1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

New to Crossfire, and anything after S939 for that matter, is this "normal"???

Discussion in 'AMD / ATI' started by Formula350, Jun 26, 2010.

  1. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    First things first: System details are accurate in my System Specs to the left
    ◄------------------------------------------------------┘
    Only temporary corrections are CPU @ 3.6GHz (1.275v), HT & NB @ 2GHz, voltages defaulted.

    One of the main things I've noticed is the card's loading. The primary generally is 97-99%, but almost all the time the 'slave' card is 8-12% lower. Is that typical operation, or should I be tweaking something to even the load out more? The only reason I'm concerned is that the primary card gets much hotter, in the order of 20C hotter... However I must confess I have changed the fan out on the slave card, but before I did that it was still 10-12C cooler than the primary. There is also the more pressing concern of the cards only using 60-70% each in DiRT 2. In the loading screen it is actually loading both cards around 97-99%, but in the menu or while racing is when it is so low. I have a feeling that because of it, I should actually be getting higher frame rate than I am (mid 50s to mid 60s, sometimes as low as 30s and as high as 80s), and that sucks since I'm not even running the graphics on high :(

    Second thing is what I replied with in another thread, in regards to the CF bridge... Here's a C&P:
    Know what I just found out earlier... NO bridge is required, at least on my 5770s >_> The computer was in sleep mode and I was just seeing if this fan duct I made would fit between the cards, which meant removing the bridge. Bumped the mouse after I was done and the PC turned on, which I cringed and thought it would crash... It didn't :confused: So I popped open FurMark and it ran fine and utilized both GPUs... Figured "has to be a performance hit though" and ran the benchmark quick. Put it back to sleep and installed the bridge again, re-ran FurMark... same performance :wtf:

    So is perhaps that why the loading is 10% lower on the slave, and perhaps why my performance is lower than what I'd expect? I've not forced HybridCF with ATi Tray Tools either, and this is a fresh install of Windows 7 (x64, but the performance & loading was the same with x86) as well. Catalyst is version 10.6, and I'm running the latest Profiles as well...

    Lastly, and this is not important really, but I'm curious what others have experienced: Using CCC's overclock auto tuning feature, it finds a max core clock of 980-1000MHz on 1.2v (default). Yet that ends up being no where near stable, with driver crashing/system freezing occurring anywhere from a matter of seconds to instantly upon FurMark loading. With gaming it seems like, with minimal testing so far, that 890MHz is the highest stable clock on the core :ohwell: Same story for a lot of you?

    Any other details, just ask and I'll do my best to provide them :)
    Thanks!
     
  2. overclocking101

    overclocking101

    Joined:
    Apr 8, 2009
    Messages:
    3,003 (1.07/day)
    Thanks Received:
    465
    Location:
    vermont
    well upon first though it seems normal to have the slave card use a little less unless you are overclocked to 4ghz or more because of the bandwidth the cpu has to put out during communicating. also CCC auto tune is garbage so thats normal as hell. with the 5770 i have i need like 1.25 or so to hit 1ghz core but some cards clock better than others.
     
  3. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    Well I can deal with the slave card not being at 95+% (it is sometimes), but like with DiRT 2 with both cards not running over 80% while racing and the low FPS, that is down right annoying :\ I mean with 1 card running @ 99% I get mid 40s to low 50 FPS, which is probably on average only 15 FPS less than in Crossfire... So it makes me think there is lots of lost FPS going on due to that :(

    If I could manage 4GHz on this CPU like cdawall managed, I'd be willing to test if it kicks the usage up more, but alas I cannot :ohwell:
     
  4. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    Alright so I don't know exactly which is the reason, but I noticed that if I run FurMark in full screen that the FPS are about 30-50fps higher, and also the load % per card is about 95-96%... So I'm not sure if it was the equaled load or something to do with full screen for the FPS boost, as the resolution is the same (1680x1050, native res for my LCD) :ohwell:

    I'm also noticing that since I've gone to x64 that a lot of the MVPU tweaks that ATi Tray Tools has don't do anything anymore. If none of them worked I'd chalk it up as some bug with my system, but I know at least the "Disable Dongless MVPU" option works since I tried it with the bridge removed and you can see the AFR trying to work as you get these pink partial frames (bars due to the high FPS). It's activation doesn't seem to actually help performance/load balance when the bridge is present though :( But the other options like the type of rendering to perform (AFR, split screen[scissor], chessbox or SFR) aren't doing anything when I change it, which it used to in x86. At least it doesn't seem to do anything now, as split screen and chessbox both resulted in the graphics going wonky, with SFR dropping performance in the game I was testing by about 30FPS. Now there is literally no change in frame rate.
     
  5. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    45,462 (10.23/day)
    Thanks Received:
    12,718
    Location:
    Australalalalalaia.
    crossfire doesnt work in windowed mode.

    Ati tray tools doesnt really work in x64


    going by my experience with 2x 4870's (very similar performance), you're CPU limited. Not by a large amount, but limited nonetheless.

