|
Odette posted:OK, so there's no way of actually fixing this until NVIDIA update their drivers? Got it. Gracefully? Not really. Best you can do is try setting it to 120Hz and hoping it fixes it--for some people that's all it takes. If that (or 119Hz I guess) doesn't work, you're stuck with trying some klunky work-arounds. NVidia Inspector w/multi-monitor mode (which doesn't actually require you to be using multiple monitors, incidentally) does work, it just doesn't work for me because I'm a giant ADHD child and jump around all the god damned time. If you're on a single monitor and just do one thing at a time, it'd be a fine solution. e; on the up-side, NVidia has finally admitted that this is actually A Problem, vice the "working as intended" line they've been giving out for the last...well however long it's been since the first high Hz monitor popped up.
|
# ? Dec 20, 2015 08:28 |
|
|
# ? May 30, 2024 11:32 |
|
DrDork posted:Gracefully? Not really. Best you can do is try setting it to 120Hz and hoping it fixes it--for some people that's all it takes. If that (or 119Hz I guess) doesn't work, you're stuck with trying some klunky work-arounds. NVidia Inspector w/multi-monitor mode (which doesn't actually require you to be using multiple monitors, incidentally) does work, it just doesn't work for me because I'm a giant ADHD child and jump around all the god damned time. If you're on a single monitor and just do one thing at a time, it'd be a fine solution. They admitted it was a problem because PCPer called their lying asses out on it, that's what.
|
# ? Dec 20, 2015 10:43 |
|
So my 6950 seems to have died today. I hoped I could wait until Pascal for an upgrade but now I need something a lot sooner. I'd rather get an Nvidia card this time around since I had a lot of weird bugs like corrupted mouse cursors with AMD over the years. Should I get a 960 and upgrade earlier or get a 970 and maybe keep it a year longer? I only have a 1080p monitor, so no super high resolutions to worry about. Or should I get an AMD card despite my bad experiences?
|
# ? Dec 20, 2015 10:59 |
|
i'd suggest a 970, because if you don't want to buy a card for a while at 1080p that and the 390 are the only real options
|
# ? Dec 20, 2015 11:03 |
|
Eezee posted:So my 6950 seems to have died today. I hoped I could wait until Pascal for an upgrade but now I need something a lot sooner. At only 1080p, a 970 should be fine for even demanding games. Prices are really good for them right now as well and is a better overall deal than a 960 4GB. Since you're running a 6950, power shouldn't be an issue unless you have an old PSU. IMHO, if you need it now and it's inside your budget, go GTX 970. It's unknown what Pascals really bringing to the table and you should only wait for that if you're okay with your hardware now.
|
# ? Dec 20, 2015 11:04 |
|
OK, thanks for the advice. I'll order a 970 today. Only being able to use my lovely laptop over the holidays would suck
|
# ? Dec 20, 2015 11:25 |
|
Dogen posted:Serious question: is there enough difference between 120 and 144 that worrying about that is worth it? In all honesty, I can't imagine there being any real difference, and I've always considered 120 to be the more logical refresh rate, seeing as with 60Hz video, you only have to display every frame twice. They both divide by 24 perfectly, so that part is a wash. But there are no doubt plenty of people around here who have a 144Hz monitor, and could set it to 120Hz to see if they can tell the difference for you. (I'm not one of them, I'm probably done replacing my monitor unless it breaks, I'm more interested in what VR headsets can do for me!) Was there a compelling reason why 144Hz supplanted 120Hz anyway? I have a vague feeling it might be purely to deliver a higher refresh rate per eye for 3D content, where of course, even 72Hz wouldn't be that high. HalloKitty fucked around with this message at 11:39 on Dec 20, 2015 |
# ? Dec 20, 2015 11:27 |
|
HalloKitty posted:In all honesty, I can't imagine there being any real difference, and I've always considered 120 to be the more logical refresh rate, seeing as with 60Hz video, you only have to display every frame twice. They both divide by 24 perfectly, so that part is a wash. But there are no doubt plenty of people around here who have a 144Hz monitor, and could set it to 120Hz to see if they can tell the difference for you. (I'm not one of them, I'm probably done replacing my monitor unless it breaks, I'm more interested in what VR headsets can do for me!) I believe it was because you need multiples of 48 for 3D video. (Because movies use 24 FPS, a number we're all familiar with, so 24 FPS per eye) So I guess they just said, hey, we're at 120 already, but 48 * 3 = 144, let's just take a short step up to 144 Hz and bob's your uncle. SwissArmyDruid fucked around with this message at 13:01 on Dec 20, 2015 |
# ? Dec 20, 2015 12:57 |
|
Huh, I'm trying to stabilize my GPU OC in 3dmark firestrike and I'm getting abysmal results. My display drivers kept crashing at a high OC so I'm going back to the drawing board and I tested at 1200MHz core, 7000MHz memory, with my voltage set to the max of 1.21V (so only the core is OC'd by a small amount) and I'm getting 5-6 FPS. Firestrike gave me a 1339 score . My whole system is a high-end build around a 6700k so there shouldn't be a single bottleneck. I'm going to try raising the core clock by 10-20 MHz and running firestrike until it crashes, but I have a suspicion that something isn't right here.
