Kallex

Added this friendly note - I try not to have toned response, but I'm confused Corsair vs. Corsair contradicting responses. This latest response so far stands out as seemingly false. I love the products, love the quality of support, now help me sort out which response is right and which one is wrong: Now that's a bit bold counter-statement late to the party, when there has been statements elsewhere (including this thread, by Corsair personnel) that pigtail is specced for 300W. So I find the pigtail capacity listed in the quote contradicting on pre-research I did and even the answers within this thread alone. I also personally am not aware of any RTX 3080 cards above 350W TDP coming with only 2x connectors, but I understand that is out-of-spec situation if those go to 380W. Our cards are 340W TDP cards (Asus TUF Gaming RTX 3080 OC - with default manufacturer limits). Other thread(s) mentioning the 300W limit (again by Corsair employee): https://forum.corsair.com/Forums/showthread.php?t=157109 Our use case isn't anymore 24/7, but nonetheless - I really would like some official clarification of the "misuse of product" part below. I just did as thorough search as I could from Asus TUF Gaming RTX 3080 OC edition. Nowhere I found the recommendations (even less requirements) for cabling. Only recommendation is 850W PSU. So, if you're effectively saying, that our (I consider high end) Corsair HX1000i or our RM750x are not covered with warranty, could you please pinpoint the exact documentation for that. We only use the cabling provided with the product, verified to the spec in this thread and the other one linked. I get the part, that for voltage stabilization and OC capacity. But if you're right and suddenly your pigtail must not be used with Corsair PSUs, there must be official statement of it. Or some actual, specific links to sources, where I can verify the claims. TLDR; I'm not arguing with you out of principle. So far I just have contradicting sources and none as frightening that "Frying up the PSU, not covered by warranty because of misuse".

Thanks for the heads up! We actually have crashing issues with Doom Eternal on one of the systems - the 750W powered one (other one hasn't played enough). I might try out the double cabling to see if it resolves (if the game gets played still).

Oh I didn't think you were debating :-). This whole thing is confusing, I've spent a lot of time trying to figure out the bits and pieces. We are running one cable (pigtailed) peaking to 320-340W power use for RTX 3080 during gaming, no stability or any issues so far. We don't do 24/7 for 300W right now, but not due to doubts for cabling, just lowering the power use down to 240W as the "last 30% of power results just 10% performance". We have run overnight 320W total load occasionally, with no issues either. Our systems have good airflow, in case it matters. Both systems are powered by Corsair; RM750x (Ryzen 7 2700X) and HX1000i (Ryzen 9 3950X). Not a single crash or game crash on either systems since 1-2 weeks of normal Folding@Home and gaming.

I think your question is proper, but you're mixing two things. 1. Wire thickness/pin thickness plays a role here. It can be also thicker up to before pigtail and then thinner for example. As mentioned here, Corsair uses thicker cable up before pigtail is split, so it can handle both the "split" parts power/current. 2. Twice the amount of wires will of course double the capacity That doesn't mean that the "undoubled capacity" can still be more than enough.

I'm not planning to SLI 3080s. I have two different Corsair-powered systems where they are going: - Corsair RM750x with single RTX 3080 - Corsair HX1000i with RTX 3080 + GTX 1070 I am doing Machine Learning research + gaming combined, want to get experience on multi-GPU setups for just that. The multi-GPU setup is my main workstation, the other one is for more regular use (and for different person).

Yeah apparently, I started to be "a bit too sure" on a matter, that I had no real expertise on. Based on common information available, its seems to be safe to connect those 2xPCIe on pigtail, but just suboptimal. Some vendors do recommend (as in here too) go with separate cables and only pigtail if needed (in 3x connector cards). I need to measure the wattage (once I finally get the two 3080s) of the cards and plan accordingly; mentally preparing to wire the additional cables if needed.

Thanks for correcting me. Actually I've continued Googling about the subject and apparently I have no fact basis on the ratings or the max currents. I'll edit my post, I can't verify the data - its all fuzzy and all. But I think its still technically safe use the provided cables and they can withstand the max power - just not necessarily optimally. EDIT: As you state 16g to first connector explains a lot of it.

