Hello

Sorry for not posting this sooner (as I said I would in the drivers forum), but I'm having an exam in less than a week...

Anyways, here's a short comparison of what speed difference I got when upgrading from Cat 3.2 to Cat 3.4.

It's only syntethic benchmarks, as I don't have any of the regular benchmark games installed, nor could I find the cd's

I tested with 5 different benchmarks:
3dmark 03, 3dmark 2001, Nvidia's TreeMark, Imagination Technologies VillageMark and Vulpines GLmark.

I chose the last three as they didn't take too long to download, gave easy to read results, and actually worked... Yes, some of them are old, so I really doubt we'll see any special optimizations for them, and thus that any speed difference should be applicable all across the board...


Test Setup:
A7V8X (BIOS 1011)
Athlon XP 2500+ (Barton 333Mhz)
1GB PC2700 (DDR 333) TwinMos CAS2
ATI Radeon 9700pro
Win2k SP3
Cat 3.2 and Cat 3.4

All test were run @ 1024*768*32.
I tested each benchmark with 2 different settings:
Fastest possible (FP), in which all sliders were set to "High Performance", AF and AA set to "Application default" and Vsync Allways OFF.
Highest Quality (HQ), in which all sliders were set to "High Quality", AF @ 16x Quality and FSAA @ 4x.

3dmark (both versions) and VillageMark were run @ default settings, whereas I used Max settings for GLmark (Note: GLmark seems to be geard mostly towards Nvidia with special settings available just for them.), and the following settings for Nvidia's Treemark:
-tltest -bpp32 -depth8

But enough about that, let's take a look at some numbers:
First off 3dmark 2001

Cat 3.2 FP got 14332, HQ 9340
Cat 3.4 FP got 14197, HQ 9848

As you can see (if not fromt he graf, then from the numbers), that you loose 135 points (-1%) when running with FP settings... But as soon as you try HQ you get an increase of 508 points (+5.4%)... Not too bad...

3dmark 03:

Cat 3.2 FP got 4625, HQ 2094
Cat 3.4 FP got 4907, HQ 2378

A nice increase of 282 (6%) and 284 (13.5%) respectively... Not bad at all!



I couldn't post it all in one go, due to maximum image per post constraints...

Over to the not-so-regular benchmarks:

Nvidia's TreeMark:

Cat 3.2 FP got 50,5919, HQ 50,3525
Cat 3.4 FP got 54,3744, HQ 51,0777

Whereas not too much in #, but it gives you an increase of about 7% and 1.5%...

VillageMark:


Cat 3.2 FP got 155, HQ 55
Cat 3.4 FP got 152, HQ 56

Once again we see a decrease when using FP, but an increase with HQ... The difference is -2% and +1.8%


Cat 3.2 FP got 156, HQ 56
Cat 3.4 FP got 152, HQ 57

Close to no difference from the first VillageMark test, which isn't too much of a suprise... (difference: -2.6% and 1.78%)

The difference between the two benchmarks are that the first test goes through 1800 frames as fast as possible, whereas the second test runs for 2 min...

Re: Re: Cat 3.2 vs Cat 3.4 (a small comparison)

Vulpine's GLmark:

Cat 3.2 FP got 76.3, HQ 51.5
Cat 3.4 FP got 76.5, HQ 56.2

Which gives us an increase of 0.2% and 9.1%


Cat 3.2 FP got 176, HQ 130
Cat 3.4 FP got 174. HQ 153

Which gives us the following, FP -1.2%, HQ 17.6%


Cat 3.2 FP got 41, HQ 24
Cat 3.4 FP got 39, HQ 25

Thus giving us -4.9% and +4.1%


I also ran Final Reality, but it didn't save the results when I told it to... But it seemed to get an increase of between 5-10% (though since I don't have the results, I can't be sure)


After looking at these numbers, it's not hard to see that ATI is working on giving us more perfomance with higher IQ...
But if the increase is enough to upgrade... Well, without any game testing it's hard to tell...
Though I will say that an increase ranging from 1% to 17% isn't bad at all...
Yes, you'll loose some points with performance settings, but when you've got a card as powerful as this, why aren't you using it?!?

If you are having problems with any of the things mentioned here, then download it!
(Just make sure you're getting the driver for you OS, the link is for windows 2000)

Just remember one thing if you choose to upgrade:
Keep a copy of your current drivers so you can re-install them if anything goes wrong.


Hope this was informative for you, it sure was for me...

Nice work Mr.Garibaldi.
Although if I may say that if you didn't you should have run the FP and HQ benchmarks several times each and taken average of the results. This would give a better overall impression of what can be expects as I have found in the past that benchmarks vary a little everytime they are run even if no changes are made to the setup. Not completely necessary anyway

When I get back to the Uk and my PC on the 28th of this month I think I'll give the new 3.4 Cats a try based on what you have posted, although if I do finally get my home PC hooked up to the net then the first thing I will be doing is verifying that CounterStrike works ok with the 3.4s

More on topic...

I reran VillageMark about 5 times getting the exact same result each time!
Treemark had a tiny difference on 3 runs (the last 2 digits went one/two up and down)...
Same with Final Reality (too bad it didn't save the scores)...
I ran both 3dmarks yesterday and got more or less the same result (but I forgot to write them down ), so I couldn't be bothered to do it again...
(To tell the truth, I was getting quite tired of watching the demos over and over and over again....)

The other benchmarks took too long for me to do any more reruns (I was doing this whilst I was supposed to study), but since the results I got from the reruns was so close, I didn't think it'd do much difference...

The next time ATI is releasing new drivers, I should have fully moved and found all my games, and have the exam behind me so I'll have more time for testing...

[EDIT]
Sorry for not including that in the post original post...[/EDIT]

A few more CAT 3.4 reviews:

http://firingsquad.gamers.com/hardware/catalyst_3.4/default.asp

http://www.technologyvault.co.uk/news/Stories/2003/05/15/10530419721.shtml