New Catalyst 9.12 - Performance boosts on 58XX Series

HalfCrazy

Senior member
Oct 3, 2001
853
0
0
Found a fixed link for the XP drivers. But the website is getting hammered big time right now. Only getting 5KB/sec download speed.
 

jvroig

Platinum Member
Nov 4, 2009
2,394
1
81
How about a link that shows the claimed 15-20% improvement?

And why only the 58xx series? Catalyst 9.12 offers no improvement for 57xx?
 

Qbah

Diamond Member
Oct 18, 2005
3,754
10
81
How about a link that shows the claimed 15-20% improvement?

And why only the 58xx series? Catalyst 9.12 offers no improvement for 57xx?

Well, the HD5770 has the same 800 shaders as a HD4870 and quite a bit less memory bandwidth. Higher clocks though on the core... And it's just a tad slower. What more do you want to squeeze out of those cards? They have a bit worse performance parameters than a HD4870 and perform a bit worse - looks fine imo.

Now the HD58xx... They doubled up on pretty much everything, yet the performance isn't there. The card has better parameters than a HD4870x2 but it's usually slower - looks like they found out the reason and tapped that hidden power - provided of course those claims are true. Though a link to some numbers would be fine :)
 

Daedalus685

Golden Member
Nov 12, 2009
1,386
1
0
There are no formal reviews on performance yet, but guru3d has a long forum post on it. The link is in the other 9.12 thread here.

Most seem to be talking about 10% increase at least, 5770 included. That being said, I'll wait for a review site to bench them.. I'll try them myself when I get home on my 4890s.
 

jvroig

Platinum Member
Nov 4, 2009
2,394
1
81
@Qbah: It has been noted by some people like BFG10K who do their fair shaire of investigation and benchmarking that the 5770 does perform erratically (lower than expected in some, then beats a much more expensive GTX card in fewer cases). I would refer you to the threads about it, but I'm sure you already have read it, and I don't remember the particular threads, and they are all too long because argumentative posts about whether or not it is driver-related or bandwidth-related abound. It's no fun trying to re-read them.

@Daedalus685: Agreed, I hope a review site does bench them when the new drivers are officially released (I am assuming the linked 9.12 drivers are leaked and not yet official? Otherwise it would have come from the AMD/ATI site and not some other site). BFG10K seemed to be interested in doing so as well, so I am sure we will also hear from him once the drivers are out.
 

Number1

Diamond Member
Feb 24, 2006
7,881
549
126
HD5770 @ 1445/960

I just removed 9.11 and installed 9.12 and ran 3DMark06 before and after.
The card was overclocked to the max setting in Catalyst.

There was no difference in performance, slightly slower if anything.
Before (9.11): 16846
After (9.12): 16735
 

Meaker10

Senior member
Apr 2, 2002
370
0
0
HD5770 @ 1445/960

I just removed 9.11 and installed 9.12 and ran 3DMark06 before and after.
The card was overclocked to the max setting in Catalyst.

There was no difference in performance, slightly slower if anything.
Before (9.11): 16846
After (9.12): 16735

Its more for 5770 crossfire or single DX10/11 AFAIK
 

yh125d

Diamond Member
Dec 23, 2006
6,886
0
76
Eagerly awaiting reviews. I agree with BFG10K, that the 58/57 cards could gain a noticable about from future drivers


@ Qbah, they still could eke out some more performance even with roughly the same hardware, as they get better and better at writing drivers for the microarch
 

lavaheadache

Diamond Member
Jan 28, 2005
6,893
14
81
I would really like to see Crossfire evolution. I hate microstutter! Even though i can mostly emliminate it with vsync, triple buffering and smooth mouse, it still needs to be tweaked. I'm sure these guys can fix this stuff they just have different priorities.
 

luv2increase

Member
Nov 20, 2009
130
0
0
www.youtube.com
These are the best drivers ATI has ever put out, period. Here are some stock runs @ 850/1200.

3dMarkVantage (Tri-Fire 5870) with 9.11 drivers = 28311 GPU Score
http://service.futuremark.com/compare?3dmv=1510935

3dMarkVantage (Tri-Fire 5870) with 9.12 drivers = 34426 GPU Score
http://service.futuremark.com/compare?3dmv=1510935


They finally fixed the scaling issue!!! I get 165FPS average on Dirt 2 in DX11 @ 1920x1080 w/Ultra Preset and 8xMSAA!!! With the 9.11 dirt2 hotfix, I would get 80FPS averaged in Tri-Fire and 97FPS in CF.


Here is what I get with these drivers and a 1000/1295 OC on all cards using 1.25v with MSI Afterburner 1.41...

3dMarkVantge (Tri-Fire 5870) @ 1000/1295 ---> 36922 GPU Score!!!!!!!!!!
http://service.futuremark.com/compare?3dmv=1693329


I've also gotten an average of 20fps increases in both Crysis and Crysis Warhead using these new 9.12 (8.681) drivers @ 1920x1080 resolution!!!

ATI PULLED THROUGH, FINALLY! I now got what I paid for LoL
 

Qbah

Diamond Member
Oct 18, 2005
3,754
10
81
(...)
I've also gotten an average of 20fps increases in both Crysis and Crysis Warhead using these new 9.12 (8.681) drivers @ 1920x1080 resolution!!!

