Please - I need every single person using CAT 6.2 to see this (and ATI reps)

jpinard

I'm posting this here and at Rage because I think it's very important.

I have found conclusive proof the CAT 6.2 drivers causing a major extra drain on the cpu, causing any cpu bottlenecked game or computer to run significantly slower (by up to 20)! If you have a high-end Radeon card (I have an X1900XTX) and a slower processor, this means CAT 6.2 may be hindering you more than helping you (vs. CAT 5.13). I've tested this in 3dMark (versions 2003, 2005), PC Mark, Rome Total War, and Falcon AF. Even though this is a fresh Win XP sintall (fully updated) I still need other people who have compiled their test results, to weigh in on this to [color=#cccccc][color=red]corroborate[/color] [/color]data.

In the following Excel spreadsheet you can see I began testing to see if a minimal or maximual X-Fi software install would function better with my new Creative Labs X-FI sound card. Data didn't look right with my old build, so I started digging. Please look at my test results here: [color=red]http://jeffpinard.com/Sound%20Test.xls[/color]

Below is an annotated graph demostrating the difference. Bigger version here: [color=red]http://jeffpinard.com/Graph3.jpg[/color]

DosCervezas

You say you have "conclusive proof", yet the only data you have is from your computer. I would not expect anyone to get too bent out of shape about this until you really do provide "conclusive proof". Not saying it is not happening, but I think you would get a better response from the community if your "proof" had a little more meat to it.

__________________
"Never argue with an idiot. They will only bring you down to their level and beat you with experience".

jpinard

EXCUSE ME? Look at my test results. These are on independant clean installs. Nothing else on them and the order of install went as follows:

1. Win XP SP2
2. Mobo chipset drivers
3. Office XP install
4. XP updates.
5. Office XP updates.
6. X-Fi Xtrememusic driver/software install.
7. Catalyst driver install.

I have 1-6 on a complete single file binary image I can blow down in 10 minutes from an external USB drive with FAT32 (NTFS not viewable in DOS via ghost).

I can then install CAT 5.13 from the CD for my X1900XTX and test... the record the results, upload them to the web, delete my Win XP partition, load up the previous image, and then install CAT 6.2

Results are 100% reproducable with the only change the version of CAT installed.

If that's not good enough for you, why don't you come over to my house and do it yourself - or better yet, why don't you spend 20 hours like I have isolating this issue and post your own scientifically reproducable result?

Zardon

jpinard, please calm down a little. I think the point he was trying to make was that your results are from a single system and it *may* be possible that your findings are the result of something not related to ATI drivers.

Obviously i cant speak as I havent tested it first hand but ill look into it.

Veridian3

Well the first thing i would do is:

Try a real game to see if this also happens in something like fear. (so run the timedemo with 6.2 and 5.13 to see if the results are much higher with 5.13).

Also, i cant see what your system spec is... i'm wondering if you have a dual core cpu and maybe the driver optimisations for dual core have a bug in them when you combine dual core +X1k+X-fi.


Let us know what you find in a timedemo.

Necrosis

Can you post the url for your 03 bench /w the 6.2s?

Also might want to include a detail list of your system's specs.

jpinard

Sorry I forgot to post my detailed system specs (see end of post). I'd like to note it would be very odd, for this to be just on my system. I've been testing for the past 2 weeks (Since I got my new mobo and vid card), and finding these results was accidental when ATI initially released the drivers. I would not have known... had I not kept a careful log of all my test results.

Necrosis, the link I provided has all the exact data from my tests. But I'll re-run it and upload it to Futuremark if that's what you want (last week none of my 3dmark scores were being accepted since they had not updated orb to accept the X1900).

I'd like to mention I have tried a "real" game... and I know it's a much better indicator than FEAR, Quake, Far Cry etc. As I mentioned before, I tested this with fraps in Rome Total War, Falcon 4 AF, and now Silent Hunter III.

The only test someone else can reliably duplicate is utilizing the "Bismarck's Revenge" scenario in Silent Hunter III. Load it up, attach the camera to the top of your sub, aim the camera at the Bismarck and start recording. In this intense battle the cpu is fairly pretty stressed, and I've consistently seen a 12% decrease in fps. My results in Silent in Hunter III are not perfectly linear to the results in 3dMark as one "might" conclude, but I hypothesise... it's the difference between the increase in graphic speed and the decreased cpu speed. ie. the 8 minute segment does not stress the cpu ALL the time, and when the cpu catches up, the overall increase in CAT 6.2 graphic performance cuts down the fps loss. Does this make sense? I'm a little disappointed so many "experts" continue to use benchmarks that are not cpu bound which skews results for those of us playing something OTHER than first person shooters.

