EPTx4/x8 replay puts almost 0 load on the GPU
Simple Questions Simple Answers - Page 312
| Forum Index > Tech Support |
|
Cyro
United Kingdom20275 Posts
x4/x8 replay puts almost 0 load on the GPU | ||
|
Belial88
United States5217 Posts
On April 17 2013 16:46 xasuma wrote: I am having a problem with I believe my gpu. When the games gets intense (watching replay at x4 or x8, when streaming and playing, or if i go more than medium settings in sc2) My gpu fan goes really high, and the temperature of my gpu goes up to about 76 c. (not overclocked- gtx 550 ti). When this happens, my computer "kinda" goes to sleep for about 3-4 seconds and comes back. ( after it's back the gpu temperature its a bit lower, but then it just rises up again). Is the gpu overheating? I cleaned it for dust just in case maybe that was affecting it, but it didn't get any better. My gpu's idle temperature is about 35 c. , in the sc2 menus (sc2 is realy the only game i play atm) about 55 c. and when playing, about 70-75 c. (And in the higher end of that is when i run into the trouble i just mentioned). My cpu is i5 2550k @4.2 mhz and my power supply is a 550 wtts. (less than 1 year old). Help.  76*C is cool runnings for most GPUs, like the 550ti. You are good up to basically 100*C on most graphcs cards, such as the 550TI, so 76*C is more than okay. It would be normal that your fan gets high under high load, everything you describe is working as intended. However, what you can do is get a program called MSI Afterburner (or rivatuner... afterburner is just rivatuner with cool skinz but it has furmark with it too). Then, go into settings, and change your fan profile. Since you are running quite cool, you can have your GPU drop it's fan speed, and then max out at 80-90*C (test with furmark stress test, 5 minutes). Or make a smoother fan profile, so your fan goes up incrementally instead of all at once, or the opposite, whatever you like better. Personally I changed my fanprofile so it goes up in steps instead of a slope, it's a lower CPU load (yes, it's already quite low, but since my max temp on my GPU is only 63*C even with max software voltage, it's okay). Also, you should really overclock your GPU. It's not a very strong GPU, and GPUs these days are built for overclocking, they are sold on that. As in, every GPU company these days supports overclocking in their warranties, all of them. The 'worst' of them simply say some grey area on 'reasonable overvoltage' but they can't tell how much you used. If you use software to overclock, like afterburner, the maximum voltage you can give, is still going to be very low, as GPU BIOS are coded so that the maximum voltage you can give in software, is very very low. What PSU do you have exactly? 550w psu doesnt tell us anything. It's about quality, not quantity. It's not like a psu company buys components and they all agree "Well that kind of circuit is good for 24 watts!". No, psu companies just arbitrarily pick a wattage rating - a good company will say "this is 400w" and most companies would sell that same psu as 1000w. And, 300w is more than enough for almost any modern, single gpu intel system, it's actually overkill for them, unless you talk about high end overclocks or the higher end GPU+CPUs. While maybe not the (entirely) correct place to ask: I recently destroyed my SGS3 (luckily, home insurance is paying) and now need a new phone. The SGS3 was a phone that I now think was much better than what I needed. If I'm looking for a new android smartphone that: runs a solid amount of apps like skype (voice only), facebook messenger, twitch, quick internet and spotify, what direction should I be looking? The HTC One X seems ok, but might still be overkill. All I want is a convenient experience with 'normal' apps, not a mobile game center or HD videorecorder. They all do that o_o Any 'media phone', let alone smart phone, can do all of that. With droid phones you generally pick one based on the form factor and keyboard (or lack of). All of the 'current' droids are just the exact same internals, the exact same phones. I think a few benches show differences but I believe that's just within the margin of error - same cpu, same speed, can't really change that. | ||
|
xasuma
Chile62 Posts
