|
What about a middle ground of a 1440P ultrawide at 100-120Hz Gsync?
|
#
?
Oct 5, 2018 22:39
|
|
|
|
| # ? Jun 5, 2024 21:34 |
|
|
EdEddnEddy posted:What about a middle ground of a 1440P ultrawide at 100-120Hz Gsync? I really want higher PPI than what my 1440p monitor offers, I enjoy having a second monitor for stuff, and the games I play don't always benefit from having more screen space (starcraft for one). I'd try a 4k ultrawide if it had the same PPI as a normal aspect ratio 4k monitor but so far they all seem to be no more than 1600p tall. VelociBacon fucked around with this message at 23:12 on Oct 5, 2018 |
|
#
?
Oct 5, 2018 22:42
|
|
|
VelociBacon posted:I really want higher PPI than what my 1440p monitor offers, I enjoy having a second monitor for stuff, and the games I play don't always benefit from having more screen space (starcraft for one). What about this?
|
|
#
?
Oct 5, 2018 23:09
|
|
|
Only 60hz unfortunately. That's a cool panel though. e: I think I asked this already but I looked through dozens of pages and couldn't find a response. What is the defacto software for GPU stress testing these days? 3dmark doesn't cut it and I hate spending hours dialing an OC on some software and then I get crashes in a game later that week. Furmark doesn't seem stressful enough?
|
|
#
?
Oct 5, 2018 23:11
|
|
|
VelociBacon posted:Only 60hz unfortunately. That's a cool panel though. Yeah, I don't think you're going to find any high-refresh 5K2K panels yet, maybe in a year or two. Even the 1600p panels are only like 75 Hz at most. Be nice if the 5K2K panels were curved, too, 34" ultrawides are nicer with a curve. I think modern GPUs/drivers notice when they're running Furmark and throttle down. Furmark pulls a lot of power but it doesn't necessarily stress the whole card in the same ways a game will. Really, there's no such thing as a perfectly stable overclock, there can always be one game that just shits out on a nominally-stable OC and forces you to widen your safety margin. Personally for me that title is Just Cause 3, my stable point there on a 780 Ti was like 100 MHz lower than any other game I played (very significant for a 780 Ti), but ymmv. Metro LL was known to be a pretty good workout back in the day. Maybe BF1? Witcher 3 (especially in anticipation of Cyberpunk 2077)? I guess if you wanted to approach it scientifically, maybe try a game from each of the major engines? So like, an Unreal game, a Cryengine game, a Frostbite game, and so on? There's Source too, but Source usually isn't very stressful and is often CPU-limited, which makes it hard to test, but maybe Titanfall 2 with DSR enabled or something? To be real honest I've had no issues with anything Maxwell or Pascal. The delineation between stability and instability is very sharp on modern chips, in my experience. Dunno if it's a factor of the node, or of the power management systems getting better, but Kepler had a big 'ol "mushy zone" of kinda-stability and Pascal pretty much just overclocks till it doesn't. Paul MaudDib fucked around with this message at 23:31 on Oct 5, 2018 |
|
#
?
Oct 5, 2018 23:20
|
|
|
Unigine usually has some good ones that can be run in a loop and even the old Haven one can push a GPU hard enough to crash still. And what do you mean you want a 2nd monitor for stuff? Like I run an Ultrawide and a 2nd 1440P monitor off to the side still for other things. 4K content editing I can see bringing a need for a higher PPI display, but outside of that unless you go physically bigger 4K starts to make things pretty small on anything smaller than a 32" without scaling. Though I agree that Starcraft (and Blizzard stuff in general) just doesn't allow an Ultrawide to utilize itself to full effect for "reasons". However man do other games very much so (Destiny, Racing Sims, Space Sims, hell even small stuff like Oxygen Not Included lol) EdEddnEddy fucked around with this message at 23:24 on Oct 5, 2018 |
|
#
?
Oct 5, 2018 23:20
|
|
|
EdEddnEddy posted:Unigine usually has some good ones that can be run in a loop and even the old Haven one can push a GPU hard enough to crash still. Yeah I have superposition and use it a bit. Maybe I'll try it this time around for my stress testing. I get what you're saying about being able to run an ultrawide with a second monitor also - for me it's just that I already have a monitor stand that works for me and I'd probably have to replace it into two seperate stands if one of them was an ultrawide. I just like the higher PPI displays for the fidelity. I can't always tell the difference between anything above 90hz refresh rate but I can absolutely tell the difference between 2k and 4k. As others have said, it's not a great time to buy a 4k high refresh monitor. Going to wait a bit and see what comes about.
|
|
#
?
