View Full Version : 4870x2. First impressions
Pages :
1
2
[
3]
4
5
6
7
8
9
10
Xion X2
08-16-08, 06:40 PM
Please post a few X2 vs CrossfireX benchmarks with high AA settings in different games.
Expect plenty. :)
Only bad thing is I'm still on this 1680x monitor for a few more weeks. But I'll pile on the AA as much as I can and see how much the 2nd card scales.
Picked up Crysis and COD4 today, so I'll have a good library to test:
Crysis
Call of Duty 4
Assassin's Creed
Oblivion
Bioshock
GRID
Stalker
Mass Effect
etc, etc.
slaWter
08-16-08, 07:08 PM
Please make Crysis and Stalker a priority ;)
Please make Crysis and Stalker a priority ;)
i secound that :)
LordJuanlo
08-16-08, 07:49 PM
Just got home now, bedtime, and tomorrow all day out again, damn social life :o
System looks stable after PSU tweak, I think under load +12V was a bit weak. I don't think it's a bad PSU at all, got excellent reviews and it's not cheap, it should be more than enough for this card.
Will try to post more numbers in about 24 hours
Just got home now, bedtime, and tomorrow all day out again, damn social life :o
System looks stable after PSU tweak, I think under load +12V was a bit weak. I don't think it's a bad PSU at all, got excellent reviews and it's not cheap, it should be more than enough for this card.
Will try to post more numbers in about 24 hours
the psu is your best next upgrade
Ambrose
08-16-08, 09:40 PM
Definitely something wrong with Crysis scaling. Let me iterate...
1 GTX280 scores an average of 25-27 at 1680x1050, V.High. The 4870x2 should eat it for breakfast, but it's slower, based on your scores... Also, note that moving from 1680 to 1920 incurs no performance hit. Again, pointing out a problem...
Note that in V.High, AF is essentially disabled even if forced... This is due to POM being enabled when you set the game to the specific settings...
Really?
8800GTX (Stock clocks) - CPU = 2.4 GHz
1680 x 1050 Very High No AA/AF
Trial 1:
Min: 5.76 Avg: 15.35 Max: 18.90
Trial 2:
Min: 5.76 Avg: 16.08 Max: 20.95
Trial 3:
Min: 5.76 Avg: 16.08 Max: 20.95
Trial 4:
Min: 5.76 Avg: 16.10 Max: 20.95
Total: 5.76 15.90 20.44
8800GTX (Stock clocks) - CPU = 3.2 GHz
1680 x 1050 Very High No AA/AF
Trial 1:
Min: 11.30 Avg: 15.45 Max: 19.65
Trial 2:
Min: 11.30 Avg: 15.84 Max: 20.66
Trial 3:
Min: 11.30 Avg: 15.84 Max: 20.66
Trial 4:
Min: 11.30 Avg: 15.85 Max: 20.66
Total: 11.30 15.75 20.41
8800GTX (621 / 1053) - CPU = 3.2 GHz
1680 x 1050 Very High No AA/AF
Trial 1:
Min: 12.05 Avg: 16.65 Max: 20.65
Trial 2:
Min: 12.05 Avg: 17.58 Max: 22.39
Trial 3:
Min: 12.05 Avg: 17.62 Max: 22.39
Total: 12.05 17.30 21.81
Lfctony
08-17-08, 01:26 AM
Ambrose, I don't see the point of your post mate. If SLI is enabled in your tests and you're trying to make a point, take a look at your average framerate... :headexplode: What's your point? I'm sorry, I don't understand what you're showing me mate. :)
Or you're just comparing your 8800 to the 280... Yep, the 280 only has an average or 25-27 at those settings, still unplayable... :( But, you can play on hacked V.High at that same resolution with great performance and amazing visuals. Also, a user just released an great texture pack that incurs no performance hit... :)
He's showing that increasing the CPU clock increases the min framerate, hence the game is CPU limited.
LordJuanlo
08-17-08, 04:33 AM
Here are the specifications of my current PSU, what do you guys think?. I really think this PSU should be more than enough for one 4870x2 (it's SLI certificated (http://www.gigabyte.com.tw/Products/PowerSupply/Products_Spec.aspx?ClassValue=Power+Supply&ProductID=2490&ProductName=ODIN%20GT%20800W)).
http://images.anandtech.com/reviews/psu/2007/gigabyte-odin-gt-800w/02.jpg
And I think here's the problem. The PSU has 4 +12V rails, and 3 of them are mostly idle. But look at 3rd rail, the computer is idle and it has a 10.3A load. When the video card starts working, values are between 16 and 18, while the rest ofthe rails remain at 2-3. I think a simple reorganization of the power cables (PSU is modular) will fix the problem, once I manage to feed the video card from two independent rails
http://img174.imageshack.us/img174/5986/psuwy8.jpg
I don't have much experience with power management, so any advice will be greatly appreciated
Lfctony
08-17-08, 07:04 AM
He's showing that increasing the CPU clock increases the min framerate, hence the game is CPU limited.
