We have 2 slot 3090's now, while the 3070 FE is still yet to be released - and I'm still patiently waiting for the 3070 mini... *sigh*Yes, for the very modest price of 2k euro
There is, its called "Founders Edition"
-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
DAN A4-SFX v4.1
- Thread starter dondan
- Start date
I've got some potentially bad news. Stress testing a 3080 FE with my SilverStone SX700-PT PSU flipped. After 5-10 minutes of FurMark, my computer inevitably powers off, tries to turn on a few times, and then eventually boots. My working theory is that the PSU has some sort of fuse when the temps get too high. (This is in line with this model, as well as most others I've looked at, having a max operating temp of 40C.) Anyone else with an FE able to replicate?
Voltage seems fine and CPU/GPU temps are entirely reasonable, so I don't know what else it could be. Currently trying the last thing I can think of, which is connecting my GPU to two separate PCIE cables from the PSU instead of just one that splits. (Not sure how this could be it, though, given that the card runs fine at 320W for several minutes before the computer powers down, and that no voltage drop is observed in MSI's monitoring widget.)
EDIT: Although, so far so good 10-15 minutes in, so maybe it was the split cable after all...
EDIT2: Nope, still powered off, but took about 20-25 minutes this time. (3DMark’s Time Spy Extreme stress test will hit this more reliably.) Will try a different wall outlet first, and will try re-flipping the PSU after that.
EDIT3: Switched to wall outlet. Time Spy Extreme stress test failed after about 7 minutes. As an aside, the behavior of the computer is a bit erratic for a few minutes after the initial shut down. The PC will try to start again several times and fail immediately. When it boots, it will likely fail again a few minutes in. After that, it's usually OK.
Voltage seems fine and CPU/GPU temps are entirely reasonable, so I don't know what else it could be. Currently trying the last thing I can think of, which is connecting my GPU to two separate PCIE cables from the PSU instead of just one that splits. (Not sure how this could be it, though, given that the card runs fine at 320W for several minutes before the computer powers down, and that no voltage drop is observed in MSI's monitoring widget.)
EDIT: Although, so far so good 10-15 minutes in, so maybe it was the split cable after all...
EDIT2: Nope, still powered off, but took about 20-25 minutes this time. (3DMark’s Time Spy Extreme stress test will hit this more reliably.) Will try a different wall outlet first, and will try re-flipping the PSU after that.
EDIT3: Switched to wall outlet. Time Spy Extreme stress test failed after about 7 minutes. As an aside, the behavior of the computer is a bit erratic for a few minutes after the initial shut down. The PC will try to start again several times and fail immediately. When it boots, it will likely fail again a few minutes in. After that, it's usually OK.
Last edited:
Haha, right. Show me where to find one of those.
Unlike Nvidia, Gigabyte might actually want to sell me one of their cards.
Please keep us up to date. Can you make a picture? Did you install the latest Nvidia driver for the GPU high clock crash problem?
Maybe the operating temp of 40°C of the Silverstone is a problem. This is why I ordered a Fractal Design 650W that has a 50°C operating temp.
Another solution could be adding a slim 92mm fan above the PSU inside the case that pull more air outside the PSU.
Drivers are 456.71, so high clock crash should be fixed. What do you want a picture of?
If the stress test succeeds with the PSU intake fan facing outwards, I might just decide to keep it that way unless the temps are terrible -- maybe with the GPU undervolted a bit. After that, perhaps I'll trade the FE for a 3080 blower or Big Navi equivalent when those come out.
(Case is great, by the way!)
If the stress test succeeds with the PSU intake fan facing outwards, I might just decide to keep it that way unless the temps are terrible -- maybe with the GPU undervolted a bit. After that, perhaps I'll trade the FE for a 3080 blower or Big Navi equivalent when those come out.
(Case is great, by the way!)
PSU entirely out of the case was totally fine. Same with a subsequent test with the exhaust facing out.
With the PSU exhaust facing out, a 20m session of Time Spy Extreme resulted in a GPU temp of ~84C and a clock of 1710MHz. Quite noisy, though. I think PSU exhaust facing inwards resulted in a ~5-10C improvement, so it's disappointing that this configuration is apparently not possible with the PSU I have.
With the PSU exhaust facing out, a 20m session of Time Spy Extreme resulted in a GPU temp of ~84C and a clock of 1710MHz. Quite noisy, though. I think PSU exhaust facing inwards resulted in a ~5-10C improvement, so it's disappointing that this configuration is apparently not possible with the PSU I have.
The Fractal PSU @dondan mentioned has the following spec: "Maximum 80% load at 50°C"
I don't know much about electrical engineering. Does this mean that as the PSU temperature increases, the max wattage decreases? If so, I wonder if this issue could be resolved with a higher-watt PSU, e.g. SilverStone's new 800W SFX-L model.
I don't know much about electrical engineering. Does this mean that as the PSU temperature increases, the max wattage decreases? If so, I wonder if this issue could be resolved with a higher-watt PSU, e.g. SilverStone's new 800W SFX-L model.
Last edited:
So the more efficient a PSU is, the less watts are lost to heat and so on. PSUs also have over tempersture protection.
When it is too hot, it shuts down. That could be heat generated by itself or from other components.
A PSU that is more efficient AND can operate at higher temperatures will not shut down as early
Discusses performance of a flipped PSU re: the 3090. Different case, potentially relevant learnings:
Last edited:
More like Turboreactor edition ?
The Reddit thread has a few people with the same problem as me.
Sounds like maybe the Corsair won't shut down, but I'm not sure that's such a good thing, given that the side panel was measured at 52C.
Based on the OT video, sounds like undervolting and biasing the front fan is a much better solution in the end.
Or: buy a card that doesn't employ a 'flow-through' design.
The new Radeon's design is intriguing. A solid IO plate so no rear exhaust, but exposed fins on top – looks like it mostly exhausts upward. A two-slot version could be a great fit in the A4.
OK, ran a 40 minute Unigine Heaven 4.0 Extreme test for 40m, with PSU intake facing out. Fans are at 100%, GPU is stable at 83C, and clocks average out to 1850MHz. Room temp is around 80-82F. Will try undervolting next.