Oct 5, 2018 23:35
|
|
|
Regarding overclocking: It looks like folks are getting about +170mhz to core and +700mhz to memory over the FE factory clocks. I'm interested in seeing how ray tracing may affect OCing. It'll use hardware on the card that's otherwise idle, and I'm curious if it'll affect stability at all.
|
|
#
?
Oct 5, 2018 23:49
|
|
|
Back when I OCd a lot more I never found anything that stress tested well enough to give you stability confidence. Same with the CPU for that matter. I tried even combining programs. However I agree with Unigen Heaven 4.0 being the best crasher out of all of them. Conveniently one of the hardest sequences is right at the beginning when you pan onto the... air ship and look at some green vial. Once I was stable there I'd be no more than 26 mhz off stable, and usually just 13. Memory was more of a crapshoot but it's easier to leave a healthier margin there. The best real world testing for me seemed to be graphically intense online fps shooters like battlefield with everything cranked including an unhealthy dose of resolution scaling
|
|
#
?
Oct 6, 2018 00:03
|
|
|
The way I like to do it is with a synthetic app overnight and then for 2 weeks of daily use stability issues probably are overclocking related and not from whatever app or game I was running having issues. After a few months it's the apps fault and not overclocking related anymore. I'm at 4.3GHz on my CPU and 2050MHz on my GPU and I'd sooner think of stability issues as app problems after they've been overclocked for more than a year.
|
|
#
?
Oct 6, 2018 00:18
|
|
|
I always used Prime 95 to test my CPU overclocking (and still do use the newest version without the AVX instructions), but I remember looking at the overclock.net guide that recommended everyone just use x264. I used it and was getting stable overclocks with it, and of course when I tried to play PUBG my computer would hard lock. I double checked the OC in P95, and of course it failed. I ended up needed to give it some more voltage to get it completely stable. Be wary of results that seem too good to be true, and don't be afraid to try different methods to see what works.
|
|
#
?
Oct 6, 2018 00:38
|
|
|
OCCT is also a good CPU OC tester. However it does sometimes pull more power than even encoding with X264/Gaming so it is good for finding the perfect OC stability, but sometimes it pushes so far beyond what any workload would ask that it can make voltage/heat excessive.
|
|
#
?
Oct 6, 2018 00:42
|
|
|
il serpente cosmico posted:Regarding overclocking: It looks like folks are getting about +170mhz to core and +700mhz to memory over the FE factory clocks. If it's like Pascal, you can crank down the power limit a bit and push the core clocks back up to "undervolt" and get the same clocks at lower power limit, which often translates to better sustained performance. In mining, you typically get ~95% of the performance as stock voltage, at 65% of the power. Really I'd see what the autotuner program comes up with first though. il serpente cosmico posted:I always used Prime 95 to test my CPU overclocking (and still do use the newest version without the AVX instructions), but I remember looking at the overclock.net guide that recommended everyone just use x264. I used it and was getting stable overclocks with it, and of course when I tried to play PUBG my computer would hard lock. I double checked the OC in P95, and of course it failed. I ended up needed to give it some more voltage to get it completely stable. Do be aware that Prime95 runs your AVX full out, which means your CPU applies the AVX offset (typically -2 multiplier). So when gaming kicks the clocks up 200 MHz higher than Prime, you can crash. Paul MaudDib fucked around with this message at 01:31 on Oct 6, 2018 |
|
#
?
Oct 6, 2018 01:26
|
|
|
il serpente cosmico posted:I always used Prime 95 to test my CPU overclocking (and still do use the newest version without the AVX instructions), but I remember looking at the overclock.net guide that recommended everyone just use x264. I used it and was getting stable overclocks with it, and of course when I tried to play PUBG my computer would hard lock. I double checked the OC in P95, and of course it failed. I ended up needed to give it some more voltage to get it completely stable. Oh I didn't know p95 was safe now. Do you need to select anything for the tests to make it safe? 
VelociBacon fucked around with this message at 01:33 on Oct 6, 2018 |
|
#
?
Oct 6, 2018 01:31
|
|
|
VelociBacon posted:Oh I didn't know p95 was safe now. Do you need to select anything for the tests to make it safe? I'd at least go with blend instead of smallfft, at least back in the X99 days Asus outright said that sustained Prime95 load could damage your CPU (probably electromigration). A 5960X can pull well over 400W in Prime95, which isn't super great for the VRMs either. Given that AVX offset is a thing now, and that you can be stable in Prime95 and not stable in other stuff, it's not even that good a test anymore 
Paul MaudDib fucked around with this message at 01:36 on Oct 6, 2018 |
|
#
?
