10 replies to this topic

[Sen]

Yea, it's me. . .I was hoping to get a little constructive feedback on my system. . .and ESPECIALLY wondering if I'm expecting more from my computer than I should be.

Desktop:

3930k @ 4.7 Cooled by a standard H100 4 fan push/pull.
2 x 7950 @ 925/1325
OCZ 1000w gold cert PSU
Asus Rampage IV extreme
This is a long convoluted fuster cluck, so I'm gonna sum up the past several days.

Monday:
Bought Far Cry 3. Can't run it in Xfire or it crashes [yes, it's patched] can't run the card O/C or it crashes. Even stock single card, I get some artifacting and high temp of 76c. When i say crash, I mean just the program. The computer itself was stable.

I suspect noobie-soft is to blame.

Today: Hook computer up to run MwO after seeing some WONDERFUL panoramic pics of eyefinity Phract and all the yummy extra field of view goodies. . . but at 6058 x 1080 it's on the edge of playable in Xfire, and 2-5 FPS higher with a single card. this is at 925/1350

Out of sheer curiosity I hooked up the 6 pin auxiliary PCI-E cable to my motherboard. Previous to this the 2nd card did not run while I was on Windows Desktop, and my idle temps were 39-42c. Afterwards my idle temps are around 55c and my 2nd card does not shut down. Since I've seen my old pair of 6970s do this when in Xfire, I'm ASSUMING I actually needed this plugged in, though I find it odd that only two cards would require this. HOWEVER, i was able to play for over an hour in Far Cry 3 on medium settings in 6058 x 1080 before I got a crash. Of course this crash was the computer, and it was accompanied by a brilliant flashing purple screen. Highest recorded temps were 79c. Hot, but not out of temp range.

Then I went back to MwO to see if I got any kind of a performance change. . but it's still the same. max FPS of 32, averages 25-27, dips on rare occasion to 19. 79c max temp on main card. This is on max settings.

Lowering the settings doesn't gain me more than a few FPS. it's ALMOST smooth. ALMOST.

What I'm trying to determine:

1) is one of my cards not entirely stable? The artifacting may suggest it, but I had *ZERO* artifacting after connecting the aux motherboard power. While I was playing, I had short periods where I got MAJOR frame lag, and then it ironed itself out. Sound like a not quite stable overclock?

2) Am I right to leave that aux power plugged into the mobo? I could've sworn that was only for 3 or 4 card setups.

3) Am I expecting too much from my machine? I have delusions that my 2 x 7950 should play all games on eyefinity @ max settings. I know for certain I'm deluded when it comes to Far Cry 3. . but MwO? Even in eyefinity I can't see it being anywhere near as demanding. This leads me into. . .
4) Am i expecting more from MwO than I should? Obviously they don't have xfire support, no DX11, they're still working on optimizing performance [with a ways to go] but i feel I should still somehow be getting more than I am. . probably again because I feel it's nowhere near as demanding as [FC3] today's new games.

5) Should I go all out and buy another PSU and another 7950? My PSU supports 3 way Xfire. . but I had to use one of the PCI-E connectors slots on the PSU to power the motherboard aux PCI-E >.> I'm ASSUMING this is ok, because the plug fit into the motherboard, and it wouldn't have if it wasn't meant to [and because it's been 5 hours and my PC hasn't blown up yet]

I'm usually a fairly educated bastion of knowledge on this stuff, but on occasion we ALL could use some feedback from our peers [and our betters, I'll swallow a little pride to sus this out] Thus I turn to our wonderful community. I know some of you out there have messed with this particular subject more than i have, so I'll leave it with you as I go back to practice my google fu


Why is attempting to post this flagging the website as infected with malware? I wonder if PGI really DOES have a leak, or if someone reported it out of paranoia? /shrug

Edited by Sen, 13 December 2012 - 05:19 PM.

[Youngblood]

Yes to question #4. Currently it seems like from almost all systems anywhere, the primary bottleneck for people right now is the CPU for MWO. There's just so much code that the game is offloading...have you tried measuring your average, minimum, and maximum FPS on just a single monitor yet?

[Barbaric Soul]

Quote

1) is one of my cards not entirely stable? The artifacting may suggest it, but I had *ZERO* artifacting after connecting the aux motherboard power. While I was playing, I had short periods where I got MAJOR frame lag, and then it ironed itself out. Sound like a not quite stable overclock?

Only way to tell that is test both cards individually

Quote

2) Am I right to leave that aux power plugged into the mobo? I could've sworn that was only for 3 or 4 card setups.

If it works, yes.

Quote

3) Am I expecting too much from my machine? I have delusions that my 2 x 7950 should play all games on eyefinity @ max settings. I know for certain I'm deluded when it comes to Far Cry 3. . but MwO? Even in eyefinity I can't see it being anywhere near as demanding.