|
# ? Dec 20, 2015 19:30 |
|
It does sound like something isn't right there, but you also shouldn't need voltage to hit 1200 core. Try it with +0mv--voltage bumps (especially on air) do very little for Maxwell, and generally are a good way to prematurely throttle an otherwise functional overclock.
|
# ? Dec 20, 2015 19:49 |
|
I'll try lowering the voltage. I have a water cooling loop and temps didn't go above 30C during the whole test. I don't know much about the throttling I hear these chips do - would it be a reaction to temperatures or could something else cause throttling to take place? E: I dropped voltage to stock (1.6v) and I'm seeing a big jump in FPS. It still isn't reaching far over 60fps (some tests cap at 40) but I certainly improved. What's going on with the throttling and how can I stop it? E2: gently caress, the sword combat part of Firestrike just started and I was getting 4 FPS. My score is now a 5911, which is still bullshit for such a modern system. E3: So I reset everything on the GPU to stock (1076/7000/1.163V) and I'd get about a 5700 score (still WTF low). I then upped only the core clock to 1200 and I just got a 2930 score. Could I have a bad card or am I just doing something wrong? Richard M Nixon fucked around with this message at 20:20 on Dec 20, 2015 |
# ? Dec 20, 2015 19:54 |
|
Maybe some other part of the card is getting hot? VRAM? I've seen that happen with custom cooling on a gpu.
|
# ? Dec 20, 2015 20:29 |
|
Dogen posted:Maybe some other part of the card is getting hot? VRAM? I've seen that happen with custom cooling on a gpu.
|
# ? Dec 20, 2015 20:57 |
DrDork posted:That's the next part I'd look at. You're loop-cooling the GPU, but are you doing anything for the VRAM or any other parts of the card? Generally you need to slap some small heatsinks on them and provide them some airflow via a fan or whatnot. Sadly, most boards do not support VRAM temp monitoring, so it's hard to tell. VRAM does not actually get all that hot, it's the VRMs that can get toasty and those really need heatsinks, heat sinks are helpful on the VRAM too especially if you are OCing the VRAM at all but heatsinks and air flow on VRMs is absolutely necessary.
|
|
# ? Dec 20, 2015 21:00 |
|
What is currently the best way to limit frame rate game by game basis? Also why is DSR still not available for laptops? Sininu fucked around with this message at 21:22 on Dec 20, 2015 |
# ? Dec 20, 2015 21:17 |
|
SinineSiil posted:What is currently the best way to limit frame rate game by game basis? Adaptive VSync, if your card supports it on Nvidia's side. I don't know what AMD calls it in settings. Or through MSI Afterburner/RTSS.
|
# ? Dec 20, 2015 21:35 |
|
Rosoboronexport posted:Adaptive VSync, if your card supports it on Nvidia's side. I don't know what AMD calls it in settings. Or through MSI Afterburner/RTSS. Standard Vsync creates awful feeling mouse lag. Is adaptive Vsync better? I'll try Afterburner.
|
# ? Dec 20, 2015 21:39 |
|
Richard M Nixon posted:I'll try lowering the voltage. I have a water cooling loop and temps didn't go above 30C during the whole test. I don't know much about the throttling I hear these chips do - would it be a reaction to temperatures or could something else cause throttling to take place? Max out the power target, don't touch the voltage, and cold boot the PC after every driver reset. The drivers watch for voltage drops and if the voltage falls out of the safe range then it locks the card to a slow protection mode until the next cold boot. You probably aren't going to be able to max out the stock voltage with the stock power limit anyways so adding voltage is just going to reduce your overclocking ability because it'll have more extreme droop from trying to maintain an even higher voltage than the VRMs are capable of delivering. Watch the graphs for the power usage and voltage and see how they're bouncing around, if you're hitting the maximum of your power limit so that the voltage is dropping then adding voltage directly isn't helping you at all (If you're not even coming close to 100% power limit you're in card protection slow mode and you need another cold boot). If you want more voltage than the stock limit of the power target will let you go to you need to edit the bios. Once you're at a power target level that's high enough that it never throttles the stock voltage then you can look at adding voltage without worrying (as much) about it dropping performance, you can loosen up the allowed voltage droop ranges and not induce driver resets. One trap is that the displayed voltage and clocks are only what the drivers are reporting, they're not a raw sensor readout. You could have a card that reports 1.3V on the graph and actually is giving 1.1V to the GPU and another different configuration that reports 1.1V and actually is giving the GPU 1.3V. If the card is in protection mode it'll still look like it's reaching even absurdly high clock speeds, but it's actually ignoring everything sent to it and locked at a very slow speed and voltage in order to protect the card.
