failing only for the purpose of the benchmark before it shuts down from overheating. You'll never achieve such score under normal usage. Even the guy that did that benchmark says in his own comments that it was overclocked to the moon (the little comment icons to the right)

4.8 Ghz, 93°C (dangerously high temps for a CPU!), loud, LM (LM most likely means liquid metal aftermarket thermal paste such as Thermal Grizzly Conductonaut so it was also likely delidded which voids your warranty and can destroy your CPU)

. The real result of 7980XE (not overclocked, loud and smoldering hot) is 37 seconds, not 27 seconds:

[jacksc02]

I just got my 2990wx's today and they are amazing with vray.
I have upgraded 4 "old" 1950x's with 2990's

[Alex_M]

Congrats! Post some Vray benchmark results, will ya! I searched on the official results page and there were none done with 2990WX.

[mitchgates]

Well, you're looking at the top result which means it was likely overclocked pedal to the metal, very likely with exotic cooling solution and to the point of failing only for the purpose of the benchmark before it shuts down from overheating. You'll never achieve such score under normal usage. Even the guy that did that benchmark

Well, that certainly puts a new spin on things I guess I've been putting too much trust in the "@ 2.60ghz" on those scores without digging into the user notes. I had been considering a refurb HP with dual E5-2696 v3 but incorrectly assumed it was scoring around the same as the stock 7980XE all this time so decided to wait for The Threadripper 2 instead. I may have missed out on a good deal on those Xeons in the process

I'm not able to access a desktop for the next few days and it seems that the little "notes" icons on the right of score details don't work in Safari on iOS. Trying to find a way to read those details to get a better idea of how these CPUs compare without overclocking.

[Richard7666]

There is only one result for the 2990WX on the V-ray bench site, and it's 30 seconds.

The median for a 1950X is about 45 seconds (and that's exactly what mine does it in, incidentally)

So it's not very impressive considering the 2990 is over 2x the price of the 1950, but it's only one result. Early days.

[Alex_M]

There is only one result for the 2990WX on the V-ray bench site, and it's 30 seconds.

The median for a 1950X is about 45 seconds (and that's exactly what mine does it in, incidentally)

So it's not very impressive considering the 2990 is over 2x the price of the 1950, but it's only one result. Early days.

That's certainly a bit strange of a result. In a lot of places I'm seeing 26-22 seconds for 2990WX. Maybe that user with the 30 seconds result didn't have his PC set up properly or was using the old motherboards. For example, the Threadripper 2990WX requires the new motherboards with beefier power phases like MSI MEG Creation X399 to unfold its full potential since it's a 32-core 250W CPU and the older motherboards have weaker power phases:

[psanitra]

I think 22-30 seconds shoud be correct. Cinebench score is 5000-6000 points depending on the core multiplier. 30-41 was tested by many, and it`s pretty much a choice of how much Watts you want to burn and cool. Der Bauer have good video on youtube and sweet spot for reasonable efficiency seems to be around 3.7-3.8 Ghz per core.

[Jiri.Matys]

Does anyone have 2990WX CPU and test it on some heavy max scene with vray please? Benchmarks are one thing, but heavy scene with much more RAM to use is another. Please is it ok, is CPU using whole RAM and is there some slowdowns due to four channel memory?

Thank you.

[jacksc02]

That 30sec result was my first attempt at the benchmark.
I noticed that windows wasn't letting me use 100% cpu due to a power management setting so I re-flashed the bios and adjusted windows power management and it has been running at 100% since.
Also I am running an older x399 motherboard so potentially that might slows things down slightly.

26-27 secs is what I am getting on the vray benchmark without overclocking. So with some overclocking that should come down. Early days

On scenes with large memory usage the 2990wx takes a little bit longer to get going than the 1950x but once the memory is loaded then it races through.

If anyone has a scene that they want to try on the 2990wx's PM me and we can sort it out

[DPS]

Do you have any stats on temperatures and Power consumption? I forget the exact number but the TR1 throttled at around 61 degrees, or when consuming more that 178W if I recall correctly.

I just had confirmation that mine is ready to be picked up from the store - I'm debating on whether it's an upgrade of a new build. Your render times are looking promising. I don't like the fact that AMD have said that warranty is void if you OC the second gen TR.

[Alex_M]

Do you have any stats on temperatures and Power consumption? I forget the exact number but the TR1 throttled at around 61 degrees, or when consuming more that 178W if I recall correctly.

AMD Ryzen CPUs use two different metrics to report the temperature so it's important to specify which metric you are talking about - Tctl or Tdie? The temperature at which 1950X throttles is 95°C Tctl (which is 68°C tDie). To understand the difference between Tctl and Tdie you may want to check this article. I highly recommending reading it but long story short, for Threadripper Tctl temperature is tDie temperature +27°C.

If a monitoring program (like HWMonitor for example) doesn't say specifically whether the temperature it displays is the Tctl or Tdie temp, then it most likely displays the Tctl temp (for which 1950X throttles at 95°C). Here's a screenshot of a program that shows the temperature as both Tctl and Tdie (notice the 27°C difference between the two). And here's example of a program that only says "CPU temperature" without specifying if it's Tctl or Tdie which typically means that it displays Tctl temperature.

I've got my 1950X running perfectly fine without throttling even when it gets to 80-85°C (as measured by HWMonitor).

As for wattage, I don't think it will throttle when consuming more than 178W like you said. I regularly see 300W-350W power draw (as measured by HWMonitor) when I run my CPU at 3.9Ghz on all 16 cores and it has never throttled.

[jasonhuang1115]

Hey Alex, what cooling solution are you using for your 1950x? I am using two fans and couldn't pass 3.8GHz while keeping it under 95C Tctl.