Does V-ray care about RAM speed?

[Richard7666]

Something I've often wondered. Does anyone have any insight into this?

Is the same amount of memory on Threadripper at 2400mhz going to be much different to a kit at 3200? Because I've been reading some things where running RAM at different speeds makes a *huge* difference. This guy was using Threadripper, so it may be specific to Ryzen/TR, but:

Just wanted to share my experience for anyone who's in a similar situation:
I have a 1950X with Asrock Taichi board.
Just for fun, I tried the ripjaws ram at motherboard default speeds of 2133mhz. It was awful. Played a few games of Heroes of the Storm, and the framerate hit lows of 60s. In comparison, with 3200mhz the lows were around 110 fps (most of the time it ran @ 130-140 or so during fights).

https://www.reddit.com/r/Amd/comment...nd_ram_speeds/

Any idea whether this sort of thing would carry over to multi-threaded applications like V-ray? Presumably not to the extent it affected that guy's gaming performance, but I'd certainly be curious whether anyone has run the benchmark with different RAM speeds selected (or different kits altogether for that matter)

[delineator]

would be curious too, but I would 100% chalk up his findings to the zen platform (and probably being exaggerated - his claims seem a bit dubious). Remember that almost EVERYTHING including the core to core stuff on ryzen/threadripper depends on fast mem speeds. Which is sort of nice and weird at the same time: very few applications actually care about mem speeds, and typically, the only time you'd see a difference even between single/multi channel is on benchmarks. But now, it matters again which is nice/weird.

[Richard7666]

I have an i7 5930K which I'll do a test on today. Change the xmp profile from 2400 to something lower, and post V-ray benchmark results.

As you say though, it may be Ryzen specific.

[delineator]

I have an i7 5930K which I'll do a test on today. Change the xmp profile from 2400 to something lower, and post V-ray benchmark results.

As you say though, it may be Ryzen specific.

Any updates?

[Richard7666]

Any updates?

Yep.

There is a small difference on i7 at least.

V-ray Benchmark averages about 2:05 on 2400mhz DDR, and 2:07 with the same RAM running at 1800mhz.

[delineator]

Eh, I would take that with a mountain of salt - thats well within the margin of error, and these tests are notorious for varying quite a bit. In the morning I could cinebench at 1400, in the afternoon, closer to 1600. Just different processes running in the background, etc etc.

But even so, its not like it hurt performance and it didn't help out either, which should be good news for those looking to add in another stick or get some cheaper ram for their render nodes.

cheers!

[Richard7666]

Yeah, it's negligible, however I ran the bench several times with each configuration (rebooting the PC between each obviously). I'd have put it down to margin of error as well if it weren't for it being consistent.
Again though, it's really no indicator of whether the same is true on Ryzen, which is what I'm most interested in.