RAM Speed Increases TVAI Processing Speed

Topaz Video AI General
1

So I played around with trying to make TVAI as fast as possible with my setup, and I decided to share my results with everyone in case this helps anyone.

My setup:
CPU : Ryzen 5800x (16 Logical Cores)
GPU : Rx 7900 xtx
RAM : G.Skill 3600 C16 32GB (Originally 16 GB)

Just as an FYI all results were with Iris ( v1 for those in the future ). I used the Windows task manager to monitor how utilized my different components were. I know some users do not like this monitor but when I double checked it with other monitoring software I found the results comparable.

Originally when I was rendering a video from 1080 to 1440, my CPU would be utilized about ~75 to 80 percent and my GPU would be utilized about ~38 to 43 percent. According to TVAI I was processing about 3.2 frames per second (FPS).

I started to think about how to speed up my processing for frames per second. I realized after a little bit I never enabled pc RAM’s XMP on the motherboard (O.K. this is an AMD platform so D.O.C.P. for those who want me to be technically correct). I found after enabling the RAM’s XMP TVAI had a jump in performance. The CPU started sitting at about ~80 to 83 percent utilization while the GPU was at ~48 to 55 percent utilization. Processing FPS jumped up to 3.8 which was about 18 percent (17 - 18 percent from 3.2 FPS? ).

After seeing how much the ram speed increased the processing FPS I decided to start using dual channel RAM configuration. I found the results was other 10 percent increase in processing FPS. The CPU still sits at about ~80 to 83 percent utilization while the GPU is now at ~55 to 65 percent utilization. The processing video FPS is now about 4.2.

So from my experience RAM speed makes a big difference. I know this should not come as a shock to anyone. 3.2 to 4.2 processing video FPS is a big jump. I don’t know why my CPU will only be utilized at the max by ~80 percent. I thought maybe originally that my CPU was data starved but that does not seem to be the case.

I decided to process two different videos at the same time and see if that would better utilize everything. I found my CPU will then run at 100 percent utilization, and the FPS of the processing videos run about ~2.4 FPS each which is equivalent to 4.8 processing FPS.

In conclusion, I found that RAM speed makes a big difference and running two different copies of TVAI will also improve efficiency. I don’t know why TVAI can’t use 100 percent of the CPU when two copies of TVAI can.

P.s After a AMD driver update the GPU is now running at ~ 48 to 55 percent utilization but still 4.2 FPS video processing. I wonder how much more optimization AMD can make with its drivers.

2 Likes
2

I’ve been wondering if it would be worth it to buy a 3600MHz kit. I’m on 3200MHz right now. Or should I higher to like 4000MHz.

3

ddr5 6000 is recommended

1 Like
4

There is no ddr5 6000 on Ryzen 3000 or 5000 series.

32000mhz is the sweet spot for 3000

3600mhz is the sweet spot for 5000

1 Like
5

I do have a 7000 series, and the RAM is not 6000MHz.
@menditsa That’s what I was wondering.

6

I thought the slowest speed DDR5 was 4800mhz ?
Are your memory settings in the BIOS correct ?

7

Sorry for the confusion.
I have two computers. One is Ryzen 5000 the other Ryzen 7000.
So 3200MHz and 5600MHz on the RAM respectively.

The only thing I’m left wondering is: would even higher memory speeds increase TVAI speed more?

8

Thanks for clarifying, i thought I was going mad.
I can’t comment on the 7000 as I didn’t upgrade to it.
I run a 5950X with 64gb of CL16 at 3600mhz
Prior to that I had 32gb of CL16 at 3200mhz
Can’t remember the exact figures but the gain in TVAI was sub 1 fps.

9

My experience, as always YMMV.

I’m an Intel user, so my Gigabyte Aorus Master MB uses Z690 chipset. I am running an i9-12900KF overclocked to 5.3GHz and cooled with a Corsair H150i Capellix LCD (360mm AIO radiator). My GPU is an EVGA RTX-3080Ti FTW Hybrid Gaming (AIO cooled with a 240mm radiator).

I started out with DDR-4800 (DDR goes down DDR-4000,BTW) which runs at 2.4GHz with 4.8GT/s. When I upgraded to DDR5-6400 (3.2GHz and 6.4GT/s) I picked up between 1 and 4 fps, depending on model and input file resolution. On a percentage basis, the improvement was between 2% and 15%, with the biggest increase in processes that were CPU heavy and/or which ran at below 10fps.

I understand that the relationship between memory, CPU and MB chipset is more complicated in the AMD world since Ryzens are tuned for optimum performance with specific memory clocks, but thought the framerate increases experienced with a substantial increase in memory clocks might prove illuminating. I would expect memory clock upgrades will have a limited effect on Ryzen CPUs when going beyond the given “sweet spot.”

1 Like