    No a tri/quad/hexa core wont help you, you dont need MORE cores - you need FASTER cores. OC more, get a better OCing chip sometime in the future, etc.
     
    Formula350 says thanks.
    10 Year Member at TPU
  6. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    Didn't know that it didn't work in windowed hah I mean, I wouldn't have even thought about it since in FurMark (MSI Kombuster) windowed mode, the cards were both getting loaded. And if I were to not use the Multi-GPU command line switch the frame rate dropped significantly, with the second GPU also not loading.

    I do recall that issue with ATT now that you mention it, but if ATT is making the registry keys, why aren't the settings taking place? :(


    I'll load up my other OC profile to bring it back up to 3.7GHz and 2.4GHz HT & NB. I'll also see if I can squeeze out 3.8GHz, but since cdawall didn't remember the voltages for everything and I'm not too sure what all the BIOS voltage settings should be (I upgraded from S939 so a lot of these are new to me) to aid in overclocks, I've not been able to stably get higher than 3.7GHz :( Which also is strange given this Phenom II is not letting my motherboard clock the same as with the Athlon II X2 250u I had for a short bit before putting it in my neighbor's PC. The HT & NB I could get to 2.9GHz with the HTT I was able to get to 363MHz, now I can't get past 2.4GHz and 240MHz, respectfully. :confused: I'll give the CPU a go though.

    Thanks :)
     
  7. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    45,462 (10.23/day)
    Thanks Received:
    12,718
    Location:
    Australalalalalaia.
    furmark may be a special case, but i know crossfire doesnt work in windowed mode usually.

    ATT can make the reg keys, but they arent in the same place in a 64 bit OS...
     
    10 Year Member at TPU
  8. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    That is what I was sort of thinking. Seeing as XP -> Vista is different as well. But then that makes me wonder, if it's not in the right place then where IS the right place in x64? The place I'm looking is the same spot where the settings are kept in x86 (HKey Local Machine>CurrentControlSet>Control>Video>[Gibberish Key]>0000>UMD), but if that isn't it I can't figure out where it would be kept :ohwell:
     
  9. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    45,462 (10.23/day)
    Thanks Received:
    12,718
    Location:
    Australalalalalaia.
    its entirely possible to update it and make it work in x64, the guy in charge of ATT simply hasnt bothered.

    also, some of those reg keys may be legacy items leftover in the 32 bit drivers, that are simply not hooked up to anything in the 64 bit drivers.

    that said, i prefer teh lazy programmers theory.
     
    10 Year Member at TPU
  10. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    I hope it's not the lazy programmer theory haha While the Split Screen and Chessbox didn't work (in x86), SFR did and I'd at least like to have that option! :\

    I have suspected that a lot of those don't work anymore though. Trueform actually has be curious, even if all the keys to turn it on aren't in the registry, I am 90% sure I know what it is... I just don't currently have a game that has it on any of the HDDs I have plugged in currently :S Wolfenstein is the only one that comes to mind though anyways.

    So if that is the case, how does one enable Single Frame Render now, or even before for that matter?
     
  11. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    Well it seems that using the Afterburner OSD is sapping a fair amount of frames :S In FurMark there is a difference of 70FPS in windowed (208 on and 278 off) and 200FPS full screen (390s on and 605 off)! So I'll have to see how that affects gaming performance with it off. Since I have a second monitor setup now, I can just run the Afterburner readings window on it... Now if I can only get a modded BIOS burned to the cards, I can take care of my fan problem! Thing is just set far too low for this shotty heatsink setup ASUS went with and Afterburner's custom fan setup causes huge FPS drops when it is changing the fan speed :ohwell: Thankfully I wear headphones atm and I can't hear it :p
     
    Last edited: Jun 28, 2010
  12. Formula350

    Formula350 New Member

    Joined:
    Mar 13, 2007
    Messages:
    864 (0.24/day)
    Thanks Received:
    49
    Location:
    Eastern Tennessee
    UGH >_< Now I don't know what the hell is up... I had one of the 5770s (the slave) fan and shroud replaced with a 90mm case fan, and it ran quite a bit cooler. So I swapped it in place of the primary to see if even in the tight conditions, it would be able to move enough airflow in order to cool it down, turns out it couldn't. SO I finished my stupid cardboard fan duct for the original primary and put it in, seemed to make it worse, so I pulled it out! Ran up the cards to temp and now BOTH cards get way hotter than they used to!? Card 1 would get to about 83C at the most, now it gets to 101C before I shut down FurMark O_O Card 2 would only get to like 66C and now it was well into the 70s o_0 This is why I never buy ASUS :(
     
  13. Mussels

    Mussels Moderprator Staff Member

    Joined:
    Oct 6, 2004
    Messages:
    45,462 (10.23/day)
    Thanks Received:
    12,718
    Location:
    Australalalalalaia.
    afterburner/rivatuners OSD DOES definitely sap performance. theres a beta on the afterburner forums i believe, where that gets solved. (but its still buggy and crap, last i checked)
     
    10 Year Member at TPU

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)