OK, my detailed system specs:

* Asus A8R MVP hardware version 1.02G
* Asus A8R MVP Bios revision 4.02
* Mobo utilizes Uli chipset verson 1.0.5.2
* Using onboard Marvell Network controller v8.41.1.3
* 2 Gig Patriot 3200LL (1x1) running in dual channel mode 2-3-2-5 1T
* ATI X1900XTX testing on CAT's 5.13 (off manufacture's CD) and 6.2 off ATI's site.
* Seasonic S12-500 revision A2 PSU
* 120 Gig Samsung SATA driver on primary SATA controller
* 300 Gig Seagate drive in Venus SATA carrier (USB external enclosure)
* MS .net framework 1.1 plus service packs (MS .net framework 2.0 not installed)
* DX 9.0C December 2005 build
* Creative Labs XTrememusic, hardware revision C - latest drivers off Creative website.
* Klipsch ProMedia 5.1 THX setup in analogue format.
* Pioneer DVD-106 DVD RW
* Mitsubishi Diamond Pro 2060u 22"CRT running 1280x960 32bb 85Hz desktop (same as testing)
* Win XP SP2 and OFfice XP fully updated

If that's not enough info, I can even post a list of all my active processes and services (I've been keeping track through all stages of testing and install).

I used to do genetic research... I'm a VERY thorough person. Needless to say this cannot be an "it's only your system" type of issue. Sure, Far Cry runs faster now... but since when was that game bottlenecked by the cpu?

I'm not angry at ATI... but I am irritated by people who come in here and say "Hey it's you", without doing any relevant testing to justify the response.

jpinard

[color=black]I'm including another Excel spreadsheet that covers the past 8 months. In it you will find the first test which has my Asus A8V Delxue (note look at cpu scores), and then it progresses to the Asus A8R-MVP, various overclocks (OC) with different cpu dividers etc. What's most important is the very last test. This is the one with CAT 6.2 - even with my previous motherboard and 20 installs ago (I do a clean build almost every month) NEVER, did the cpu score drop so dramatically. [/color][color=black]http://jeffpinard.com/3dMark%202003%20-%20FSAA%20&%20AF%208%20month%20span.xls[/color]

[color=black]I urge people to test this themselves. Test a game that's cpu limited - not some first person shooter, and run different benchmarks. Try PC Mark too... the results are consistent (at least on AMD processors). I can go back to 5.13 drivers and the cpu loss is fixed...[/color]

jpinard

Adding 3dMark 2005 tests. This chart contains raw data from many builds (testing), but the relevant comparative results are in bold red and should pop up automatically when you open the file.

[color=red]http://jeffpinard.com/3dMark%202005%20-%20FSAA%20&%20AF%20v1.xls[/color]

Necrosis, here's my published results for 3dMark 2003 CAT 6.2 (more to come)

[color=red]http://service.futuremark.com/compare?2k3=4620926[/color]

Added 3dMark 2005 with CAT 6.2's (more to come)

[color=red]http://service.futuremark.com/compare?3dm05=1803709[/color]

tech.Me

Very interesting.

Maybe the cause to 4.12s running faster then 6.2s in my computer.

__________________

jpinard

Very well could be. It's being born ut by my data, and appears to be something was introduced with these v6.2 drivers (not sure if it's intentional or not) that really hurts processor performance.

Necrosis

I'm scoring quite a bit higher than you in all CPU test in 03 or 05. I don't know if this is because of different settings or what. I'm not OC'ing either.

Hmm..wonder if it's a bug between the drivers & your chipset?

H3X4D3C1M4L

Didn't notice mine change period

Veridian3

ok...

I've fired up and ATI motherboard, an FX-60, AIW x1900 and 1gb of mem. Sata HD etc and pretty much everything i could to make the same system as the one you've listed. The only major diff is i'm using onboard sound...

I ran the CPU tests from 3dmark06 and here are my results

5.13 6.2
CPU score 1920 1924

CPU t1 0.614 0.615
CPU t2 0.960 0.963


That leads me to believe that its possibly a clash with your soundcard drivers/hardware. Can you use onboard sound and see if the prob goes away?

jpinard

The ony way it could be a sound card driver issue is if ATI introduced something in CAT 6.2 that causes it. Like I said, everything is fine with CAT 5.13. I've run the cpu test alone with no sound card drivers installed and the results are the same (in the cpu scores).

