PDA

View Full Version : RealBench and BF3



xmanrigger
08-10-2013, 09:54 AM
Can pass numerous runs of RB, but cant play BF3 without crash at same speeds, 4400/2133. ???



Originally Posted by Arne Saknussemm View Post
Probably SLI related that xman....

Not sure if SLI related. Hit F5 in BIOS and run default settings, all is well. Set mild OC, pass RB all day long, crash in BF2. 4400@1.35v and 2133/9-11-9-27-1@1.65v


Originally Posted by Zka17 View Post
That could be because of the graphics... the component(s) not stressed in this version of RB... - what is the exact nature of those crashes? Any BSOD code, Event viewer explanation?


Blue scrns to game freezing but able to alt-ctrl-del and reboot. 3800/1333 no crash.

Arne Saknussemm
08-10-2013, 11:36 AM
Is that straight multi OCing Xman...no bclk OCing (because as yopu know touch BCLK and you screw with RAM and PCIe frequency)?

Have you run SLI without a bridge...to eliminate bad bridge......tried a 314.xx series driver instead of the 320.xx ones which seem to give a lot of people headaches (though I must say I have not had a problem on any of them) there is a new beta out 326.41.

And finally we come to your CPU ...the IMC and PCIe handling are on the same bit of the CPU if I remember right....maybe, given your RAM OC experience and now PCIe not being able to take the strain....you should be first in line for IB-E ;)

OC your bike for a month mate and then come back with a CPU instead of a lemon in the socket :D

xmanrigger
08-10-2013, 12:36 PM
Is that straight multi OCing Xman...no bclk OCing (because as yopu know touch BCLK and you screw with RAM and PCIe frequency)?

Have you run SLI without a bridge...to eliminate bad bridge......tried a 314.xx series driver instead of the 320.xx ones which seem to give a lot of people headaches (though I must say I have not had a problem on any of them) there is a new beta out 326.41.

And finally we come to your CPU ...the IMC and PCIe handling are on the same bit of the CPU if I remember right....maybe, given your RAM OC experience and now PCIe not being able to take the strain....you should be first in line for IB-E ;)

OC your bike for a month mate and then come back with a CPU instead of a lemon in the socket :D


Yes Arne, bclk at 100, untouched, and multi at 44. How can I run SLI without bridge? My memory is at a very mild speed, half-assed loose timings, and sufficient voltage. I had the memory at much higher speeds with same or less voltage running and passing RB. No OC on video cards whatsoever.

Next round of BF3, I will try 125 strap at 2133. Never really could get my mem anywhere at 100, 2240mhz tops. 2395mhz and passing RB using 125 strap. All my good RB scores were at 125 strap and then up.

HiVizMan
08-10-2013, 01:19 PM
Actually you can run your SLI without a bridge, the performance will not be as good but it will be in SLi.

The fact that RB is passed but a graphic intensive game is not certainly points to the issue being a video one, however memory (system that is) is key to how your VGA performs with your system.

xmanrigger
08-10-2013, 09:32 PM
Actually you can run your SLI without a bridge, the performance will not be as good but it will be in SLi.

The fact that RB is passed but a graphic intensive game is not certainly points to the issue being a video one, however memory (system that is) is key to how your VGA performs with your system.

I think you are correct with the sys memory theory. I am 99.9% sure my GTX680s are solid. Could it be not enough voltage to dramm for 2133/9-11-9-27-1 at 1.65v? I would think if 2400mhz is spec'd to run at 1.65v, 2133mhz with half-assed timings should be a breeze.

And also as I stated in 1st post, 680s are at stock speed. No overclock on mem or GPU. I was under the assumption that gaming benefits more so with higher sys mem speed than tighter latencies.

BF3 runs butter smooth with everything at stock and graphics at Ultra on this system. I was trying to see if it would run with RB stable overclocks. Strange. From what I seen messing with RB to the extent I have, any clocks that pass multiple runs of RB should shred any game. Apparently not.

Regardless, I still hold RB as one of my tools to establish system stability.

As far as posting a higher score than the 842, I hoping to post an 850 soon. That I believe, will be the limit,n my 3930K's IMC.

Zka17
08-11-2013, 01:04 AM
As HiVizMan said, the the graphics setting rely a lot on the system memory settings when it's about the overall system stability...

How about trying XMP settings from BIOS, leave the CPU at stocks for now, then test how your graphics is behaving... if it's fine, then start pushing the CPU multi - but still let your voltages on auto...

Arne Saknussemm
08-12-2013, 08:02 AM
Plus the number of SLI threads with BF3 crashing is legion and may well be driver related....does this happen with any other games stuff like Metro for example which is a stern test.....benchmarks like Unigine Valley...etc.

Nodens
08-12-2013, 07:27 PM
And also as I stated in 1st post, 680s are at stock speed. No overclock on mem or GPU. I was under the assumption that gaming benefits more so with higher sys mem speed than tighter latencies.


This is not true. Both matter and depending on the game engine, its settings and the amount of VRAM available on the video card you may see situations that benefit the latencies more. But in general there is a balance. High latencies can cause frame stuttering or driver timeouts.