ATI PULLED THROUGH, FINALLY! I now got what I paid for LoL

What are your scores in Crysis Very High / Warhead Enthusiast 1920x1080 AAx4 then? Please share :)
 

BFG10K

Lifer
Aug 14, 2000
22,709
2,997
126
These drivers just might interest me enough to drop my 5770 back into my system and retest it.
 

Xpoc

Member
Feb 17, 2008
114
0
0
I have gained 700 points on previous high score in Vantage. The points all came from the GPU score which gained a whopping 1000 points!!

Nice jump in performance.
 

Zargon

Lifer
Nov 3, 2009
12,218
2
76
I have gained 700 points on previous high score in Vantage. The points all came from the GPU score which gained a whopping 1000 points!!

Nice jump in performance.


so what was the old score and whats the new one??
 
Last edited:

Qbah

Diamond Member
Oct 18, 2005
3,754
10
81
I'm seeing 47fps avg 2560x1600 4xx all enthusiast in Warhead, frost level.

That is crazy ! O_O

And where are the emotes ?

EDIT: Check this SH1T OUT :D Taken from the Guru3D thread - the dude is running 4x HD5770 and a PhII - INSANE :D AAx16 Link to his post. I put AVG in bold too. MIN almost 40FPS.......

16-12-2009 16:37:01 - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX10 1920x1080, AA=16x, Vsync=Disabled, 64 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
================================================== ============
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 37.69s, Average FPS: 53.07
Min FPS: 38.72 at frame 1950, Max FPS: 79.92 at frame 1586
Average Tri/Sec: -33658988, Tri/Frame: -634282
Recorded/Played Tris ratio: -1.45
!TimeDemo Run 1 Finished.
Play Time: 32.53s, Average FPS: 61.48
Min FPS: 38.72 at frame 1950, Max FPS: 82.71 at frame 1613
Average Tri/Sec: -38734532, Tri/Frame: -629999
Recorded/Played Tris ratio: -1.46
!TimeDemo Run 2 Finished.
Play Time: 32.61s, Average FPS: 61.33
Min FPS: 38.72 at frame 1950, Max FPS: 84.05 at frame 1654
Average Tri/Sec: -38615616, Tri/Frame: -629600
Recorded/Played Tris ratio: -1.46
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============

Completed All Tests

<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>

16-12-2009 16:37:01 - Vista 64

Run #1- DX10 1920x1080 AA=16x, 64 bit test, Quality: High ~~ Overall Average FPS: 6140,5
 
Last edited:

luv2increase

Member
Nov 20, 2009
130
0
0
www.youtube.com
What are your scores in Crysis Very High / Warhead Enthusiast 1920x1080 AAx4 then? Please share :)



Crysis with the Real Lifesys Mod using TOD2 with settings of "very high" in the benchmark tool (although I do get the warning saying I have a custom autoexec.cfg) @ 1920x1080 with 8xAA (in-benchmark) and 8xAF (forced via CCC), I am getting 66FPS average in DX10...

One thing I have noticed with Crysis/Crysis Warhead is that when you force AA via the CCC, you will have hardly any hit to your FPS, but AA is still indeed working. When you set AA in-game or in this case in the benchmark, it will have a bigger hit.

Literally, when AA is set in CCC and AA is set to 0xAA in-game with Crysis/CW, there is "hardly" and drop in FPS.

I play both games with 24xAA (edge-detect), 16xAF, Highest Mipmap detail and highest settings and with mods.

Real Lifesis does put a smack on your FPS. I would say I've gained a good 15-20FPS in Crysis/CW. ():)

edit: I use vsinc too and set my 46" 1080P Samsung HDTV to 120hz automotion. Nothing beats it :)
 

Qbah

Diamond Member
Oct 18, 2005
3,754
10
81
Crysis with the Real Lifesys Mod using TOD2 with settings of "very high" in the benchmark tool (although I do get the warning saying I have a custom autoexec.cfg) @ 1920x1080 with 8xAA (in-benchmark) and 8xAF (forced via CCC), I am getting 66FPS average in DX10...

One thing I have noticed with Crysis/Crysis Warhead is that when you force AA via the CCC, you will have hardly any hit to your FPS, but AA is still indeed working. When you set AA in-game or in this case in the benchmark, it will have a bigger hit.

Literally, when AA is set in CCC and AA is set to 0xAA in-game with Crysis/CW, there is "hardly" and drop in FPS.

I play both games with 24xAA (edge-detect), 16xAF, Highest Mipmap detail and highest settings and with mods.

Real Lifesis does put a smack on your FPS. I would say I've gained a good 15-20FPS in Crysis/CW. ():)

edit: I use vsinc too and set my 46" 1080P Samsung HDTV to 120hz automotion. Nothing beats it :)

That is CRAZY :D Check the result I posted from the Guru3D thread. 4 HD5770s scale incredibly now! Looks like ATi found a way to have CrossFire scale amazingly in Crysis ! What a Christmas gift to all HD5xxx owners, especially the ones with a CrossFire setup... :)

YES, IT DOES PLAY CRYSIS MAXED - always wanted to say that :D
 

OVerLoRDI

Diamond Member
Jan 22, 2006
5,490
4
81
Are there any other hosts for this file or perhaps a torrent? getting ~10kb/s on those links.