Blur Busters Forums

Very interesting!
Thanks for posting.

Yeah, I tend to forget how everything changes from build to build when it comes to latency.
There's just so many variables (sample-to-sample variances in each component of the PC), plus the differences we're trying to measure are so small, that this is bound to be the case.

A Solid lad wrote: ↑

10 Feb 2022, 09:36

Very interesting!
Thanks for posting.

Yeah, I tend to forget how everything changes from build to build when it comes to latency.
There's just so many variables (sample-to-sample variances in each component of the PC), plus the differences we're trying to measure are so small, that this is bound to be the case.

In my experience, the biggest, most consistent factor where raw "average" (but not necessarily distributional) latency and gaming is directly concerned is typically the game engine.

That said, there's no doubt there can be latency differences from system-to-system, but again, yes, it's highly relative to the mix of components and software being run, as well as the given user configuration (sometimes regardless of platform), so that can make it difficult (often impractical) to get a 1:1 comparison in every situation.

I think the real answer to "Is it true Ryzen has higher input lag than Intel?" is "It depends." That doesn't mean it can't be discussed, I just don't think there's any "one" easy, globally applicable answer.

For instance, my G-SYNC article's latency results are globally applicable, yet the "total" latency of each scenario will vary per system. The only constant in that case is the "difference" between those totals.

.

BTRY B 529th FA BN wrote: ↑

10 Feb 2022, 23:53

Interesting, jorimt! Thank you for posting.

I'm trying to run it on my 5800X machine and it's running but it say:
*** Unable to access mlcdrv.sys to modify prefetchers
*** So, enabling random access for latency measurements

It also says that on my i7 970 X58 rig. I'm running it from my downloads folder just by clicking the mlc.exe The mlcdrv.sys driver is also in the same folder. Using mlc v3.9a

EDIT: Also can one make a direct comparison between running MLC on Intel VS AMD or does there need to be some changes made when running it? What I mean is with the inability to access mlcdrv.sys to modify prefetch; assuming I'm running MLC right.

I provided instructions on how to properly run MLC much earlier in this thread:
viewtopic.php?f=10&t=7033&p=53073&hilit=mlc#p53073

It should work the same on an AMD system as it does on an Intel system, as far as I'm aware.

.

On the topic of latency measurements, does anyone know who I can contact or where I can get an LDAT?

So, I can confirm, that Zen3 platform is very strange. In csgo I have input desync, for example - cant bhop. For guys, who dont know - you need to scroll mouse to jump and press ADADAD and do it in same timing.
First of all - no WHEA errors, no TM5 errors, all stability tests (prime, linx, aida, occt) passed. PC is bought about half year ago.
I turned off RAM XMP profile, used just 2400mhz 1.2v default profile - desync disappeared. OC'ed manual to 3000, 3200, 3600 with 1.35v (like XMP) - some have desync, some not.
Deep inspecting is not possible, because IF (memory controller) is integrated to CPU, some motherboards have problems with RAM OCing because of layers thickness (lolwut?) and so on. And to inspect problem I need to get new MB, CPU and RAM to be sure. And new ryzens will be on new socket, so this is not cost my money and time. And i7-12700K outperforms r9-5900x at the same price, and you still have i9 for upgrade. And AMD offering to turn PCIe4 back to PCIe3 to eliminate USB bugs.
Maybe ryzen is ok for someone, but I'm better switch to Intel back.

.

BTRY B 529th FA BN wrote: ↑

14 Feb 2022, 11:18

Are you using any BCLK ocing with these numbers?

The G.SKILL (both DDR4 and DDR5) were using XMP (the OLOy didn't support XMP), and neither the 8700k nor the 12900k were manually OCed in those captures. They were at stock with default MultiCore/turbo boost enhancements (both ASUS motherboards) enabled.

The 12900k is barely worth OCing past stock (for gaming) unless you're going for benchmark scores. I'm already getting a ~230w draw from the CPU at stock 5.1GHz turbo in Cinebench R23, for instance

.