MWO is the most demanding game I have, except for maybe Metro 2033 which I just got and haven't fully check out yet. At 2560*1600 resolution, with a single 7970 and 2600k OC'ed the 4.2ghz, I average in the 45-50 fps range, all settings at max. Considering your trying to play at 6058*1080 while effectively only having a single 7950(crossfire not being supported, you may as well take the 2nd 7950 out and save on electricity), I think you may be asking too much of your system with the unotimized MW:O, but not asking too much in Far Cry(I would think Far Cry should be properly set to use crossfireX set ups).

Quote

4) Am i expecting more from MwO than I should? Obviously they don't have xfire support, no DX11, they're still working on optimizing performance [with a ways to go] but i feel I should still somehow be getting more than I am. . probably again because I feel it's nowhere near as demanding as [FC3] today's new games.

Yes, you are expecting too much from MW:O in it's current form.

Quote

5) Should I go all out and buy another PSU and another 7950? My PSU supports 3 way Xfire. . but I had to use one of the PCI-E connectors slots on the PSU to power the motherboard aux PCI-E >.> I'm ASSUMING this is ok, because the plug fit into the motherboard, and it wouldn't have if it wasn't meant to [and because it's been 5 hours and my PC hasn't blown up yet

No. Let the Devs get the game optimized. You have a killer system, just enjoy what you have. A 3rd 7950 will not help you out IMHO.

[xenoglyph]

Your other questions have been answered, but regarding the OC stability...sounds like it's not completely stable. I don't know what to recommend to test it. I used to run bitcoin miners and used OCed AMD cards in 3 way xfire. So basically I did my stability testing with the GPGPU software. It actually took me a ton of tweaking to get the most out of the 6950s I was using at the time.

All my OCing was done with MSI Afterburner.

Some of the unstable overclocks would last for 3-8 hours before crashing. Stable overclocks could run for days at a time. Like I said, I don't know what to recommend for testing, but make sure you run it overnight.

CPU OC stability is generally easy to ascertain IMO, but you should prolly run Prime95 Large FFTs overnight too

[Sen]

thanks for the replies guys. . helps to get outside feedback after you've had your head in your rig for awhile. As an update, after I got done playing tonight I pulled the top off my case. . . and 2nd GPU was cold, fans weren't even spinning. This was after going back to a single monitor [which was getting over 80 FPS], so i guess that answers that part. I'm just surprised, I thought there was at least basic Xfire support baked into the drivers. . guess MwO killed THAT idea.

Going back and priming now. Decided to update BIOS to latest this morning, they COMPLETELY changed how manual vcore voltage adjustment is displayed, so I had to go back and do it anyway. On the bright side, either because of the update OR my not doing enough testing, it looks like I was WAY over volts at 1.31. With vdroop, I'm currently priming @ 1.248 [it's set for 1.275] and no issues so far.

The only thing bothering me is my CPUTIN. It's an ASUS board. THEY ARE INFAMOUS for mis-displaying CPUTIN. it's showing between 69 and 93c, but it goes up and down in single digit changes, which makes it LOOK like it's reading correctly. Max temp on hottest core so far is 66c, but it shows 69-70c on CPUTIN at IDLE [most cores high 20s, low 30s]

Edited by Sen, 14 December 2012 - 06:38 AM.

[Oderint dum Metuant]

It sounds to me that the overclock on the cards was not quite stable.

Also is that OCZ PSU multi or single rail? my line of thinking here, is that if it is multi rail, do you have a dodgy or overused rail powering one of the cards.
the 6pin AUX cable from another rail may somewhat stabilize that issue in shunting power to the card.

For me your GPU temperatures are safe, but my OCD says too warm, i can't seem to find a clear answer if the 7 series has a thermal throttling point, i know with Nvida its in the 70's.

[Sir Roland MXIII]

As far as question 1, I had the same thing happen to me when OCing my APU on the graphics side - it was because I was trying to find the right amount of voltage. So, to my eyes, that looks like OC instability - just my two cents.

Also as far as the, is it CryEngine or is it me? Honestly I think the engine is at BEST a decent engine but there are much, MUCH, better ones out there, some which are 5 years old now and still more impressive than CE3. So, in short, my answer to the "is it me or CryEngine", my answer is always - it's probably not you.

I assume you've tried this setup in other, non-CE3 games? If not that'd answer the question of Cry vs Computer.

[Sen]

Quote

is that OCZ PSU multi or single rail?

Linked for most efficient communication:

http://www.newegg.co...N82E16817341046

Quote

For me your GPU temperatures are safe, but my OCD says too warm

You're inside my head now Before I put the PCI-E connector on the motherboard things liked to get a bit . . dodgy at 76c. With it on, even last night in MwO temps were pushing 82c! [but not a single gfx flicker or artifact or microstutter did I see/feel] intellectually I am aware that these cards should be fine at 90-95c +, and they will throttle before they melt. . but it just doesn't FEEL right somehow. Since someone will ask:

http://www.newegg.co...N82E16811119265

4 front fans [two on either side of the H100, vented sides and top, and a 120mm exhaust fan. Cards exhaust into the case, through the top vent, and out. I tried rigging a 200mm fan to the top, made 0 difference and looked bad.

