Originally Posted by Shintai
EDIT: LOL ty v_rr
what i think is grid is a strange game for a test
look what anandtech found.
4870x2 84
4870 43
4870cf 30
9800gx2 17
http://www.anandtech.com/video/showdoc.aspx?i=3354&p=7
When i'm being paid i always do my job through.
I would like to further add that trying to find out what FPS are really doesn't matter/help. The discussion is whether or not the X2, CF or a single card have any difference in AFR stuttering.
These illustrations show this.
I am sure Sampsa will be releasing more information about resolution, FPS, etc. But the most concerning of our inquiries has been answered.
.
Hi,
Ok here is a short explanation:
First of all, I'm using Race Driver: GRID demo and 4870 CrossFire and 9800 GX2 seem to work just FINE. With 1920x1200 resolution, 4xAA and 16xAF I get 66,9 FPS with single 4870, 73,4 FPS with 4870 CF and 75,6 FPS with 4870 X2:
(Red bar keskimääräinen is average FPS)
Second, the graphs I've showed here related to microstutterin ARE NOT intended to display performance but difference between frame rendering times. I have only recorded 2 seconds of gameplay and used first 30 frames from data.
If performance is 60 FPS.. that would mean graphics card renders 120 frames in 2 seconds. Since I have only used 30 frames to analyze the results and draw the graphs, it means gameplay of half a second (0,5s, 500 ms)!!! That's not enough to draw conclusion about perfromance...
... it might not be enough to draw conclusion about microstuttering either but in my own tests 30 frames already show clear pattern with ATI Radeon HD 3870 X2 and rendering problem where every other frame is being rendered after different time than every other.
Last edited by Sampsa; 07-15-2008 at 03:51 AM.
Favourite game: 3DMark
Work: Muropaketti.com - Finnish hardware site
Views and opinions about IT industry: Twitter: sampsa_kurri
but if dual gpu is not working can we have a clear information about microstuttering.
When i'm being paid i always do my job through.
SLI is total garbage for fast paced fps games. You suffer from massive lag using the AFR rendering mode. Half of the 6 pre-rendered frames are out of date by the time they get displayed.
Indeed an important thing to note. We need a game where we see 80%+ scaling to have a educated idea ( or even high levels of IQ on a 30" LCD should give a nice indication ). I didn't realize scaling in Grid was nearly non existent. Perhaps higher levels of anti aliasing may create a spread between the single 4870 and CF?
Feedanator 7.0
CASE:R5|PSU:850G2|CPU:i7 6850K|MB:x99 Ultra|RAM:8x4 2666|GPU:980TI|SSD:BPX256/Evo500|SOUND:2i4/HS8
LCD:XB271HU|OS:Win10|INPUT:G900/K70 |HS/F:H115i
ty for fast answers.Hi,
Ok here is a short explanation:
First of all, I'm using Race Driver: GRID demo and CrossFire and 9800 FX2 seem to work just FINE. With 1920x1200 resolution, 4xAA and 16xAF I get 66,9 FPS with single 4870, 73,4 FPS with 4870 CF and 75,6 FPS with 4870 X2:
Second, the graphs I've showed here related to microstutterin ARE NOT intended to display performance but difference between frame rendering times. I have only recorded 2 seconds of gameplay and used first 30 frames from data.
If performance is 60 FPS.. that would mean graphics card renders 120 frames in 2 seconds. Since I have only used 30 frames to analyze the results and draw the graphs, it means gameplay of half a second (0,5s, 500 ms)!!! That's not enough to draw conclusion about perfromance...
... it might not be enough to draw conclusion about microstuttering either but in my own tests 30 frames already show clear pattern with ATI Radeon HD 3870 X2 and rendering problem where every other frame is being rendered after different time than every other.
now i want to ask something else. now we are sure that dual gpu is working and as the test results shows we see that single card has more stutter then cf or x2 versions. i think this might be because of single 4870 is at its limits so when rendering it stutters more because some frames are more harder to render. so can it be more healthy to make test at lower resolutions or detail settings for not to push single 4870 to its limits to see what is really going on stuttering side.
When i'm being paid i always do my job through.