(Antec BP550 Plus 550W Continuous Power ATX12V V2.2 80 PLUS Certified Modular Active PFC Power Supply) I just ran furmark with the my current settings (i5 2550k @4.2 mhz and stock GPU values) : And my card got as hot as 86c.(auto fan speeds, that went up to 100%). and at about 1/3 into the testing my computer pretty much went to sleep for 3 seconds. and I got the following error. -"the Nvidia open gl driver lost connection with the display driver due to exceeding the windows time-out limit and is unable to continue, the application must close." --> this is what happens in sc2 also. Since I hadn't tested out my CPU oc with any stress test, I lowered it at @4.0 mhz, however the same thing happened. At about the same time into the test. ----------------------------------------------------------------------- The reason why I wasn't overclocking is because I am still not sure what is causing this, and since sc2 is giving me very good FPS at low/medium (always over 100). I thought I should probably find out about this problem before trying out values for the card. (Which I have tried to oc a little with msi afterburner, but I thought that was the issue and put the values back to stock) . The only thing that I think might be is the fact that I didnt move the voltage numbers at all when I overclocked the cpu ration using AsRock extreme tuning utility. Edit!: I just tested my CPU with Intel Burn Test @4.0 mhz. However during the test, the cpu ratio kept switching between x36, x37 and x40. (3.6 mhz, 3.7 mhz, and 4.0mhz) I did the stress level at standard and at high. But I am just wondering why isn't it locked at 4.0 mhz? :/ | ||
|
divito
Canada1213 Posts
I was curious if anyone happens to know what the lowest wattage would be for a fully modular power supply and who makes one. Thanks. | ||
|
Rachnar
France1526 Posts
OCZ, seasonic come to mind | ||
|
xasuma
Chile62 Posts
On April 18 2013 02:13 divito wrote: I want fully modular power supply for my next system. I currently have a 1050W (fully modular) but this is immensely overkill for my current and future system. From a cursory viewing of some retailers, the lowest I've seen for a fully modular one is 750, which still is more than I need. I was curious if anyone happens to know what the lowest wattage would be for a fully modular power supply and who makes one. Thanks. Amazon list | ||
|
skyR
Canada13817 Posts
| ||
|
Myrmidon
United States9452 Posts
OCZ ZT 550W and Silverstone Strider Plus 500W (ST50F-P) would be cheaper but of course lower quality than the Seasonic X / Platinum units at 400W and 460W, as well as Silverstone Strider Gold 450W. I mean, the market for people who want full modular to sleeve or whatever else, isn't insignificant, but it's not that large. Most people are okay with a lot of the modular power supplies that keep say the ATX 24-pin and ATX12V / EPS12V connectors hardwired, or maybe a few others. | ||
|
invisible.terran
United States280 Posts
| ||
|
Craton
United States17233 Posts
| ||
|
Cyro
United Kingdom20275 Posts
On April 18 2013 09:01 invisible.terran wrote: Why do some streams like Demuslim's seem so smooth compared to my game. There's a noticeable in smoothness and since I can see it on streams, it is not my monitor's inability to display 120Hz. My ingame fps is also consistently above 60 in low settings. What else can make this difference? Thanks in advance Stream program will sample a frame every 16.7ms, while real world playing with realtime rendering can't have perfectly consistent frametimes without adding in input lag, that's basically what Vsync does, for an observer (but feels horrible for the player) | ||
|
Belial88
United States5217 Posts
On April 18 2013 01:54 xasuma wrote: Okay, my power supply is the following: http://www.newegg.com/Product/Product.aspx?Item=N82E16817371016 (Antec BP550 Plus 550W Continuous Power ATX12V V2.2 80 PLUS Certified Modular Active PFC Power Supply) I just ran furmark with the my current settings (i5 2550k @4.2 mhz and stock GPU values) : And my card got as hot as 86c.