AMD Threadripper 2990WX vs 3970x

[marko.zeko]

Hi!
I have recently bought AMD Threadripper 3970x and tried to compare rendering times with my 2990wx config.

So configurations are:

1.
CPU: AMD threadripper 2990wx
COOLING: Coolermaster masterliquid ML360 TR4 edition
MOB: Gigabyte x399 Aorus pro
RAM: 64Gb Corsair Vengeance LPX DDR4
GPU: ZOTAC AMP! Extreme 1070 8 Gb
DISK: WD SN750 m.2 500 Gb
PSU: EVGA supernova 750 G3 750 W

2.
CPU: AMD threadripper 3970x
COOLING: Coolermaster masterliquid ML360
MBO: Asus ROG STRIX TRX40-E gaming
RAM: 64Gb Corsair Vengeance LPX DDR4
GPU: ZOTAC AMP! Extreme 1070 8 Gb
DISK: ADATA SX8200 Pro 512 Gb
PSU: Thermatake Toughpower Grand RGB 850W Gold


Test scene was archexteriors 34 scene 10. Render was started out-of-the-box. No tweaking.

2990WX rendered it in 61 mins, while 3970x finished it in 19 mins.

As far as I know 3970x is not 3x faster than 2990wx.
I think that something is not OK with my 2990wx.
Does anyone has experience with 2990wx?
What could be wrong with 2990wx config?

[Karol.Osinski]

Wow! That's a huge difference damnit. They say the new gen is much faster and more stable but this seems strange...

[Neilg]

Your liquid cooling might have shit the bed.
Happened to all of our new workstations, we had to take it out and go with air. Install the ryzen master tool and check the temperatures / cpu throttling when rendering. Windows performance still registers this as 100% of available cpu use, even if your processors are capping at 2ghz.

We're going to back to air across the board - between NYC and Miami we are currently at about a 70% failure rate within the year on 30 workstations.

[super gnu]

Your liquid cooling might have shit the bed.
Happened to all of our new workstations, we had to take it out and go with air. Install the ryzen master tool and check the temperatures / cpu throttling when rendering. Windows performance still registers this as 100% of available cpu use, even if your processors are capping at 2ghz.

We're going to back to air across the board - between NYC and Miami we are currently at about a 70% failure rate within the year on 30 workstations.

given i can safely assume your watercooling didnt actually shit the bed, what happened to it ? ive used watercooled workstations for the last 15 years and never once had any problem. are we talking AIO cooler or custom loop?

[Neilg]

given i can safely assume your watercooling didnt actually shit the bed, what happened to it ? ive used watercooled workstations for the last 15 years and never once had any problem. are we talking AIO cooler or custom loop?

AIO cooler. CPU temperatures rocketed up and the ryzens started to throttle themselves for temperature control - they will not allow themselves to go above 68 degrees out of the box. They were all Enermax coolers, built mostly at the same time but some a couple months apart. No issues with thermal paste or how they'd been installed.
The fans were running, so we dont know what happened inside, but they were not functioning as a cpu cooler whatsoever - which to me is shitting the bed.
Took it out, replaced it with a heatsink and fan setup and the cpu's run at 100% speed again.

The problem being described here and the cooling system being the same type makes me think it's worth installing the ryzen monitor to see if it's the same issue. that cpu should be rendering faster, it's being throttled.

This is what mine looked like - this was grabbed while rendering.

[super gnu]

yeah you are not he first person ive heard complaining about enermax AIO.. a 70% failure rate is quite impressive!

[Neilg]

it's astonishing. outright negligence. fortunately ryzens do what they can to make sure they dont cook themselves so we only lost a few frames long term.

I know coolermaster is better but maybe this is that .1% for them.

[bildform]

I just changed my enermax aio for a "Noctua NH-U14S TR4-SP3" air cooler. I went from 1000 mhz to 3400 to 3600 mhz at around 65° C in a warm room. This is for a 1950 Threadripper.
Now i am researching a cooling solution for the 3970x. It has 100 Watt more that have to be cooled. I will never buy enermax again.

https://www.youtube.com/watch?v=g-YgNSclpC8 This happened to the enermax aios.

[marko.zeko]

Problem was in memory disposition.
I had 128 Gb of memory in first config. After I got 3970 x I took out 64GB and put it into new config.
Remaining RAM modules were put in first four slots. So only two channels were used. Memory clocks were set to default / auto.
In previous configs setting to auto meant that would work at declared clocks, but in both AMD configs that meant clocks were set to 2133 MHz even tough memory was 3000 GHz.

After I put modules one to each four channels and clock was set to 3000 GHz my vray benchmark result was 27000 and mentioned scene render time was 35 mins.
Which is huge difference compared to first result but difference is still 1,6 x (vray benchmark) and 1,8-1,9x (evermotion scene) slower compared to 3970x.

[Ihno]

The fans were running, so we dont know what happened inside, but they were not functioning as a cpu cooler whatsoever

Here you go:
https://forums.chaosgroup.com/forum/...71#post1042471

Problem was in memory disposition.
I had 128 Gb of memory in first config. After I got 3970 x I took out 64GB and put it into new config.
Remaining RAM modules were put in first four slots. So only two channels were used. Memory clocks were set to default / auto.
In previous configs setting to auto meant that would work at declared clocks, but in both AMD configs that meant clocks were set to 2133 MHz even tough memory was 3000 GHz.

After I put modules one to each four channels and clock was set to 3000 GHz my vray benchmark result was 27000 and mentioned scene render time was 35 mins.
Which is huge difference compared to first result but difference is still 1,6 x (vray benchmark) and 1,8-1,9x (evermotion scene) slower compared to 3970x.

I have awful experiences with ripping ram kits apart / mixing them. This can cause a lot of trouble down the road. Like blue screens/freezes and stuff like that.

[PIXELBOX_SRO]

To add to the mess the ram kits for ryzen need to be arranged in a different way than kits for intel.

[marko.zeko]

new interesting observation.


The other day 2990wx config rendered evermotion scene in 35 mins.


Today with no change in BIOS (CPU clock and RAM clocks are the same) it took 27 mins for the same scene with same settings.


Only thing that was different I have restarted comp just before rendering. This is 20% faster.