Blur Busters Forums

Slender wrote: ↑

13 May 2025, 08:59

You can play with the numbers as much as you want. I just said that I had to downgrade 2 drivers to make the problems go away.

You can also play around with words as much as you want
That's where the term "verifiable test data with large sample size" comes into play, because anyone should be able to replicate said performance uplifts from a driver change.
Computers are simple input & output devices, results should be replicable if given a repeatable test environment (e.g.: the CS2 benchmark map)
You also mentioned 566.03 as a good driver for the 3060ti and 537.58 as "good drivers" for CS2, without any context or data to back the claims up.
I'm interested in you showing me a source which can back your claims up.

https://youtu.be/QW_XUkcIzTE?si=n6m0L0mtBu1TjuDh

That is a completely useless test which is not repeatable and is not even close to remote representation of what happens in realtime...

Here's how data should be represented at minimum. Relying on overlays which average values over x ms such as MSI/RTSS is not a good way to represent real-time data.
This image not related to this subject, just a screenshot from a user comparing tREFI in a Vulkan API test)

kyube wrote: ↑

13 May 2025, 09:05

Slender wrote: ↑

13 May 2025, 08:59

You can play with the numbers as much as you want. I just said that I had to downgrade 2 drivers to make the problems go away.

You can also play around with words as much as you want
That's where the term "verifiable test data with large sample size" comes into play, because anyone should be able to replicate said performance uplifts from a driver change.
Computers are simple input & output devices, results should be replicable if given a repeatable test environment (e.g.: the CS2 benchmark map)
You also mentioned 566.03 as a good driver for the 3060ti and 537.58 as "good drivers" for CS2, without any context or data to back the claims up.
I'm interested in you showing me a source which can back your claims up.

https://youtu.be/QW_XUkcIzTE?si=n6m0L0mtBu1TjuDh

That is a completely useless test which is not repeatable and is not even close to remote representation of what happens in realtime...

these tests don't mean anything at all.
I rely only on my own feelings, nothing else.
at 566.36, my game was slowing down and literally freezing.
at 566.14, the freezes disappeared, but I was faced with a loss of fps.
on the 566.06 driver, I have no problems with fps at 14 and freezes at 36.
Naturally, I will play on 566.06 because I feel better.
The only figure I pay attention to is the DPC. I know the values that should be there, and I know the values when something is wrong.

a lot of people recommend driver 537.58 for cs2, noting good p1 fps.

Slender wrote: ↑

13 May 2025, 08:38

kyube wrote: ↑

13 May 2025, 08:28

Slender wrote: ↑

12 May 2025, 20:29

Coldplayer wrote: ↑

12 May 2025, 19:47

Anyone here have a consensus on the best cs2 driver?

537.58

Are you able to substantiate any of these claims you're throwing around with verifiable testing data, with a decent enough sample size?

you can find video with comparision driver in cs2 on youtube. But for me, im just test driver only about dpc. If dxkernel / nvidia driver not spike above 50, that driver be ok for me, but some driver is broken.
im play a lot on 566.36, it cause very big stuttering in hard cpu games like cod, im decrease to 566.14 it fixed my stutters, but that not good driver (cuz bad perfomance, can google it). 566.06 very good for me (good perf/no dpc spike).

how does 566.06 compare to 537.58 for you? im still on 537.58 its the only driver which doesnt cause my latency to randomly spike in latencymon. using a 3070

Slender wrote: ↑

13 May 2025, 09:18

a lot of people recommend driver 537.58 for cs2, noting good p1 fps.

thats only count if the driver supports that card right ?
i might be wrong but 5080 for example wont go below 572.16 without issues.

OostBlokBoys wrote: ↑

13 May 2025, 14:55

how does 566.06 compare to 537.58 for you? im still on 537.58 its the only driver which doesnt cause my latency to randomly spike in latencymon. using a 3070

LatencyMon is terrible at evaluating DPC/ISR driver performance on Windows.
Refer to: viewtopic.php?t=14578#p115130 where I've discussed as to why that is the case.

Use 1st party solutions hosted by Microsoft, such as Windows Performance Analyzer or 3rd party scripts relying on .ETL's such as XTW or xperf

Lukas_ wrote: ↑

14 May 2025, 04:59

Slender wrote: ↑

13 May 2025, 09:18

a lot of people recommend driver 537.58 for cs2, noting good p1 fps.

thats only count if the driver supports that card right ?
i might be wrong but 5080 for example wont go below 572.16 without issues.

Of course, if the video card came out later than the driver, there is no point in trying to install it. It is possible, but pointless.
test with the first available driver for your video card.

kyube wrote: ↑

13 May 2025, 08:49

I don't care about this "old wives tales" type of arguement, nor do I care about "u can find video" arguement when 99% of users online don't have a clue on what makes up performance.
I was looking substantiated data with PresentMon and/or XTW, xperf or anything ETL-based... but it seems that you do not possess such data & instead of that, someone should rely on your current mood on which driver is "better"

Hey, Kyube. It's seems you're knowledgeable on the subject and that's cool, but there ain't no reason to reply in such a condescending manner. Simple question, simple answer, move the f*** on with your lives, get off to the numbers and graphs on your quad-monitor setup or whatever it is that you do.

Imma go with @Slender's recommendation or figure it out myself if it fails.

I have just registered on this forum to say this. Do better. And if you possess the power to help someone, do it without the mockery and rambling.

gg

aymancader wrote: ↑

13 Oct 2025, 19:10

Hey, Kyube. It's seems you're knowledgeable on the subject and that's cool, but there ain't no reason to reply in such a condescending manner. Simple question, simple answer, move the f*** on with your lives, get off to the numbers and graphs on your quad-monitor setup or whatever it is that you do.

Imma go with @Slender's recommendation or figure it out myself if it fails.

I have just registered on this forum to say this. Do better. And if you possess the power to help someone, do it without the mockery and rambling.

gg

You should read these forums more often to understand why my manner is variable depending on the type of post in question.
I also fail to understand why you've necro'd such a old post in the first place, but let's move past that...

The user you're trying to defend (let's assume you're not one of the 50 different alt accounts which "spams" nonsense here) has shown clear lack of interest in approaching this topic in a matter of how Windows performance troubleshooting is supposed to be handled.

If someone doesn't want to read the official, recommended and "bulletproof" way on how to figure out performance problems on Windows & then continues rambling clear nonsense without any sources or proof, what else is left other than point out the errors?

These things being discussed here aren't intangible, nor something abstract.
They're practical, concrete & specific software "problems" which can be "fixed" using methods described by 1st party sources (Microsoft Nvidia, AMD,...), most of which are written by HW/SW engineers
"Feelings" are prone to human (physiological, psychological, environmental) bias, therefore cannot be used as extrapolation onto every possible hardware & software combination.

Either way, majority of these "problems" purported around here have been solved ages ago or lack incentive to be tested properly due to lack of knowledge & lack of dedication.
Trolling seemingly sells better.

Have fun with whatever you're trying to do.