The average would have gone up as well, if it was CPU limited... The average is exactly the same here. So the CPU doesn't really help, besides that increase in minimum. If frames were consistently higher from the higher CPU, the average would have increased as well.
Also, V.High Crysis is GPU limited. The GTX280 vs 2 GTX280s scales from 26 to 39FPS... The 4870x2 doesn't scale that well vs one 4870 which again indicates a problem. Xion X2 has stated that the newer drivers improve the scaling a lot. Again, problem and being worked on...
Edit: LordJuanlo, it's quite possible that the given rail is overloaded when you start games. The CPU and the video card are very likely both drawing from that one, since it idles at 10A. Try to split the load, and put the CPU or Video card on V4, which is completely unused from what I see...
He's showing that increasing the CPU clock increases the min framerate, hence the game is CPU limited.
There may be some truth to that. Take a look at a test I ran yesterday with no other changes other than CPU clock speed increase. Look at the framerate increase going from 4.1GHz to 4.2GHz. Tests were performed with a 4870x2.
1680x1050:
8/16/2008 1:03:40 PM - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1680x1050, AA=No AA, 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: 36.70s, Average FPS: 54.49
Min FPS: 38.37 at frame 1964, Max FPS: 79.45 at frame 896
Average Tri/Sec: -39143504, Tri/Frame: -718350
Recorded/Played Tris ratio: -1.28
!TimeDemo Run 1 Finished.
Play Time: 32.24s, Average FPS: 62.04
Min FPS: 38.37 at frame 1964, Max FPS: 84.55 at frame 108
Average Tri/Sec: -44398156, Tri/Frame: -715612
Recorded/Played Tris ratio: -1.28
!TimeDemo Run 2 Finished.
Play Time: 32.15s, Average FPS: 62.21
Min FPS: 38.37 at frame 1964, Max FPS: 85.55 at frame 93
Average Tri/Sec: -44466660, Tri/Frame: -714829
Recorded/Played Tris ratio: -1.28
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
8/16/2008 1:03:40 PM - Vista 64
Run #1- DX9 1680x1050 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 62.125
__________________________________________________ ___________________________
8/16/2008 10:56:21 PM - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1680x1050, AA=No AA, 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: 34.27s, Average FPS: 58.36
Min FPS: 41.55 at frame 1935, Max FPS: 80.28 at frame 877
Average Tri/Sec: -41959600, Tri/Frame: -719034
Recorded/Played Tris ratio: -1.27
!TimeDemo Run 1 Finished.
Play Time: 31.09s, Average FPS: 64.33
Min FPS: 41.55 at frame 1935, Max FPS: 80.28 at frame 877
Average Tri/Sec: -46030716, Tri/Frame: -715524
Recorded/Played Tris ratio: -1.28
!TimeDemo Run 2 Finished.
Play Time: 31.00s, Average FPS: 64.51
Min FPS: 41.55 at frame 1935, Max FPS: 80.97 at frame 1015
Average Tri/Sec: -46183904, Tri/Frame: -715950
Recorded/Played Tris ratio: -1.28
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
8/16/2008 10:56:21 PM - Vista 64
Run #1- DX9 1680x1050 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 64.42
__________________________________________________ ___________________________
1900x1200:
8/16/2008 1:12:17 PM - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, 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: 39.26s, Average FPS: 50.94
Min FPS: 38.90 at frame 1976, Max FPS: 65.92 at frame 879
Average Tri/Sec: -34492576, Tri/Frame: -677174
Recorded/Played Tris ratio: -1.35
!TimeDemo Run 1 Finished.
Play Time: 34.53s, Average FPS: 57.93
Min FPS: 38.12 at frame 1964, Max FPS: 70.52 at frame 887
Average Tri/Sec: -39058836, Tri/Frame: -674265
Recorded/Played Tris ratio: -1.36
!TimeDemo Run 2 Finished.
Play Time: 35.15s, Average FPS: 56.89
Min FPS: 38.12 at frame 1964, Max FPS: 70.52 at frame 887
Average Tri/Sec: -38435800, Tri/Frame: -675603
Recorded/Played Tris ratio: -1.36
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
8/16/2008 1:12:17 PM - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 57.41
__________________________________________________ ___________________________
8/16/2008 10:52:02 PM - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, 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: 37.65s, Average FPS: 53.12
Min FPS: 38.79 at frame 1964, Max FPS: 71.29 at frame 860
Average Tri/Sec: -36022740, Tri/Frame: -678149
Recorded/Played Tris ratio: -1.35
!TimeDemo Run 1 Finished.