Oct 6, 2018 01:32
|
|
|
To clarify: I use an old version of prime without AVX instructions. I didnt word my post well. I use version 26.6, and I don't run it beyond overnight.
il serpente cosmico fucked around with this message at 01:37 on Oct 6, 2018 |
|
#
?
Oct 6, 2018 01:33
|
|
|
il serpente cosmico posted:To clarify: I use an old version of prime without AVX instructions. I didnt word my post well. Which version and which settings do you use for testing? I'd love to test with it but I'm scared of pooching my cpu.
|
|
#
?
Oct 6, 2018 01:36
|
|
|
Prime is more a cooling and power supply test. I don't think it was ever dangerous, but it tests a limited set of CPU functions and makes the most heat. If you're doing something weird like running it for a week it could be bad, but I ran it overnight at 90C and 4.7GHz on a Haswell with no ill effects.
|
|
#
?
Oct 6, 2018 01:36
|
|
|
VelociBacon posted:Which version and which settings do you use for testing? I'd love to test with it but I'm scared of pooching my cpu. I always use version 26.6 with the default settings overnight.
|
|
#
?
Oct 6, 2018 01:38
|
|
|
il serpente cosmico posted:I always use version 26.6 with the default settings overnight. Thanks.
|
|
#
?
Oct 6, 2018 01:39
|
|
|
I run the latest for a few hours (sometimes it crashes/errors when other stuff doesn't), which I can't imagine to be that much more damaging than running "normal" 100% loads with the same overclock for hours at a time over years.
|
|
#
?
Oct 6, 2018 01:45
|
|
|
Remember how Fermi was so bad Nvidia spun stability testing into running a virus on your GPU? Fun times, the only Nvidia GPU I never got. I went to SLI 280s at the time, SLI keeps getting worse so I'll never do that again either.
|
|
#
?
Oct 6, 2018 01:46
|
|
|
http://www.tomshardware.com/answers/id-2985195/4790k-hitting-100-degrees-load-seconds.html
|
|
#
?
Oct 6, 2018 01:47
|
|
|
That's just wrong, I have 0 AVX offset because everything uses AVX these days and have no problems. If you have bad cooling it could be a problem, but watch your temps and you'll be fine.
|
|
#
?
Oct 6, 2018 01:52
|
|
|
Asus's Haswell-E Overclocking guide:quote:Users should avoid running Prime95 small FFTs on 5960X CPUs when overclocked. Over 4.4GHz, the Prime software pulls 400W of power through the CPU. It is possible this can cause internal degradation of processor components. 24 hours of 400W load can easily be worse than years of 200W load. Electromigration is an exponential effect, not a linear one. The more power you pull and the hotter you run, the worse it's gonna be.
|
|
#
?
Oct 6, 2018 01:53
|
|
|
Paul MaudDib posted:Asus's Haswell-E Overclocking guide: In any case I'd expect anyone pushing the limits of >300W CPUs to just use some actual common sense in this case.
|
|
#
?
Oct 6, 2018 02:06
|
|
|
craig588 posted:That's just wrong, I have 0 AVX offset because everything uses AVX these days and have no problems. If you have bad cooling it could be a problem, but watch your temps and you'll be fine. I don't use any AVX offset, but I also don't use newer versions of prime95 because it pulls an insane amount of power and increases the CPU temperature beyond anything else I've ever seen. I've never had an OC that was stable in 26.6 overnight that was unstable anywhere else, so I figure it's a good enough test for my purposes.
|
|
#
?
Oct 6, 2018 02:09
|
|
|
FFXV has a free downloadable benchmark and it’s really nice. Same graphic options as full game, you can loop it, looks pretty and uses both CPU and GPU. Counts frames, etc. I also use Asscreed Origins benchmark because goddamn that game is badly optimized, but it’s extremely annoying to test things because it feels like every setting change wants a restart Edit: there’s also blender’s BMW scene Comfy Fleece Sweater fucked around with this message at 02:21 on Oct 6, 2018 |
|
#
?
Oct 6, 2018 02:14
|
|
|
I was really curious to see how much wattage P95 pulls at load with small FFTs. I'm running a coffee lake at 4.8ghz across all cores. According to my UPS, at idle I'm running it's putting out 85 watts (this includes my PC, monitor, router and modem). On 26.6, Small FFTs, the wattage ramps up to 223 total, so an increase of 138 watts from idle to full load. On 29.4, Small FFTs, the wattage ramps up to 255 total, so an increse of 170 watts from idle to full load. That's a pretty significant increase between the two versions.
|
|
#
?