Veridian I'd be curious to see what your 2003 comparative results are. 3dMark 2003's cpu test is more sensitive than 2006's. Also... what version(s) of .net framework are you using? 1.1 alone? 2.0 alone? or both?

Maybe this is isolated to single core users? I know ATI is slowly introducing some tweaks to take advantage of dual core prcoessors - and maybe this has hurt those of us with high-end single cores?

MindlessOath

if you havent already done so... ati has a feedback form that you can fill out. when filling this out it lets them know there MAY be an issue and send the beta testers to test it out.

the link should be in my signature but i belive it may be old, still its easy to find - just go get new drivers and you should see a link somewhere (at ati's website).

this still may be a related problem to your system only. you will need to test differntly... dont be stubern... you know its differnt with each driver version, now do the rest of the testing and find out what peice of hardware is causing this issue so you can help ati resolve it.

so... take everything out. just mobo, cpu, mem, video, obviously psu and cdrom drive. then do some testing. dissable all audio onboard and dont include any addin cards.

test with the default chipset drivers that came with your motherboard etc.... then download up to date ones for that chipset. now test again - yes each time with 5.13 and with 6.2

now insert peice by peice and keep testing.

hey you want this fixed right... dont be stubern just do it (i say this cause its a common thing people get stubern on issues like this that they bring up). ya its alot of work but it has to get done to find the issue at hand and help ati resolve a problem for you and mabe even a lot of other people.

ati may look at this as a great fix for something and give extra performance that was lost.

ahem, also make sure that all driver settings are the same. its not unlikly that they used a new more CPU intense algarithim in the catalsyt AI, so dissable cat a.i.

read and do all or no one is going to listen or help you.

and i would like to ask everyone else here with simular hardware (even if not) to test this out please - if ya have the time (mabe not so indepth but at least try something).

__________________
ATI Radeon X800XT PE <---- **LOOK UPDATED!!**
AMD Athlon 2Ghz, nforce2, 512mb ddr3500
LINKS U SHOULD KNOW!!!!
DriverCleaner - use every time u install new drivers!
ATI driver feedback - to report driver problems
Tacticalcenter - my Gaming website and forums
ATI 3rd party Optimized Driver Compairison

jpinard

I spent this morning doing just that. No sound card. Same results. I'm about to try a .net framework 2.0 only install to see if it's framework related. In my previous image I only had 1.1 installed. Granted I've heard CCC does not use 2.0, but someone else posted they did it and it works great.

I'm not a stubborn person either. If I were stubborn, I would not have put 30+ hours into this already to use process of elimination to narrow down the culprit. I think you misinterpreted part of my testing. It's "because" I've narrowed it down to CAT6.2 - not because it just happened when I upgraded that I've come to my conclusions.

Another note you should know. I'm already using the latest uli chipset drivers, because BIOS 402 for the A8R-MVP requires them (1052a). Being a Crossfire motherboard with an ATI chipset - I'd be terribly surprised if it was an issue because of ATI's own Crossfire motherboards.

I just wish there was a single core user with an X1900 that could run and PUBLISH these 3dMark 2003 tests too.

Veridian3

Just tried it with an FX53 and again, 6.2 are slightly better in performance for me. (both the first test and the 2nd used .net 1.1 with the latest patches).

Have you installed the AMD CPU driver?

jpinard

Veridian, can you post links to your CAT 5.13 scores and CAT 6.2 scores (3dMark 2003) at Futuremark? That would be wonderful. Thanks!

Also, is that with CAT AI on or off?

I know you said you used .net framework 1.1 - do you have 2.0 installed or is it still an install option via windowsupdate?

Thank you so much for helping out, I really appreciate it!

The_Neon_Cowboy

Are your PC mark scores are affected?

and at this point I remind you 3DMARK03 = IMO was and still is "Trash"
I'd rely more on other avilable benchmarks then what ever 3dmark03 says...

Also remember those CPU tests in 06 are optumized for Agias "physix" code
and that a genral CPU wont do nearly as well as a RISC CPU designed just
to do PPU work...

I don't really care much about 06's cpu scores mabye if they included
a more regular or genral CPU test one like in the past not "just" ones for
PPU use...then I'd give it more attention.

(Well back to my light saber wielding in KOTOR II)

__________________

H3X4D3C1M4L

Just discovering that game?