Play Time: 33.04s, Average FPS: 60.53
Min FPS: 38.79 at frame 1964, Max FPS: 74.76 at frame 988
Average Tri/Sec: -40794008, Tri/Frame: -673900
Recorded/Played Tris ratio: -1.36
!TimeDemo Run 2 Finished.
Play Time: 33.02s, Average FPS: 60.58
Min FPS: 38.79 at frame 1964, Max FPS: 74.76 at frame 988
Average Tri/Sec: -40848164, Tri/Frame: -674312
Recorded/Played Tris ratio: -1.36
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
8/16/2008 10:52:02 PM - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 60.555
Xion X2
08-17-08, 10:11 AM
And I think here's the problem. The PSU has 4 +12V rails, and 3 of them are mostly idle. But look at 3rd rail, the computer is idle and it has a 10.3A load. When the video card starts working, values are between 16 and 18, while the rest ofthe rails remain at 2-3. I think a simple reorganization of the power cables (PSU is modular) will fix the problem, once I manage to feed the video card from two independent rails
Yep. Use that monitoring program after you've swapped the cables around. Does the manual not tell you which 12v rail each cable pulls from?
Lfctony
08-17-08, 10:16 AM
There's a big difference in your tests though.
1. You're running at High and not V.High. I've already said that V.High is GPU limited, even on GTX280 SLI...
2. You're running a 4870x2 at High. It's very likely that you are CPU limited.
That's a very different scenario...
I've also noticed that the 32bit exe runs the game faster for me in a 64bit environment. Can you test at 32bit? Your frame rates should be a little higher under the same settings...
Xion X2
08-17-08, 10:22 AM
There's a big difference in your tests though.
1. You're running at High and not V.High. I've already said that V.High is GPU limited, even on GTX280 SLI...
2. You're running a 4870x2 at High. It's very likely that you are CPU limited.
Agreed. Turn Crysis up to very high and enable 4-8xAA and see if you're still CPU-limited...
There's a big difference in your tests though.
1. You're running at High and not V.High. I've already said that V.High is GPU limited, even on GTX280 SLI...
2. You're running a 4870x2 at High. It's very likely that you are CPU limited.
That's a very different scenario...
I've also noticed that the 32bit exe runs the game faster for me in a 64bit environment. Can you test at 32bit? Your frame rates should be a little higher under the same settings...
I know that it was run at high settings which is what most people are going to use. Very high is out of reach for everyone, so it's pointless to test.
I tested 32-bit mode a few pages back and noticed a 12FPS drop compared to 64-bit mode.
Here it is again though:
64-bit
8/16/2008 1:12:17 PM - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, 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: 39.26s, Average FPS: 50.94
Min FPS: 38.90 at frame 1976, Max FPS: 65.92 at frame 879
Average Tri/Sec: -34492576, Tri/Frame: -677174
Recorded/Played Tris ratio: -1.35
!TimeDemo Run 1 Finished.
Play Time: 34.53s, Average FPS: 57.93
Min FPS: 38.12 at frame 1964, Max FPS: 70.52 at frame 887
Average Tri/Sec: -39058836, Tri/Frame: -674265
Recorded/Played Tris ratio: -1.36
!TimeDemo Run 2 Finished.
Play Time: 35.15s, Average FPS: 56.89
Min FPS: 38.12 at frame 1964, Max FPS: 70.52 at frame 887
Average Tri/Sec: -38435800, Tri/Frame: -675603
Recorded/Played Tris ratio: -1.36
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
8/16/2008 1:12:17 PM - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 57.41
__________________________________________________ ___________________________
32-bit
8/16/2008 1:20:59 PM - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, Vsync=Disabled, 32 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: 53.62s, Average FPS: 37.30
Min FPS: 23.10 at frame 154, Max FPS: 55.07 at frame 1649
Average Tri/Sec: -25289358, Tri/Frame: -678000
Recorded/Played Tris ratio: -1.35
!TimeDemo Run 1 Finished.
Play Time: 43.82s, Average FPS: 45.64
Min FPS: 23.10 at frame 154, Max FPS: 56.61 at frame 1007
Average Tri/Sec: -30826988, Tri/Frame: -675393
Recorded/Played Tris ratio: -1.36
!TimeDemo Run 2 Finished.
Play Time: 43.43s, Average FPS: 46.05
Min FPS: 23.10 at frame 154, Max FPS: 58.99 at frame 97
Average Tri/Sec: -31019980, Tri/Frame: -673675
Recorded/Played Tris ratio: -1.36
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
8/16/2008 1:20:59 PM - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 45.845
Agreed. Turn Crysis up to very high and enable 4-8xAA and see if you're still CPU-limited...