That has been answered in many threads already pertaining to GRID. (it's understood)
And now it has been answered yet again... thanks Sampsa!
.
We also need to be very careful when we talk about microstuttering,
If you play game at avg 15 FPS and it seems laggy, that IS NOT microstuttering. Your quality settings are just too high for your system.
If you play game at avg 30 FPS and minimum FPS drops in some situtations to below 20 FPS and game feels laggy, that IS NOT microstuttering. Your quality settings are just too high for your system.
If you measure the time between frames are being rendered and draw a graph of difference of times between rendering the frames and see WOBBLING line (like yellow line in above graph), that IS microstuttering.
Correct me if I'm wrong
PS. Here is a good FAQ about microstuttering: http://www.hardforum.com/showthread.php?t=1317582
Favourite game: 3DMark
Work: Muropaketti.com - Finnish hardware site
Views and opinions about IT industry: Twitter: sampsa_kurri
Yeah but if it's not too much to ask, could you do a short test with Unreal Tournament 3 @ 1680×1050? A ton of games use that engine and most people play at that resolution so the outcome would be much more universal...We also need to be very careful when we talk about microstuttering,
If you play game at avg 15 FPS and it seems laggy, that IS NOT microstuttering. Your quality settings are just too high for your system.
If you play game at avg 30 FPS and minimum FPS drops in some situtations to below 20 FPS and game feels laggy, that IS NOT microstuttering. Your quality settings are just too high for your system.
If you measure the time between frames are being rendered and draw a graph of difference of times between rendering the frames and see WOBBLING line (like yellow line in above graph), that IS microstuttering.
Correct me if I'm wrong
PS. Here is a good FAQ about microstuttering: http://www.hardforum.com/showthread.php?t=1317582
Yes you a right. What i want is to be sure what is going on here. Now your tests show that single gpu stutters more then a dual one. So as the stutter we are looking for shouldnt stutter more in single gpu then dual gpu this situation needs an answer. As i said below maybe single card is stuttering more because it is at its limits.but As you said this might not be shown on the graph like this so this theory might be wrong. Then i am thinking of something else. Maybe ATI did somethink about microstuttering in drivers or etc so by this adding cf maybe stuttering less then a single gpu. Also single 4870 stutters like a 9800gx2 as shown in the graph. If 9800gx2 has stuttering problem then we should say single 4870 has stuttering problem too.
I am just seeking answers for this situation.
When i'm being paid i always do my job through.
I agree. The HD4800 series seems to handle GRID very well. A game like Oblivion or Unreal Tourney would give some parity to this discussion.
.
I have already tested UT3 @ 1920x1200 no AA and 16xAF:
http://www.xtremesystems.org/forums/...&postcount=529
No issues with any of the cards I tried.
Favourite game: 3DMark
Work: Muropaketti.com - Finnish hardware site
Views and opinions about IT industry: Twitter: sampsa_kurri
Is it really that HARD to read the graph
As you can see 4870X2 and 4870 CF show almost equal variation between times frames are being rendered.TBH I'm a bit confused about the graphs as well.
Sampsa if it's possible can you add some comments on the methodology and what we are looking for in the graphs and provide the theoretical ideal graph as reference.
Would be really usefull
This could be helpful?
Only thing which is important is how big is variety in that yellow horizontal line (how much zig-zag it goes up and down). It's 5 milliseconds (it differs between 15 and 20 milliseconds) for 4870, which is clearly unnoticeable. Intrestingly both 4870x2 and 4870CF plays more smoothly than single 4870, but as I said thats unnoticeable.
Last edited by L7R; 07-15-2008 at 05:41 AM. Reason: typos
You won't notice microstutter in UT3. It's there, but the framerate is too high for it to be perceptible, as well as UT3 engine already optimized for multi-gpu and phys-X. Because the engine will delay to wait for Phys-X data, stutter is almost impossible to notice.
But try out bioshock, or Gears.
I think sometimes stutter is from onboard composting engine. It takes some time for final rendered frame to go from second/third/fourth card to main card for output. The previously used TMDS chips showed this very quickly when pushed...it is like alot of time the second frame just never gets there in time...
