My system is a prebuilt Lenovo Ideacentre Y700 with these specs:
Windows 10 pro 64 build 16299
Intel i5 6400
1x8GB @ 2400mhz
SSD 120gb
GTX 1070 Lenovo (seems it’s a MSI Founder’s Edition & 1506mhz base clock) stock bios, no updates available
Lenovo h170 motherboard, very very basic, it has very little settings available.
450w PSU 80 bronze
Symptoms
Random driver reset, crashes to desktop after a couple of hours, in Prey 1080p ultra, vSync (card gets disabled in device manager) - this actually doesn’t happen in busy areas where the card may work hard
Stutters in Unigine stress tests (Heaven, Valley)
The issues appear without overclocking
What I did:
I set it already to maximum performance in NVidia Control Panel.
I tried the following drivers (without DDU, just overwritten, using the “clean install” option):
391.01
388.71
381.65
368.39
I already updated the bios of the motherboard, reset bios settings to defaults.
Stress testing for 30+ minutes in Unigine Valley, didn’t crash the drivers, but it stutters a bit (see screenshot)
What other types of stress or gpu stress tests can I do to make sure the card works 100% ?
Does it make sense to stress test the ram or other components?
Is that 450w psu enough? Stress testing and even raising the power limit in MSI Afterburner doesn’t seem to crash anything.
An h170 doesn’t have heatsink on the VRM’s. Without at least passive cooling it will overheat under sustained loads, and this may cause stuttering with other components.
Check this out to learn more. I timestamped it at 5:35.
I have a push-pull configuration by two 120mm fans (front-back). The card is a blower style so it’s exhausting outside of the case. I’m not sure if this means that there is somewhat of an airflow directing heat away from the motherboard or at least cooling it a bit.
Guess I’ll try next a 2-3 hours stress test, that should confirm wether or not there is a temp issue somewhere.
Is the single channel memory also a thing that may cause the stutters?
I guess you’e not understanding. Do you have a rear case fan to draw air across the board in a passive style, or do you have a fan directly blowing air onto your motherboard (active style)?
Also, no. Single channel memory would not be the cause of this.
See the top of the motherboard where the CPU fan header is? Thats what should have a heatsink on it. So that area is getting passive air cooling by air being pulled across it by the rear fan and this might be fine for short burst of high intensity loads, but any long term load is bad news; especially if you’re over clocking.
Now this, may be the issue, but if it is due to your parts becoming overheated during sustained work loads, then you will want to slap some heat pads on there.
So we need to know if you are doing any overclocking, and what you are typially doing when the stuttering occurs. Like after a couple hours or so or only when you fire up a game?
No overclocking whatsoever, the drivers crash/reset after 2-3 hours of gaming, and stutters happen pretty much all the time, judging from the GPU load drops seen in the screenshot.
Only thing I remember doing is swapping the gtx 750 with this new card, but after that I reinstalled the drivers quite a bit. Gonna try a new fresh install of win10 or a win8/7 as I understand stutters get introduced to other gtx 1070 users with win10.
My other bet is that the low cost components this prebuilt system has, combined with the fact that it’s resealed item, resulted in a component having an actual defect that I’m not yet able to troubleshoot.
I got one more week until I can return this item but I badly want this GTX 1070, might as well sell the rest of the parts and build a new PC from scratch, but I want to rule out hardware faults on the GPU first.
That GPU is too hot. Pascal performance suffers dramatically past 62 degrees C. Now that 1070 has enough power, its drawing 75 W from the PCI slot and another 75 W from the 6pin; So its getting 150W total which is up to spec for the 1070. Although if its getting that hot under load then its efficiency is tanking. I would suggest undervolting/underclocking the card and try to get it under 70C.
I would tend to agree the card is running too hot. 82°C IIRC is the temp at which the card drops clock speed to keep temps down on the FE. I dont think that would cause the driver to lock up. I’ve never seen that happen to any 1070. I have my doubts about that power supply though. Its 450w but whats the 12v rail actually rated for? Cheap PSUs often bundle all of their rated wattages together to a max wattage they cannot actually deliver.
I would go with more case fans first. You have an airflow issue. You can replace the cooler on the card if you want, but its not necessarily going to lower GPU temps if theres a pocket of hot air in the case not going anywhere.
Noting is wrong with blower per se. But a better cooler would definitely help the card. YMMV, I think that card is custom.
Yeah its fine. The CPU draw is only 64 Watts, and the 1070 max draw is 150 Watts. So he’s sitting at around 225 Watts under full load. As to whether the PSU can handle the voltage ripple remains to be seen, we need to know about its efficiency rating.
A better static-pressure exhaust fan would help. Definitely needed if he switches out the blower cooling for once that churns the heat inside the case.
But if its 80+ bronze then you’re usually fine, with the caveat being that its only for efficiency. Where it can handle load-wise what it says is another story.
Totally fine. Its not churning out 450 all the time. It means that it can handle sustained loads, with the specific efficiency, up to that number (and somewhat beyond). So I don’t think its your PSU, unless it is actually bad.
Edit: for clarity. Thanks for the catch @MazeFrame
I think you are probably on the right track with the VRM not having any cooling. I think were both on the same page with “needs more cooling”. I only actually see 3 vrm in the machine linked, which doesnt sound right but if true could be a real issue without airflow.
@nemuro I cant actually read the info on the side of the PSU there. Its not a high end one for sure but it shouldnt be crap either. I might suggest making your rear exhaust fan an intake fan instead. This will make it louder but should help VRM temps. You might also consider a new case for your hardware. It looks like the board is standard ATX so it should fit in most cases. It would give you better cooling options. I can only see 3 spots for fans in that case so your options are more limited. Perhaps noctua fans are in your future. I know the arctic F12 and F14 are also great fans if you are looking for something a little cheaper.