Well of course not, but again, who cares about very high settings when it's not playable for anyone?
http://www.youtube.com/watch?v=rU5GkRsucQY
anyone seen this
Lfctony
08-17-08, 12:23 PM
Weird, that my situation is reversed (Crysis 32b vs 64b exe). Could it be the ATI card?
Xion X2
08-17-08, 12:45 PM
Weird, that my situation is reversed (Crysis 32b vs 64b exe). Could it be the ATI card?
ATI cards work much better in Vista64. Especially in AA scaling.
Xion X2
08-17-08, 12:47 PM
Well of course not, but again, who cares about very high settings when it's not playable for anyone?
Even on high settings with a decent amount of AA enabled (4-8x) you're not likely to be CPU limited on Crysis. But of course if you don't enable AA then you might.
Never understood why most can't understand this simple concept. The more AF/AA/resolution you increase, the less and less the CPU matters.
And very high is playable for lots of people. The X2 plays it just fine (35+fps) and so do SLI setups with 2** series cards.
I haven't tested it in quad-Crossfire yet (rig should be up and running tonight), but there's a video of some guy w/ a tri-SLI 280 setup on youtube playing it w/ 8xAA on very high DX10, and he looks to be averaging the same or better. Framerate even jumps to 60fps in several places:
http://www.youtube.com/watch?v=8Orsk65ib5c
slaWter
08-17-08, 12:48 PM
ATI cards work much better in Vista64. Especially in AA scaling.
Come on Xion, where are your benchmarks? :p
Xion X2
08-17-08, 01:02 PM
Should be up and running tonight. I take extreme care w/ my builds and have done some additional cosmetic work on this one. I also swapped everything out.. mboard, PSU, CPU, graphics and SSDs.. so it's taking some time. :)
http://www.nvnews.net/vbulletin/showpost.php?p=1747285&postcount=47
slaWter
08-17-08, 01:16 PM
Should be up and running tonight. I take extreme care w/ my builds and have done some additional cosmetic work on this one. I also swapped everything out.. mboard, PSU, CPU, graphics and SSDs.. so it's taking some time. :)
http://www.nvnews.net/vbulletin/showpost.php?p=1747285&postcount=47
No worries, I have time. Cards aren't available here anymore :o
Mhhh that ATI Radeon sticker on the side of the X2 is sexy :D
Lfctony
08-17-08, 01:19 PM
ATI cards work much better in Vista64. Especially in AA scaling.
In my case, I'm only running Vista 64. The 64bit Crysis exe gives me lower results than the 32bit exe... Very strange...
Whoever did that video for ATI is an idiot. Nice video and all, but the 4870x2 is 51% faster than the GTX280 in Lost Planet and not 35%... In Vantage the 4870x2 is 28% faster and not 22%, same with CoD4, 28% and not 22%... They should take second grade math again...
Skinner
08-17-08, 01:30 PM
In my case, I'm only running Vista 64. The 64bit Crysis exe gives me lower results than the 32bit exe... Very strange...
Whoever did that video for ATI is an idiot. Nice video and all, but the 4870x2 is 51% faster than the GTX280 in Lost Planet and not 35%... In Vantage the 4870x2 is 28% faster and not 22%, same with CoD4, 28% and not 22%... They should take second grade math again...
Same here with two HD4870X2's in quad CF:
dx10 64 bits, 1600x1200 x16AF, 2xAA VH, :
!TimeDemo Run 0 Finished.
Play Time: 86.45s, Average FPS: 23.13
Min FPS: 14.85 at frame 1939, Max FPS: 39.31 at frame 998
Average Tri/Sec: -8018893, Tri/Frame: -346633
Recorded/Played Tris ratio: -2.64
!TimeDemo Run 1 Finished.
Play Time: 75.54s, Average FPS: 26.48
Min FPS: 14.77 at frame 1950, Max FPS: 42.45 at frame 63
Average Tri/Sec: -8822207, Tri/Frame: -333206
Recorded/Played Tris ratio: -2.75
!TimeDemo Run 2 Finished.
Play Time: 75.44s, Average FPS: 26.51
Min FPS: 14.60 at frame 1946, Max FPS: 42.68 at frame 991
Average Tri/Sec: -8833640, Tri/Frame: -333204
Recorded/Played Tris ratio: -2.75
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
17-8-2008 17:16:21 - Vista 64
Run #1- DX10 1600x1200 AA=2x, 64 bit test, Quality: VeryHigh ~~ Overall Average FPS: 2649,5
DX10 32 bits average = 30 fps (32 bits results won't be saved in the logfile for some reason)
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.