(auto fan speeds, that went up to 100%). and at about 1/3 into the testing my computer pretty much went to sleep for 3 seconds. and I got the following error. -"the Nvidia open gl driver lost connection with the display driver due to exceeding the windows time-out limit and is unable to continue, the application must close." --> this is what happens in sc2 also. Since I hadn't tested out my CPU oc with any stress test, I lowered it at @4.0 mhz, however the same thing happened. At about the same time into the test. ----------------------------------------------------------------------- The reason why I wasn't overclocking is because I am still not sure what is causing this, and since sc2 is giving me very good FPS at low/medium (always over 100). I thought I should probably find out about this problem before trying out values for the card. (Which I have tried to oc a little with msi afterburner, but I thought that was the issue and put the values back to stock) . The only thing that I think might be is the fact that I didnt move the voltage numbers at all when I overclocked the cpu ration using AsRock extreme tuning utility. Edit!: I just tested my CPU with Intel Burn Test @4.0 mhz. However during the test, the cpu ratio kept switching between x36, x37 and x40. (3.6 mhz, 3.7 mhz, and 4.0mhz) I did the stress level at standard and at high. But I am just wondering why isn't it locked at 4.0 mhz? :/ Antec BP is outdated, and not really great quality, but for your build it should be fine since you aren't doing insane overclocks and you aren't running a ton of peripherals. Nvidia driver crashing could be from a lot of things, but it's a sign of instability. I've gotten that with: - Faulty PSU. See if you can use a different PSU to test. Or, what you can do, is take another powersupply, attach it to the PCI-E connectors on the GPU, jump the motherboard connector (google it), turn it on, then turn on your computer. So like if you have some piece of crap, 10 year old computer, you can do this if you dont have a spare psu (who does?). What I would recommend is that you run prime95, just for like 5 minutes. If that passes, do an overclock of like 4.4ghz@1.3v (im assuming you have an aftermarket heatsink because you did a low overclock? dont ever overclock on the stock heatsink) and see if that can pass prime95 (it might not be stable for you, but we want to see if you can get that same error via power consumption). - Faulty GPU. Downclock your GPU. Use msi afterburner, just incrementally downclock it. If you can pass stress testing on furmark on downclocked settings, that is an indication of a faulty GPU. Your cpu bouncing around in ibt sounds like you dont have the settings correct, to use all 4 of your cores, and so turbo ratio is being applied. At stock settings, it's x40 on single core usage, x37 on 2 cores, x36 for 3-4 cores used. Dont use ibt, it's a terrible program (it's written by 'agentgod', that should tell you something). Dont use asrock extreme tuning utility. I dont know where you got the idea that you should use software to overclock, literally every single guide and person out there will tell you to never use software to overclock the cpu or ram (only for GPU is it okay). You should really post your FULL system specs, as in the exact model, of everything. Your ram, case, motherboard, the more information you give me, the more i can help you. You mention an asrock board, well, asrock boards are notoriously low quality on certain chipsets, I didn't even think about your motherboard but now that i know you have an asrock, it's very possible your motherboard is the issue. Don't hide things from me unless you just want to continue having your problem. Get what I mean? You need to make it easy for us to help you. | ||
|
xasuma
Chile62 Posts
First, thank you for your answer. I just tested my CPU with prime95 in all tests @4.0 mhz. Everything fine, temps got as high as 68 Celsius. Then I tried @4.4 mhz and everything was still fine, temps got as high as 72 Celsius. Now for my GPU test (which has a clock core speed of 950 mhz, and 2150 mhz memory clock as it's stock values) I tried taking the clock core speed down to 920 mhz, tested it, and I got the same error during testing . Then I took it down to 900 mhz, and it was fine with furmark. (I left the memory clock alone at 2150). ----------- My pc's specs are: i5 2550k (I have it with the coolermaster hyperplus 212) nvidia gtx 550 ti (stock fan) asrock p67pro3 (motherboard) 4gb ddr3 ram @1600mhz (this is the closest model i found online, just mine doesnt have the cool stickers.. Link Antec BP550 Plus 550W (power supply) 60 hz 22" monitor , 1080p. / normal old case/normal hard drive (i just ordered an ssd :D). So from my furmark test, you think there is a problem with my GPU then? SInce i Had to take it down 50 mhz from its stock speed for it to work properly? Merci! | ||
|
Jonoman92
United States9101 Posts
Are there any programs to try and run iOS apps on android? Probably a dumb question with a resounding no answer.. | ||
|
Belial88
United States5217 Posts