Oct 6, 2018 02:30
|
|
|
Deviant posted:
Can't wait to see how Serious Sam 4 makes it poo poo itself when it comes out.
|
|
#
?
Oct 6, 2018 04:37
|
|
|
There's nothing special about prime95 that makes it more dangerous than anything else, it's just a way to generate a highly artificial stress test of the AVX units without ever waiting for memory or anything really. Overheating the VRM on the motherboard will shorten its lifespan though, sure, but if you have a good motherboard it's usually not a problem. Stability testing with prime95 is not very useful because it's about as far from a real workload as you can get and it only really stresses a small part of the CPU anyway - it's only really useful to stress test VRM and cooling. OCCT's small dataset test will draw almost as much power but is better at actually finding instability. One of the main reasons these tests use so much power (other than using the very power hungry AVX units) is that they use a very small dataset that mostly fits in the CPU cache, so they never wait for RAM and just run the compute units as fast as they can possibly go. No real workload you will ever run behaves like that. On my 8700K (running 4.9GHz at around 1.35v with no AVX offset), prime95's small FFT's and OCCT's small dataset test will draw around 180W to the CPU. A more realistic workload like Asus Realbench (which is mostly x264 plus some other stuff that also stresses the GPU at the same time) will draw 140W, so we're talking 25-30% more power for p95/OCCT small. A general rule of overclocking thumb says that for ambient cooling you're pretty safe as far as power consumption goes as long as you don't exceed double the stock TDP, which on the 8700K is 95W. Maybe this is a better fit for the overclocking thread, but in my admittedly limited experience, if it passes ten minutes of OCCT's small dataset test without any Windows WHEA warnings, it's usually going to pass 8 hours of Realbench too, and if it passes 8 hours of Realbench it's good enough for daily use. TheFluff fucked around with this message at 07:47 on Oct 6, 2018 |
|
#
?
Oct 6, 2018 07:38
|
|
|
I just occt test them overnight now.
|
|
#
?
Oct 6, 2018 08:15
|
|
|
You know you can just edit a file to tell Prime95 (any version) to run without AVX right?
|
|
#
?
Oct 6, 2018 11:45
|
|
|
Odly enough my 2080 FE is rock solid on OCCT but fail all the time when playing any game (crash to desktop), the worst culprit being Shadow of the tomb raider with a crash every 30min on average. However disabling the factory overclock (debug mode) make everything super stable. Really tempted to RMA it as I feel i've exausted all options.
|
|
#
?
Oct 6, 2018 13:41
|
|
|
Krogort posted:Odly enough my 2080 FE is rock solid on OCCT but fail all the time when playing any game (crash to desktop), the worst culprit being Shadow of the tomb raider with a crash every 30min on average. Could be PSU related, have you tried swapping it out (or just running the PCIe power from a spare PSU). If that's fine yeah I'd RMA it sooner than later but that's really weird to me.
|
|
#
?
Oct 6, 2018 15:12
|
|
|
Should I be worried that my cpu is possibly holding me back? I wasn't really worried about it before, but all the cpu talk has me concerned.
|
|
#
?
Oct 6, 2018 15:28
|
|
|
Is it overclocked? It's not. If it's not overclocked it's time to read up on that. Even back to Sandybridges if they overclock well you're not missing much unless you can use 6-8 cores.
|
|
#
?
Oct 6, 2018 15:34
|
|
|
Spaseman posted:Should I be worried that my cpu is possibly holding me back? I wasn't really worried about it before, but all the cpu talk has me concerned. Whether or not your CPU is throttling you depends on what you're doing with your PC and what your GPU is. If you have non-stock cooling I'd read up on overclocks just because there is nothing to lose.
|
|
#
?
Oct 6, 2018 15:39
|
|
|
How much depends what you're playing, what resolution/settings, and how OC'd it is, but yes, modern CPUs are faster. Additionally, ram speed keeps mattering more and more, especially for minimum frame times. If you bought the slowish ram goons are always saying you can get away with, it's probably making a significant performance difference 4 years down the line.
|
|
#
?
Oct 6, 2018 15:45
|
|
|
|
| # ? Jun 5, 2024 21:34 |
|
|
VelociBacon posted:Could be PSU related, have you tried swapping it out (or just running the PCIe power from a spare PSU). If that's fine yeah I'd RMA it sooner than later but that's really weird to me. Unfortunately I only have 1 Seasonic Focus gold 650w on hand. I used to be running a GTX 1070 with a big OC without problem before swaping to the 2080.
|
|
#
?
Oct 6, 2018 15:59
|
|