Video Enhance pcie performance

I have a pcie v3.0 bus motherboard with an rtx 3070.
Since the rtx 3070 is a pcie 4.0 bus, how much difference in fps can I hope to gain if I build a pc with a 4.0 pcie bus?

It makes no discernible difference. Better buy a second 3070 if your current system can handle it.

The PCI-e speed is all but meaningless here, and would only kick in when your GPU were to be fed data at max possible rate, all the time. And naturally that’s not the case (else your full-movie 4k upscale would be done in like 10 seconds).

Also, I wouldn’t get a 2nd 3070 (SLI is sooo 1990). Just get like a RTX 3080 Ti, for a very reasonable price these days, and get it over with. It’s really their computational speed that matters, not PCI-e bandwidth.

N.B. Ran some synthetic benchmarks myself, once, like setting your link speed to x8, instead of x16. Even that only makes makes like 1-1.5% overall differece (gaming).

I didn’t mean SLI. The 2nd card is recognized as an additional GPU and used by Topaz. Two 3070s are significantly faster than 1x 3080 TI. I myself use 2x 3080 Ti.

That’s interesting. My 1 RTX 3080 Ti is being used for only ca. 50-80% (with full saturation on all my 8 P-Cores of my i9 12900k). So, surprised a second card helps, as the first one isn’t even being used to the max.

On Windows I use 3 Topaz video instances. 3 videos are processed in parallel. I use version 2.6.4. Version 3x is extremely much slower.
Try to edit at least 2 videos in parallel!

1 Like

I don’t get how that works. Sure, it would get my GPU to near 100%, but my CPU is already at capacity too. So, what would be gained by the second process, if each process can now only use 50% CPU?

Start the Topaz program and start with the 1st video. After that starts open the Topaz program for the 2nd time. Now load the 2nd video and start editing. Overall performance will improve between 15-35% depending on PC performance. Important!! : Always only 1 video per program! Do not put several in a row! This leads to problems with the individual settings of the respective video.
Your CPU load is not always at 100% on an instance. 10-20% makes a big difference in the minimal breaks that your CPU isn’t at 100%.

Maybe not literally fully at 100%, but pretty darn close to it:

Disabling (‘parking’) the E-Cores yields the best performance (so 100% P-Cores): with E-Cores on, the latter just start to be used first, become saturated, leaving the P-Cores idling too much.

I will give it a try, though (after this job is done).

How fast does task manager update? Not fast enough .) Don’t trust the task manager. Especially not with GPU usage!
Set the Topaz program to “reduced machine load”. That’s good for me in general.

1 Like

N.B. That is, conceptually, actually the original idea behind hyperthreading. Except that VEAI already does an impressive job maxing out all P-Cores. :slight_smile:

I think everyone has to test it out with their own hardware. My information is based on experience, which shows me much faster results. Take the time and test this.

Task Manager updates every second, iirc. GPU loads I measured with GPU-Z.

But, indeed, with “reduced machine load” I might actually achieve the required ‘hyperthreading’ effect. Good tip. :slight_smile:

In one second your CPU has time to think 1000x how to get out of this damn PC case
 to freedom :slight_smile:

Lieutenant Commander Data: “0.68 seconds, Sir. For an android, that is nearly an eternity.” :crazy_face:

Maybe I’ll do a test once I get my storage drive back. This article shows that there is a possibility of a speed increase between PCIe 3 and 4 with some video processing software. Games—Not at all.

I’m also waiting for the processing speeds of TVAI 3 to stabilize. They weren’t constant last time I tried to do a speed test.

Not constant at all, really. Not sure that’s a big issue per se, though, as, like I said, VEAI, for starters, doesn’t produce nearly enough data to saturate the PCI-e bandwidth. My RTX 3080 Ti, for instance, offers a whopping 912 GB/second bandwidth. Such speeds are only really relevant for games. VEAI isn’t pusing out high-speed video, of course, but uses the GPU primarily for computational stuff. Theoretically, there could be a scenario in which VEAI could saturate the entire GPU too, but since the CPU-to-GPU ratio is overwhelming in favor of the CPU, it’s simply not possible for enough data to be fed to the GPU per second.

One can see each instance in the stack looks very different in taskmgr, but all of a sudden the GPU maxes out. One good thing though, you can use the computer at the same time :stuck_out_tongue_winking_eye:

Personally i need them both (versions).

What is your CPU and Ram ?

I have an i9 12900K + 64G RAM, plus an RTX 3080 Ti (with 12G).