And you say gtx 550ti, i really wish you would tell us exactly what model it was... what manufacturer you have makes a difference. This was that difficult thing I was talking about, remember? Because certain manufacturers may, or may not, cover warranty on a 550ti nowadays. And what do you mean 'stock fan'. It's not really a 'stock fan'. It's called a 'reference model', which is what Nvidia (or radeon) sell, it's like the plain, original design. And then other companies, like MSI, gigabyte, sapphire, etc, they will often sell a reference model, and then they'll sell their souped up versions, which are basically with different heatsinks at stock, maybe a minor overclock on the bios (which they sell as SUPER OVERCLOCK EDITION!! or FTW!), maybe different amounts of VRAM if nvidia/amd allows for it, which nowadays they won't. So it's not really a 'stock fan', it would be like saying 'my stock heatsink on my cpu' if intel boxed every i5 with an nh-d14, and then msi, and gigabyte, and sapphire all sold their own i5's which was the exact same chip but maybe it was clocked to 4ghz instead of 3.9ghz, and came with a phanteks or silver arrow instead. | ||
|
Rannasha
Netherlands2398 Posts
On April 18 2013 15:42 Jonoman92 wrote: I'm new to smart phones. Got an app I'd like to try but it's only available for iOS while I have a S3 w/ android obv. Are there any programs to try and run iOS apps on android? Probably a dumb question with a resounding no answer.. Not possible. | ||
|
xasuma
Chile62 Posts
So by what you said I should take the core voltage up a little bit, and that should help? Here I have a screenshot of my msi afterburner, with the current values that are working for now. I unlocked the core voltage option just to see it's value, but I didn't touch it yet. + Show Spoiler + ![[image loading]](http://s13.postimg.org/sa1lwo0pj/afterburner.png) As for the "power limit %" I don't really know how to check that value. Thanks again! | ||
|
Belial88
United States5217 Posts
See the voltage limit you have in afterburner? That's a voltage limit that was hard coded into the firmware of the GPU. It's not 'really' the voltage limit of the card, your card can do wayyy more voltage safely, the TRUE max voltage is much higher. What GPU companies do, is hardware a much lower voltage limit into the bios, to prevent people from damaging their cards and then RMAing them. So, you got companies, like msi in fact, who say "we support overclocking and overvoltage completely in our warranties" because you can't really overvolt the GPU. Does that make sense? Yes, you are overvolting, but even the max you can overvolt your GPU without modding the BIOS, is not very much additional voltage. My point is don't worry about increasing voltage. Obviously, keep temps in line, but that basically means stay below 90-100*C. Yes, increase voltage a bit, see if that helps. Either way you should RMA your card, but I would suggest running a few more tests before RMAing, just to be 100% sure it's your GPU. Increasing voltage to see if that helps stability, try plugging the GPU into a different GPU slot on your motherboard and see if the issues are the same... i mean it's pretty cut and dry that your problem is the gpu, but do you understand what Im getting at here? Dont worry about the power limit %, it's greyed out because it's not an option you can adjust on your card. It's basically a meaningless option, just like CPU power and current limits in the bios. The GPU or CPU will draw what the CPU/GPU wants. If you have a power limit that is too low, it'll just throttle, and you don't want that. So you should generally always max such settings, but the default should always be so high that it's a non-issue anyways. Like I can set a power limit and current limit of 1000w on my cpu, but my cpu will never use more than 250w so it doesn't much matter. Now if I set my power limits to 150w, then my cpu would throttle. It only really matters to people who downclock their stuff, looking for power savings. People who might do compute workloads, run servers... not gamers. | ||
|
( bush
321 Posts
I am from Brazil and i am very sad with the internet here, i can barely play on NA with 200 ms ping (at least thats what "tracert us.logon.battle.net" tells me on cmd). Recently i heard VPN services could help me with this problem. Do you guys know if this really helps and could you recommend me any? Thanks. | ||
|
Shikyo
Finland33997 Posts
| ||
| ||