So, what happens, is in generations previous to RV770, is that a memory control ringstop is used for crossfire communication. Should that ringstop be busy when the data comes in, ther is a delay to free it up, and we see "stutter".
Crysis makes this more pronounced because of it's already high memory usage.
So I'd agree with you totally, Sampsa, as to what is going on. And again I want to thank you for taking the time to show that it IS an issue with soem cards, and that those of us that we complaining of such weren't off our rockers. It's highly appreciated.
I will say though, i tihnk your testing methods are a bit flawed. you need to let a game load up, and then let the game idle for a few seconds before taking data. You notice "stutter" in single cards due to fraps, so giving the system some time to settle into it's task will highlight this problem a bit better, as ANY app is going to be inconsistent during the first couple of seconds.
Last edited by cadaveca; 07-15-2008 at 05:35 AM.
Kinda off topic, but any explanation on why your CF / X2 scaling in GRID is so much 'different' than what AnandTech preview shows?Hi,
Ok here is a short explanation:
First of all, I'm using Race Driver: GRID demo and 4870 CrossFire and 9800 GX2 seem to work just FINE. With 1920x1200 resolution, 4xAA and 16xAF I get 66,9 FPS with single 4870, 73,4 FPS with 4870 CF and 75,6 FPS with 4870 X2:
(Red bar keskimääräinen is average FPS)
Second, the graphs I've showed here related to microstutterin ARE NOT intended to display performance but difference between frame rendering times. I have only recorded 2 seconds of gameplay and used first 30 frames from data.
If performance is 60 FPS.. that would mean graphics card renders 120 frames in 2 seconds. Since I have only used 30 frames to analyze the results and draw the graphs, it means gameplay of half a second (0,5s, 500 ms)!!! That's not enough to draw conclusion about perfromance...
... it might not be enough to draw conclusion about microstuttering either but in my own tests 30 frames already show clear pattern with ATI Radeon HD 3870 X2 and rendering problem where every other frame is being rendered after different time than every other.
It seems they got close to 100% scaling with X2 compared to single. True the res they used was higher but I don't think this is just a matter of CPU bottlenecking. Your test shows only what, 10% scaling?
it's first 30 or so frames, which CANNOT indicate performance accurately.
To understand that one needs to compare system settings, driver settings, OS versions/updates installed, game version, ingame settings, testing methodology, testing period, testing scene. guess I cover most of the things
I bet thats because using different motherboards. I have had many differences in crossfire scaling between different boards. I have seen Gigabyte boards gives most coherent results between different games. MSI and asus may be great on some game but not so consistent between different games.
One intresting thing what you find in some gigabyte mobos is extra molex socket in pcb just for crossfire use! I was little surprised and wondering what the heck is that for.
Last edited by L7R; 07-15-2008 at 05:52 AM.
GRID only plays nice with Crossfire/SLI after patch 1.1. Sampsa?
What you say makes little sense to me, as most that changes is some internal chipset latency, which will never have more than a FPS or two impact in games. Multi-gpu this may havea slightly larger impact..maybe double..but not more than that.I bet thats because using different motherboards. I have had many differences in crossfire scaling between different boards. I have seen Gigabyte boards gives most coherent results between different games. MSI and asus may be great on some game but not so consistent between different games.
One intresting thing what you find in some gigabyte mobos is extra molex socket in pcb just for crossfire use! I was little surprised and wondering what the heck is that for.
The extra socket is to provide power directly to pci-e slots, eliminating extra electrical noise in the pcb. But now that cards are getting higher-draw, this socket is disappearing, as the 24-pin mobo connector went up from 20 to provide the extra power anyway.
Ok, nm...it was explained already. :p
Thanks for clearing this. But that doesn't really explain the difference between motherboards even from same manufacturer.
Last edited by L7R; 07-15-2008 at 06:08 AM.
lol bad results. how does 1/2 less pci-e bandwidth = double performance?(x38 vs p45, nevermind p45 TRD issues)
Why is it my own testing, i see the opposite....oh, right, my testing doesn't put money in my pocket!
Anyway different chipset is not good compare, eh? Maybe try same chipset to see difference.
Posting Permissions
Bookmarks