|
# ? Dec 20, 2015 22:01 |
|
SinineSiil posted:What is currently the best way to limit frame rate game by game basis? NVidia Inspector will let you set hard caps on framerate on a per-game basis.
|
# ? Dec 20, 2015 23:35 |
|
craig588 posted:Max out the power target, don't touch the voltage, and cold boot the PC after every driver reset. The drivers watch for voltage drops and if the voltage falls out of the safe range then it locks the card to a slow protection mode until the next cold boot. You probably aren't going to be able to max out the stock voltage with the stock power limit anyways so adding voltage is just going to reduce your overclocking ability because it'll have more extreme droop from trying to maintain an even higher voltage than the VRMs are capable of delivering. Watch the graphs for the power usage and voltage and see how they're bouncing around, if you're hitting the maximum of your power limit so that the voltage is dropping then adding voltage directly isn't helping you at all (If you're not even coming close to 100% power limit you're in card protection slow mode and you need another cold boot). If you want more voltage than the stock limit of the power target will let you go to you need to edit the bios. Once you're at a power target level that's high enough that it never throttles the stock voltage then you can look at adding voltage without worrying (as much) about it dropping performance, you can loosen up the allowed voltage droop ranges and not induce driver resets. Good info, thanks. I wasn't sure what the difference between power target and voltage was so I maxed them both. I'll try with just the target at 110. Is there some kind of motherboard setting to increase power to the PCI express slot or change the PCI frequency that might help stability? DrDork posted:That's the next part I'd look at. You're loop-cooling the GPU, but are you doing anything for the VRAM or any other parts of the card? Generally you need to slap some small heatsinks on them and provide them some airflow via a fan or whatnot. Sadly, most boards do not support VRAM temp monitoring, so it's hard to tell. I have this cooler from EK: There are thermal pads over the memory and the VRM that contact the block, so they should be kept cool. Is there a thermal sensor reading I can see with hwmonitor or one of the GPU tools that will show me those temps?
|
# ? Dec 21, 2015 02:55 |
|
Richard M Nixon posted:I have this cooler from EK: If everything was installed correctly you don't need to worry about VRM temps. I don't think your card supports VRM temp monitoring, I don't even know of any that do.
|
# ? Dec 21, 2015 03:12 |
|
Weird, I switched over to MSI Afterburner from Asus GPU Tweak and now Firestrike is working great. I'm hitting 16k at 1350 MHz and I'm slowly scaling up. I wonder why the Asus tool wasn't working well...especially with my Asus card. So can someone explain the difference between power limit and core voltage? I read a post on Reddit that said it was basically limit = amount of draw the card is allowed and and voltage somehow helps stabalize the OC but it didn't say how. My current strategy is: 1) Should I set draw to 110% 2) Increase clock by small amount (10hz) 3) Run firestrike 4) If score is higher than previous run, goto 2, else 5) increase core voltage by 5mV, then goto 2 Does that sound right?
|
# ? Dec 21, 2015 03:47 |
|
Richard M Nixon posted:Weird, I switched over to MSI Afterburner from Asus GPU Tweak and now Firestrike is working great. I'm hitting 16k at 1350 MHz and I'm slowly scaling up. I wonder why the Asus tool wasn't working well...especially with my Asus card. That's not really correct. Here's your basic stepping stone guide: a) Set power to 110% or set temp to the max it allows you to. ----->Test for stability/benchmark b) Increase core clock by 20mhz ----->Test for stability/benchmark c) When you get instability, raise voltage some amount (12mv?) to increase stability. ----->Test for stability/benchmark Conversly, find the max vcore you're willing to run on your card and then set it to that and just increase your core clock until it's not stable, then go back to the last stable clock.
|
# ? Dec 21, 2015 03:55 |
|
VelociBacon posted:That's not really correct. Here's your basic stepping stone guide: Awesome, I'll give it a shot. I finally hit 1400MHz and I'm up to a 16500 firestrike score. I'm hoping to be able to reach 1450-1500, but we'll see. I checked my ASIC quality and I'm only at 66.5%. I know it's a dubious metric of overclocking ability but it is on the lower end of what I've seen posted online.
