Heck, we were thinking the same thing not too long ago. So, in this article we’re going to dive into what you need to know, and whether you should suddenly give a hoot, or two, about electrical engineering.
The scope and scale of which has not yet been determined nor confirmed by any official party.
Over the past weekend, however, a handful of theories have been put forth to explain why some users are taking to forums to report system crashes with brand new RTX 3080 graphics cards. Most of these theories boil it down to a single issue: capacitors.
But before we hammer the final nail into the coffin, let’s dive into what it all means, and why it might be best to wait for official statements and explanations before jumping to any conclusions.
It’s not entirely unlike a new GPU launch to also entail some instability issues, most of which are fixed in following months through updated drivers or firmware fixes. However, an article from Igor’s Lab (or two, for that matter) posits an in-depth explanation, one which suggests that the choice of capacitor on the underside of the GPU may have some impact on subsequent crashes, yet crucially also suggests that a lack of GPU testing across a wide variety of voltages and frequencies is key to the whole affair (read: it’s not quite as simple as some capacitors good, some capacitors bad).
The two types of capacitors most commonly found on the underside of your graphics card’s chip are MLCC and SP-CAP. These are easily distinguished from one another by their arrangement: either a single large SP-CAP, or a group of, likely 10, MLCC capacitors.
You’ll also hear the terminology POS-CAP thrown around too. This appears to be an incorrect attribution, stemming from Panasonic Polymer Tantalum Solid Capacitors, which are not often found in modern graphics cards according to Buildzoid on YouTube. Instead, the more widely manufactured SP-CAP is found in its place.
There’s a bit of disagreement between those in the know on the term POS-CAP’s use, as POS-CAP may be common despite the technical inaccuracy of its use. That, however, doesn’t really factor in all that much to the issue today.
All you really need to know is that the capacitors on the different cards varies between make and model of graphics card, and that their stable operation can depend on many factors. There are heaps of variables for consideration when trying to explain crashes, spanning across power delivery to a GPU, and we’re far from a definite answer in this case. There’s also the fact that modern graphics silicon is able to essentially auto-overclock at will thanks to GPU Boost, whether you’re running it at stock speeds or not.
A few graphics cards manufacturers have released statements on the issue in the past few days.
EVGA reports that a 6-capacitor configuration, of SP-CAPs, was insufficient for “real world applications testing”, and that a greater number of MLCC capacitors would fix the issue (reported by Videocardz). It also states this as the reason for the delay to the EVGA FTW3 series models.
What is GPU Boost?
(Image credit: Nvidia)
GPU Boost is Nvidia’s dynamic clocking technology. Essentially, it is an algorithm for maintaining a max clock speed for the longest period, based on certain pre-determined limits, such as power draw, acoustics, and temperatures. Nvidia’s Boost technology will even overclock a compatible card beyond its rated max boost clock, effectively applying a temporary overclock. Any manual overclock subsequently applied to an Nvidia GPU then acts as offset to Boost, stepping the frequency curve by a given OC MHz value every step of the way.
Galax, Gainward, and Inno3D have all similarly confirmed that their graphics cards would not be sold with 6 SP-CAP capacitor configurations. Galax also explained that some of those sent to media beforehand may feature the 6-capacitor configuration, but that these were only production samples. Not one of these three manufacturers confirmed any crashing or fault due to the 6-capacitor configuration—Gainward insists it has not received any such feedback, in fact.
Zotac is also investigating the issue, and while MSI initially said a driver issue was at fault, it has also updated its RTX 3080 Gaming X Trio design to feature more MLCC capacitors, Videocardz later notes.
Generally, then, it seems as though MLCC capacitors can be more stable than SP-CAPS, and that those with SP-CAPs may see instability with some of the least performant GPUs. This is only a working theory at this time, and also doesn’t necessarily mean your SP-CAP-only card will crash, nor that a MLCC capacitor card is guaranteed to remain stable.
Perhaps that’s not the answer you were looking for right now, but that’s kind of the point. The truth is we actually know very little about the scale, exact cause, or permanent fix for this issue, only that indeed some RTX 3080s are crashing when pushed into high frequencies, at certain resolutions, in certain games, and that this may have some correlation to capacitors. All other theorems remain just that.
For our part, we’ve experienced issues testing third-party versions of the RTX 3080, specifically a factory-overclocked card. Interestingly, the driver crash has only reared its ugly head when we’ve been running 1080p benchmarking, not at 1440p or 4K, and only in specific games too. We’ve tested with new pre-release firmware, which disabled the factory overclock, but we still encountered the same issue.
Underclocking the GPU itself, however, fixed the stability issue, and we only needed a slight offset of -30 percent to get there. Which does speak to potential stability issues in the power subsystem of the card.
It is also worth noting that while the Founders Edition worked flawlessly throughout our benchmarking, we were unable to stably benchmark even that card with an overclocked GPU—there’s not much OC headroom to play with, it seems.
And it’s not just us either, as once again anecdotal reports suggest other RTX 3080 users have also benefited from a marginal downclock.
There’s still the very real possibility of a solution via firmware, as opposed to a complete RMA. GPU manufacturers charge a great deal of cash for the best parts for custom overclocks, such as first choice of capacitors, and perhaps we’re simply seeing the result of a judicious boost frequency curve coming into play on less-than-suitable (read: cheaper) cards. The capacitors may be causing some instability, even at stock, but that insufficiency may only be brought to light by another, fixable variable.
Nvidia’s latest driver, launched today, also promises some stability fixes for RTX 30-series GPUs.
If you’re one of the unfortunate few, that we know of, who is affected by intermittent crashes, then we suggest reaching out to the card’s manufacturer for further information, you could always send your card back for a new one, although we’re still very much in the ‘wait and see’ stage right now and that may not deliver a firm fix.
We have reached out to Nvidia for comment and will update if we hear anything back.