I started researching a full closed loop water cooling system, but oi ve' that stuff gets expensive quick

Quote

All my OCing was done with MSI Afterburner

in a fit of not awake, I missed this this morning. . . I tried afterburner yesterday so I could set a custom fan ramp up. . but if I tried to O/C the cards at all, I'd get weird graphics shake everytime I drug the mouse across the "monitor" side of the program. Open a browser window, and the whole screen shook in exactly the same way. I'm gonna have to go back and re test everything on single monitor at this point, get it stable, then throw it back on the eyefinity and see what happens.

Quote

is it CryEngine or is it me?

You're probably half right on this one. In my delusions of graneur to get 1k FPS across 3 screens in MwO [and Far Cry 3. . just to torture the kid] I'm starting to really shake the cobwebs out of this thing. It's always been hit-or-miss stable, but [sadly] I don't always push this thing as hard as I could. I can transcode videos all night with no issues, then it'll random BSOD during web surfing [LLC issues] but the instances so rare it's hard to wonder if it's just a one off occurrence. if nothing else, the feedback is helping me chase down small transient things, think a little differently, and chase down 100% stability so I can come back and get some more predictable results. Today's BIOS update *REALLY* seemed to help with this. Now I actually input the manual voltage, so instead of "add .030v to random number " it's "1.3v = 1.3v minus Vdroop.

Next step [after paying for a new computer's worth of car repairs on my wife's vehicle] is to go back into FC3 with a single screen, max everything out, keep my O/C, and see what shakes loose.

Cy'all in a few hours ^^

Edited by Sen, 14 December 2012 - 12:17 PM.

[Barbaric Soul]

Quote

in a fit of not awake, I missed this this morning. . . I tried afterburner yesterday so I could set a custom fan ramp up. . but if I tried to O/C the cards at all, I'd get weird graphics shake everytime I drug the mouse across the "monitor" side of the program. Open a browser window, and the whole screen shook in exactly the same way. I'm gonna have to go back and re test everything on single monitor at this point, get it stable, then throw it back on the eyefinity and see what happens.

I get the same thing with my 7970 when I OC it ever since the 12.11 beta drivers were relased. Try older drivers. I don't have my 7970 OC'ed so it isn't an issue for me right now.

[awdwikisi]

1. mwo is still beta, so while that shouldn't be the catch all for oddities, atm its just testing that way
2. YES, for my ch5 board it wasnt even remotely stable over all untill i plugged in the extra power in.
(my board is a bit different, have an extra 4pincpu and a 4 pin molex for the pcie slots)
3. unsure to a possiblity of no, having tested multiple bulldozer's, quadsocket opterons, thubans, phenoms, 1x6970/3x5870's/2x7950's/3x7970lightning's
i can safely say that anyone running similar system and rez at 5760x1080 should easily run medium @50+frames in bf3/ps2 and the like,
considering you have a much better cpu, i would think high, @50+frames.
4. see 1 and this is kinda like crysis 1 imo in terms of hardware requirements
5. about powersupplies, yes and no, yes because powersupplies are never 100% efficeint and most likely only 80% efficeint
and you could be close to that. no because i cannot believe that a 3930 draws more watts than a 8120 ocd, can be wrong but...
another video card will not help as through multiple people that have replied here xfire doesnt work atm. second getting another card will only exacerbate what issue you already have, and in my experience you need 2 powersupplies to run 3+card configs well, as single 1500watt psu's are not enough
the result i got must be anomaly as i cannot reproduce it in another computer outside my 8120/ud3, ive tried all of last week and nothing to show

imo any graphic card past 70c gets me twitchy
fresh installs of windows for every oc test, windows/catalyts seems to get angsty when oc fails....also could be my paranoia
what type of ram dimms, i noticed 4x8gig modules do no like to work well with high oc's

power draw of cards
http://forums.atomic...p?showtopic=264

[Sen]

At this point it's as stable as I think I can get it. If i leave chrome open and leave it for a few hours, it'll freeze, the screen won't come back on. . reminiscent of not enough voltage at idle/LLC issue. If I don't leave chrome open, it doesn't do it.

Far Cry is now stable at 900/1300 xfire minus an occasional glitch where it'll crash when saving. . and the very occasional random artifacting. . but so rare as to no longer be a concern.

At this point, it's MwO just not using my hardware. Just have to wait until support is added in.

Oh, and about that thing with 21.11 beta and afterburner. . the gain from afterburner isn't worth the loss from the 12.11 beta drivers. . .the 7k series performance boost is AMAZING ^^