|
# ? Dec 21, 2015 04:02 |
|
I have a problem with what I assume is Shadowplay keeping gpu alive when computer isn't doing much. Fans are off when nothing uses the gpu, but sometimes Shadowplay gets stuck active or something (I see it at gpu activity tray icon) and fans turn on for a bit every 5 minutes. Has anyone experienced and solved this? It's a laptop so fans are quite audible.
|
# ? Dec 21, 2015 04:02 |
|
Richard M Nixon posted:Awesome, I'll give it a shot. I finally hit 1400MHz and I'm up to a 16500 firestrike score. I'm hoping to be able to reach 1450-1500, but we'll see. I should mention that even if you're using my stepping stone thing you need to also be looking at what vcore and temp are acceptable to you.
|
# ? Dec 21, 2015 04:10 |
|
If I found a cheap used 670 to SLI my two cards what kind of performance gain would that give me? A decent boost or also not really worth it and I should just hold for Pascal?
|
# ? Dec 21, 2015 05:15 |
|
Richard M Nixon posted:I checked my ASIC quality and I'm only at 66.5%. I know it's a dubious metric of overclocking ability but it is on the lower end of what I've seen posted online.
|
# ? Dec 21, 2015 05:16 |
|
VelociBacon posted:That's not really correct. Here's your basic stepping stone guide: Maxwell cards clock in multiples of 13MHz boost bins, so it is best to increment by 13MHz. Depending on the card and situation it'll round down to 13 or up to 26 from that 20 so it increases variability. On cards that are hitting their TDP limit, max vcore doesn't necessarily mean maximum clocks, as the high voltage means you hit your TDP limit faster. This isn't an issue on the MSI or Gigabyte 970s with their massive TDP limits, but the posters aforementioned Asus card has a very low TDP limit on the stock BIOS (the Asus card at max power limit actually has less wattage to use than the MSI at stock PL).
|
# ? Dec 21, 2015 05:57 |
|
DrDork posted:Lower ASICs are, at least in theory, preferable if you're going with a water loop, because it should let you jam more voltage through it (after a BIOS mod). It'll suck down more watts than an equally clocked high-ASIC chip, but whatever. Oh that's great news. I wasn't sure what it was implying when it said low ASIC quality was good for water cooling. Like I've posted before, I can't seem to warm the card up over 33*C so as long as I'm not putting enough voltage to actually fry the card, I'm happy to pour in as much as it can take. VelociBacon posted:I should mention that even if you're using my stepping stone thing you need to also be looking at what vcore and temp are acceptable to you. Temp was addressed above, but what could constitute an unacceptable vcore? I see people with 980 TIs registering 1700+ MHz on leaderboards so I assume they're taking some significant overclock. My use case for the card is going to be heavy gaming so I don't want to burn the thing out after a month of overclocking. BurritoJustice posted:Maxwell cards clock in multiples of 13MHz boost bins, so it is best to increment by 13MHz. Depending on the card and situation it'll round down to 13 or up to 26 from that 20 so it increases variability. I've only read a small bit on doing a BIOS mod to remove the voltage cap on my card. I know Asus is pretty crappy with their low limit. Is there a good guide you can point me to for doing the mod? Am I right in understanding that I can either grab someone else's modded bios or save mine and mod my own? Is it preferable to DIY or does it require serious skill and it's best to just take a pre-done bios? Should I be trying to find a bios that came from my exact model of card or does vendor not matter? Here's where I'm currently stable in 3dmark. At least, stable through one run. I'm getting a 16688 in Firestrike 1.1. Any thoughts on voltages? I'm not familiar with most of the measurements and all the different volt metrics. E: Why is the GPU clock reported in hwinfo 1499MHz but the boost reported in GPU-Z is 1385? Richard M Nixon fucked around with this message at 06:23 on Dec 21, 2015 |
# ? Dec 21, 2015 06:17 |
|
BurritoJustice posted:On cards that are hitting their TDP limit, max vcore doesn't necessarily mean maximum clocks, as the high voltage means you hit your TDP limit faster. This isn't an issue on the MSI or Gigabyte 970s with their massive TDP limits, but the posters aforementioned Asus card has a very low TDP limit on the stock BIOS (the Asus card at max power limit actually has less wattage to use than the MSI at stock PL). It's pretty funny seeing what exactly this means when review sites pretend Furmark's a representative sample of anything other than Furmark and turn a G1 into a space heater.
|
# ? Dec 21, 2015 07:06 |
|
I do like that I barely touch 100% TDP with my MSI card at 1450/7800.