EDIT: Lots of apparently unverifiable "nice looking numbers" - edited away. In short, my understanding is, that its safe to use pigtail; its designed to deliver both ends at the standard (150W max at both totaling 300W). Johnnyguru has better and shorter answer below. So eventually it boils down to your preferences. Electric-wise individual cables are better, for other reasons, they're not necessarily. In my use case, I have little interest to "overwire just to be sure", because of cable clutter and extra flat cabling potentially intervening with the airflow of already tight-spaced dual GPU setup. I will try first to cope with one cable to 2x PCIe 3080 GPU and reconsider if I bump into any issues.

I asked the same question when going for 2x GPU build with GTX 1070 and RTX 3080 - 2x8 pin version. PCIe standard allows 150W per each 8 pin connection and the cables that come with PSUs are rated for that; the pigtails too for both their connections being fully connected. That means single cable with 2x8 pins is rated for 150W + 150W = 300W itself. The benefit you get boils down to basic physics; you get less resistance and thus more stable voltages, if you use more cables to do the same job. So if you plan to OC and want to get and rely on as stable voltages as you need, you might/will benefit from more cables. You can safely start with lower amount and if you encounter problems, then add more, if you seem to need/benefit from them. My question is below. My use case is 24x7 use (we do Folding@Home during nights) so I wanted to check whether the cabling is OK for sustained high load. https://forum.corsair.com/v3/showthread.php?t=200109

I asked the same question a while ago (in a bit different flavor), and the answer I got was: "If the cables are included with the PSU, they are rated for that, but you might lose some voltage that affects the OC." I myself don't have especially thick cables, but have 2x GPU setup and overall don't prefer to overly "wire-clutter" that area of the case if not needed. https://forum.corsair.com/v3/showthread.php?t=200109 EDIT: What you REALLY DO NOT DO, is to get custom adapter to turn into sub-rated smaller cable (like 6 connector one) to output that 8 connector watts. There are images of melted adapters, which is likely outcome. So rely on your PSU's cabling, its a standard after all.

EDIT: I just realized I repeated what was already said in the thread before. Erased my post. It boiled down to that I think it should be verifiable that the warranty period clearly wasn't over before the case was raised (if its still valid today).

Oh this is a good point! I'll edit this to my main reply too. Might prevent my suggestion - as I don't use iCUE myself for this very reason.

DISCLAIMER! This might prevent my suggestion below - as I don't use iCUE myself for this very reason. DISCLAIMER! Accumulating data for longer term, such as full month (if you plan to) starts to be solid use case for time series database + dashboard solution. That being said, I'm author of HWiNFO PromDapter; Prometheus adapter for HWiNFO that lights up exactly that. https://www.hwinfo.com/forum/threads/prometheus-adapter-for-hwinfo-grafana-dashboard.6281/ NOTE! You will need to set up Prometheus and Grafana in some system, which can be an overkill for single-system setup. However if you have multiple systems to monitor, you only set them up once and just fetch data from all the systems (I'm myself monitoring 3 folding systems). Also Prometheus has default setting for 14 days, so you'll need to configure it properly. If you go this route, I'm happy to help you set it up. I just recently added my own HX1000i to it and its nice to see the efficiency and stuff realtime as well as for longer term. Here's example screenshot where PSU data is for the recent 7 days (my main system) and other parts are from 3 different systems CPU + GPU power usages. https://www.hwinfo.com/forum/threads/prometheus-adapter-for-hwinfo-grafana-dashboard.6281/post-27027

Those cards can still be very efficient while idle ;-). But - speaking of the current state of the world... you *could* run something like BOINC or other generic scientific grid-computer with just 10-20% of your CPU, enough to keep it "off idle". BOINC has enough options to keep it off when comp is not idle etc, limiting the CPU usage enough; possibly allow it to keep you off-idle, but not in too powerhungry mode (to introduce other unpleasanties). While it would boost your electric bill also a bit, it would benefit for greater good. My (personal) company made a commitment to run those things overnight, which adds definite electricity costs, but puts our hardware for good use.

Just a word of warning, be cautious of underestimating the PSU requirement of new RTX 30x0 (I think that's what you're referring to). It's not the continous load only that needs to be enough, but also the short term voltage spikes that trip the PSUs. If you Google around the topic, you'll see. It's not black magic, but its part of reason why the PSU requirement was that high. They always tend to overstate the requirements, but this time there is point to be made. Given also the issues with RTX 30x0 its easier to RMA your card, if you can say that you meet the PSU requirement and its not the reason for instability or whatnot. EDIT: I'm not saying 650W wouldn't do it, but if you can find out whether it has the characteristics (to tolerate the spikes) to match the RTX 30x0 requirements you're planning to meet.