Look at how many hours I've pegged playing that game over and over

jpinard

Yes, my PC Mark scores are affected, as are my 3dMark 2005 scores as I've already shown.Hopefully ATI is tracking this. Here's my latest results. [color=red]http://service.futuremark.com/compare?2k3=4622512[/color] This is a MS .net framework 2.0 (only) install (no .net framework 1.1), no sound card, and CAT 6.2
CPU results:

CAT 5.13 ............ CAT 6.2 ................. % difference
133.8 ................. 101.1 ..................... -24.5 %
17.7 ................... 14.2 ...................... -19.8 %


I'm not sure how many times I have to say/prove this over and over and over, that cpu limited games are running slower with CAT 6.2 Why are people so abject to accepting these results without proving me wrong with their own tests? [color=red] [/color]
Doesn't anyone care their cpu-heavy games are running slower with CAT 6.2? Fear, Quake, and Far Cry aren't the only games on Earth, and I am very concerned with the single-minded nature of some (too many) people - only caring about first-person-shooter performance and saying "screw you" if you play flight sims, wargames, or other military simulations.

I should note you couldn't have a more ATI fan friendly person than me. I've purchased every generation of ATI card since the 9700, I've been around for years, and now I'm using ATI-based motherboards. I originally turned from NVidia because of their dishonest marketing practices (which they still do - paid shills) - but my God, where's all the help? Where's the data (except from me)?

Necrosis

Well that is the point man it's your machine. No one else is seeing this thus far. So there is no reason to get ticked off at them.

For one you're trying to prove your case with a benchmark. A benchmark..that is why people are having a hard time believing you. Especially on a benchmark that is like four years old. Most folks don't even own the Pro version of these benches. Go poke around other places here to find other folks results on those benchmarks. Look at their settings, and match them.

pr0digal jenius

try installing the driver without the CCC maybe?

Or, just out of curiosity, what else is running in the background? My guess is there's a logocal explanation somewhere, since your results don't seem to be matching what others are seeing

jpinard

I think it's a motherboard/CAT 6.2 issue. I finally found the closest match possible on Orb - it can be viewed here:[color=red] [/color][color=red]http://service.futuremark.com/compare?2k3=4620618[/color] This individual is running CAT 6.2 also, and has a slightly increased FSB.

I wish there was one person on this planet with an Asus A8R-MVP or comparable x200 based mobo to either verify or disprove my results.

I'd like to point out for the millionth time as well - that CAT 5.13's are perfectly normal... cpu tests, video tests are just fine. Necrosis, don't you think if it was a hardware issue it would be a consistent result across all drivers... NOT just in the CAT 6.2's?

jpinard

I've got NOTHING running in the background outside of normal base Win XP services and the ATI CCC. These builds I'm testing are clean enough to eat off. PLus I've doen several variations. A lot of people hated their Asus A8R-MVP's so maybe I'm just the dumbest person who never shipped theirs back I think if the CAT 5.13's hadn't worked fine it'd be easier to accept I had a bad motherboard.

jpinard

In case anyone is keeping score... here's what I've tested recently to try and alleviate the issue:

* Installed beta drivers from ULi - same poor cpu performance with CAT 6.2
* Installed latest AMD FX drivers from AMD - same poor cpu performance with CAT 6.2
* Re-ran all tests with CAT AI disabled - same poor cpu performance with CAT 6.2
* Ran driver only package (no CCC) - same poor cpu performance with CAT 6.2

This can only mean it has to be a driver related issue... I can't think of anything else I can change... except maybe I'll try replacing individual ATI system files?

jpinard

OK everyone, this is very interesting. I did two DRIVER ONLY (no CCC) installs of CAT 6.2 and CAT 5.13 (off CD) and the tests show the exact same results.

So this means it is NOT a CCC issue, but is most definately a driver issue. Throughout the last 20 hours I've made wildly different images with differing motherboard chipset drivers and through this the cpu tests are all the same.

If someone with the ATI driver team would just talk to me (heck I'll even let them call me collect so it costs them nothing) and tell me what changed, maybe we could figure out what's causing the problem.

Necrosis

Like I said earlier I do this it's some thing with your MB & the drivers. That is that only thing that makes sense. I wonder if it's DirectX specific though. Do you have any thing OpenGL to see if it does the same thing?

MindlessOath

right now this is more than likly an issue with your crossfire board... ie the crossfire function.

__________________
ATI Radeon X800XT PE <---- **LOOK UPDATED!!**
AMD Athlon 2Ghz, nforce2, 512mb ddr3500
LINKS U SHOULD KNOW!!!!
DriverCleaner - use every time u install new drivers!
ATI driver feedback - to report driver problems
Tacticalcenter - my Gaming website and forums
ATI 3rd party Optimized Driver Compairison