Gx00 driver bug fixes...

[Humus]

CAS 2 increases you system performance by 5-8%, using CAS 3 when your RAM handles CAS 2 is just idiotism. It's like clocking down you 500 MHz processor to 450MHz just to be safe.

[Admiral]

My PC100 SDRAM won't handle CAS 2 even at 100 fsb, but it sure can handle a fsb of 138 and would post up to 144

[Wombat]

Kosh,
I happen to disagree with you. I accept that the CAS3->2 performance boost is roughly 3%, but on average that's all you usually get from a 50MHz CPU boost.

[Kosh Naranek]

Going from CAS3->2 does give you a boost but it's only in some special cases like Q3 you are able to measure any real difference, in most cases you can't measure any differnce at all.Whereas a 50 Mhz cpu boost can be measured in many more cases / apps / games.

[smokva]

CAS2 can improve performance in apps like Seti@home very much.
It doesn't improve anything in gaming.

------------------
ABIT BX6v2
Celeron300A@450
Alpha PAL6035, 27CFM fan
64MB SDRAM 6ns
G400 16MB DH
Yamaha DS-XG
Hitachi GD2500BX+

[dZeus]

Quake3Arena is VERY memory-subsystem dependant, so CAS-2 will _definately_ improve q3a scores.

This is also the reason why q3a kicks so much ass on a P4 1.4GHz with dual rambus memory interface.

[Humus]

So, where's the benchmarks?
Here they are:
3dmarks2000 @ Athlon 500, G400 16MB, 128MB PC-100 SDRAM

CAS-2, L2 enabled: 2454
CAS-3, L2 enabled: 2333
CAS-2, L2 disabled: 2385

See, CAS-2 gives a 5.2% performance boost. Going to CAS-3 gives a higher performance drop than disabling the L2 cache!

BTW, Q3 is much more dependent on fillrate than memory subsystem.

[Kosh Naranek]

Humus
At what resolution are you testing ?
Is it 1024x768x16 bit ?

[Kosh Naranek]

Humus
The reason I ask is because 3DMark 2000 is not consistent in its 3DMark Scores......
And the difference between worst/best case can be 5%.
I'm not saying it is but it can be...
BTW the difference on my p3@770 and a K7@1050 is 3432 against 3550 both running in 1024x768x16 bit so the G400 is close to it's maximum here.

[Rags]

Kosh, they are right. CAS2 does give you a decent performance boost. And it isn't just in selected apps, it's pretty much across the board. It's kind of like getting a bit extra RAM without having to go out and buy it.

Rags

[Gurm]

I don't mean that I always LEAVE my RAM at CAS3 (did I say I did? oops!) but rather set it that way when setting the machine up and testing/overclocking.

However, none of this addresses the fact that Haig advised us to make our RAM run FASTER (which some RAM can't even handle) rather than turn off buggy code, and I was wondering what possessed an otherwise rational being to advise such a thing?

- Gurm

------------------
Listen up, you primitive screwheads! See this? This is my BOOMSTICK! Etc. etc.

[Haig]

GURM,

I did not advise anyone to do anything. I simply asked people to try something out and give me some feedback.

Gentlemen,

This was not aimed at pointing the blame towards anyone else. All I wanted was some feedback in order to help solve our dual cpu issues with the abit boards.

Forgive me for even thinking about asking this here.

Haig

[This message has been edited by Haig (edited 18 September 2000).]

[RAB]

Sorry Haig, I didn't mean to stir up a hornets nest. I was just a bit amused that Gurm was using CAS 3 and was trying to get one of his "calm" responses. Sorry Gurm, I just couldn't resist.

Haig, please continue to ask questions here. As you know, we are collectively pretty ADD and often go off on tangents. We will try to stick to the subject from now on.

AS far as CAS 2 helping on dual CPU systems, I'll leave that for someone else since I only have single CPU machines.

RAB

[Humus]

Originally posted by Kosh Naranek:
Humus
At what resolution are you testing ?
Is it 1024x768x16 bit ?
----------------
Humus
The reason I ask is because 3DMark 2000 is not consistent in its 3DMark Scores......
And the difference between worst/best case can be 5%.
I'm not saying it is but it can be...
BTW the difference on my p3@770 and a K7@1050 is 3432 against 3550 both running in 1024x768x16 bit so the G400 is close to it's maximum here.

Yes, it's 1024x768x16. The score's is consistent for me, I've runned 3dmarks2000 a lot of time always getting scores about the same. It's like +/- 10 or something like that.

BTW, I tested UT later yesterday. I knew UT is very cpu dependent so I guessed it would gain a lot more, and it sure did!

UT @ 800x600x32, with everything posssible turned on.

CAS-3: 35 fps
CAS-2: 39 fps

That's 11.4% performance boost, like upgrading my Athlon 500 to Athlon 550.

[Gurm]

No no Haig, it's a useful question, it's just not relevant to what I felt was a helpful post about the bugginess of the P3 optimizations. That's all. Mayhap a separate topic would have been the way to go with it?

- Gurm

------------------
Listen up, you primitive screwheads! See this? This is my BOOMSTICK! Etc. etc.