I'll also second Arne's suggestion and test with Metro which is one very heavy beast as the issue may be game (or game+driver) specific that manifests easier on certain settings.
On a last sidenote you may also have residual effects after messing with BCLK, specially if you were playing with skew settings in order to stabilize an edgy BCLK setting (debug code 01 even after reverting to stock BCLK is such an example). In that case you should drain the board from all power.

EDIT: Also if a minidump gets created could you post it here so I can take a look in the debugger?

chrisnyc75
08-13-2013, 05:02 PM
Similar problem here. I can pass RealBench, Cinebench, Heaven, & Valley all day long, but this weekend (when I *finally* had time to sit down and do some actual gaming on my new system) I experienced 2 crashes in the midst of gaming (my gpu's are at stock speed, and my RAM is solid as a rock 8+ hours of memtest86+). Turns out my oc wasn't quite as stable as I thought it was, back to the drawing board.

Anyway, while I do think RealBench is a valuable tool for verifying initial stability, and is certainly a lot of fun to play with, I'm not so sure I think it holds up against tried-and-true stability benchmarks like IBT, Prime, LINX, etc. It has its place, but I don't think it's the all-in-one stability testing tool some have made it out to be....yet.

Zka17
08-13-2013, 05:19 PM
Turns out my oc wasn't quite as stable as I thought it was

OK, so what did you have to do make it stable for those "true stability benchmarks like IBT, Prime, LINX, etc." ?

Are you aware that those are not stressing the graphics together with the CPU and RAM too?

chrisnyc75
08-13-2013, 05:35 PM
OK, so what did you have to do make it stable for those "true stability benchmarks like IBT, Prime, LINX, etc." ?

Still working on that. My time is limited to a few hours in the evening after I get home from work, so it's a slow tedious "to be continued" process. Last night I tweaked down my dram, vtt, & vcssa voltage to make sure I'm not producing unnecessary heat. Tonight I'm going to work with vcore and see where that gets me.


Are you aware that those are not stressing the graphics together with the CPU and RAM too?

I am aware of that, but if I'm passing RB, CB, Heaven, Valley, & Firestrike (all of which should put stress on both gpu and cpu, right?) but still failing IBT, Prime, & LINX, and my real-world symptom is that I'm crashing while gaming, the logical conclusion is that it's probably a weakness in my cpu oc that is causing the crash -- perhaps just a statistical thing (gaming for 3+ hours is a lot more calculations than 3 passes in RB). And again, my gpu is not overclocked at all, so unless one of my brand new 770's is a lemon (I suppose this is possible, but I haven't tested for that yet....maybe I'll do that tonight) all signs point to the cpu oc.

I'm not trying to knock RealBench - I am a fan. I just wanted to mention my real world experience with it to play "devil's advocate" during its development. It's a good tool, but if it's passing with flying colors while I'm crashing while gaming it may have room for improvement.

Zka17
08-13-2013, 05:52 PM
As far as I know, games are stressing both CPU+RAM and graphics for long time... the best test for your OC would be those games (it is really hard find any comparable stressing) - but, if you want heat and you don't mind cooking your hardware, do Prime95 with Furmark... :cool:

chrisnyc75
08-13-2013, 06:08 PM
As far as I know, games are stressing both CPU+RAM and graphics for long time... the best test for your OC would be those games (it is really hard find any comparable stressing) - but, if you want heat and you don't mind cooking your hardware, do Prime95 with Furmark... :cool:

lol I'm not a fan of roasting my hardware -- trying to avoid that is what brought me to RealBench in the first place. But unless there's some other test that can help me avoid crashing in the midst of a boss fight it's pretty much the reality I have to work with.

Fwiw, I'll be satisfied when I can pass IBT x5 and Prime for 1 hour. I won't slow-roast it (unless it comes to that.... *sigh*)

chrisnyc75
08-15-2013, 01:58 PM
Just a quick update: after some trial & error, I eventually settled on 1.465v vcore (@4.75ghz) at which I'm stable enough to pass 30 minutes of Prime, and run Heaven on loop while Norton virus scanning for 2 hours (I thought this was a good way to test simultaneous stress to gpu+cpu without subjecting my hardware to Prime95/IBT/LINX torture). After all that, as of last night I was able to game for 4 hours without crashing and with temps well under control, all of which means my oc is finally rock solid [I really hope!].

The pertinent take-away, though, is that I was able to pass multiple passes through RealBench a full .04v lower than the ultimate vcore it took me to get "functional 24/7 gaming stable". In my mind, I think that means that if RealBench is to be considered a true test of oc stability, the multitasking portion should be more intense, or a segment that tests 3d rendering while doing cpu calculations should be added.

Just my 2 cents worth of constructive criticism about my experience. :)

Nodens
08-19-2013, 11:24 AM
Version 2 of RB will improve on it's stress test mode (which is currently weak) in a way that is better than the 2nd pass of the encoding test and on par if not better than Prime95/LinX while not reaching the temperatures they do. I plan on stressing every subsystem available. Patience, I'm working on it.:)

chrisnyc75
08-19-2013, 05:34 PM
Version 2 of RB will improve on it's stress test mode (which is currently weak) in a way that is better than the 2nd pass of the encoding test and on par if not better than Prime95/LinX while not reaching the temperatures they do. I plan on stressing every subsystem available. Patience, I'm working on it.:)

Music to my ears. :D

HiVizMan
08-20-2013, 11:20 AM
Nice one mate, gaming is the best stress test there is for the entire system.