|
# ? Dec 21, 2015 07:17 |
|
Nvidia has released 361.43, which includes the following fix: > [GM20x] Clock speeds remain above idle at 144 Hz on desktop. [1631144] So, anybody experiencing that should update and check if the issue is resolved.
|
# ? Dec 21, 2015 16:07 |
|
Can't wait to check at home.
|
# ? Dec 21, 2015 16:52 |
|
ChiTownEddie posted:If I found a cheap used 670 to SLI my two cards what kind of performance gain would that give me? A decent boost or also not really worth it and I should just hold for Pascal? Check out the EVGA B-stock sales. I've seen 680s for $40 and 690s as well (forgot the price). If you're spending more than that I'd suggest bumping up to a refurb GTX 700 or Radeon 7000 series card, at a minimum. Pascal is gonna be cool but it's still almost a year out practically speaking. If you need more power now just look for something around the $250 mark (B-stock 970 or refurb 290/x or retail 390). Those are going to hold value better than a 980 Ti.
|
# ? Dec 21, 2015 18:06 |
|
Upgrading this week from a 2gb 660ti to an 8gb 390, am I gonna be happy for atleast 2/3 years with this, assuming that pascal poo poo isn't loving ridiculously faster?
|
# ? Dec 21, 2015 18:12 |
|
Sgt. Cosgrove posted:Upgrading this week from a 2gb 660ti to an 8gb 390, am I gonna be happy for atleast 2/3 years with this, assuming that pascal poo poo isn't loving ridiculously faster? Probably? I'll let you know when I take off my time travel face bag.
|
# ? Dec 21, 2015 18:16 |
|
Paul MaudDib posted:Pascal is gonna be cool but it's still almost a year out practically speaking. Is it really a year away? Goddamn, I want to replace my 290(non-x) and all the upgrade paths just don't seem worthwhile at the time.
|
# ? Dec 21, 2015 18:28 |
|
Tab8715 posted:Is it really a year away? Goddamn, I want to replace my 290(non-x) and all the upgrade paths just don't seem worthwhile at the time. Practically speaking. They will probably have something out in Q2, but it will probably be expensive and it'll take 3-6 more months before the other lines start coming in and the price stabilizing around the Budget/Mid/High/Ultra prices. Or not, who knows?
|
# ? Dec 21, 2015 20:35 |
|
|
# ? May 30, 2024 11:32 |
|
Tab8715 posted:Is it really a year away? Goddamn, I want to replace my 290(non-x) and all the upgrade paths just don't seem worthwhile at the time. Crossfire 290s aren't half bad and probably the best $200 you can spend from where you are. I've seen return ref cards as low as $185 recently which is a steal. Loud and hot but cheap and powerful. I think 390s should also be compatible if you want a non-ref card, but you'll pay an extra $50-100 for it. If you wanted to sell, I've seen Fury Vanilla as low as $400 and that would be about the same total sum as CF 290s. Probably slightly slower in games supporting CF but faster in single-card games and keeps CF as an option down the road. If you have to pay more than that you might as well just jump to a 980 Ti. They'll drop value once the new Pascal Titan variant releases and then again once the x80 Ti releases, but that also means you can just do SLI 980 Tis once that happens. Yeah, apart from Fury and the 980 Ti the market looks pretty much the same since Kepler and Hawaii. The 970 now takes the performance class of the 780 Ti, and AMD released upclocked 8gb Hawaii to match the 980, nothing significant has improved. It's hard to say what will be futureproof with 144hz and 1440p becoming the norm and 4K, ultrawide, and eyefinity starting to make inroads on the enthusiast market. More is definitely better but there are also more shrinks and architecture improvements coming, like always. A nice high end setup will last a while before it drops down to midrange or low-end performance (at which point it will be considered a hot power-guzzler), a bargain setup will go obsolete more quickly. Consider that the 7950 or 7970 (aka R9 280/x) are still very good performers at the low end of midrange 4 years down the road. Personally I'm going to upgrade once people start dumping 980 Tis once Pascal comes out, then upgrade again or SLI once Pascal x80 Tis have passed the shortage level. Given how many 980 Tis are out there I expect they'll drop to $400-425 and then ultimately down to around $375 once there's a 1080 that matches/outperforms it with less power. I also think that while CF Fury X will outperform SLI 980 Tis in many future games and scenarios (4k, etc) it will underperform in some situations and some games due to lower VRAM and be less compelling as an option due to much lower sales volume (thus limited secondhand/return market) compared to 980 Ti. Paul MaudDib fucked around with this message at 21:17 on Dec 21, 2015 |
# ? Dec 21, 2015 20:35 |