Yesterday, someone asked if it was worth getting a more expensive 50XX, and I think this community did them wrong

Posted by the_lamou@reddit | buildapc | View on Reddit | 41 comments

Turns out, the budget brands [massively skimp on cooling](https://www.tomshardware.com/pc-components/gpus/igors-lab-uncovers-hotspot-issue-affecting-all-rtx-50-series-gpus-says-it-could-compromise-graphics-card-longevity), creating hot spots around the power regulation circuitry and potentially significantly decreasing the lifespan of the cards. Not a problem for someone who upgrades every year; big problem for budget card buyers that can't afford not to hang on to their hardware for a while. There's way too much of a "premium/expensive == scam" mindset here (and Reddit in general). The reality is and has always been that it's way more nuanced than that and a lot of times expensive == better, even if it's not exactly "1:1 better benchmarks." So I guess this is just a discussion / PSA / mini-complaint: saying things like "you will never need better cooling/build quality/reputation, just buy the cheapest whatever you can find" is bad advice. Sometomes, it's true. A lot of times, it isn't. And most of those latter times, buying the cheapest is just going to cost you a hell of a lot more.

Reply to Post

41 Comments

[-]

Dik-in-fish@reddit

Am I missing something? Cuz that article says all 50 series cards not just budget ones

[-]

the_lamou@reddit (OP)

>Am I missing something? Yes, you are. They very clearly explain that the testing was performed on Palit and PNY cards (two bottom-tier budget brands) and how better cooling completely solves the problem. The only conclusion they come to for *all* cards is that NVIDIA'S cooling specs are for normal use rather than worst case scenario, but none of the premium AIBs rely on the base spec document anyway, making it not an issue. This is why you read to the bottom of the article instead of stopping after the subhead.

[-]

DarthXDaddy@reddit

Is okay this reddit. Reddit don't read

[-]

Dik-in-fish@reddit

Your just blatantly lying, the article pertains to all 50 series cards and they admit the reason is that they packed and layered the PCB way to much, the article you link is quite literally about how 50 series cards are fundamentally flawed, so I guess congrats on proving yet another reason 50 series is trash.

[-]

the_lamou@reddit (OP)

The article says that many AIB card builders potentially have similar issues because they may be trying to squeeze too many components into too small a space, points out that using better materials (as are common in server and workstation cards) can offset the problem, and then spend the entire back half of the article talking about how just putting thermal pads on the VRM basically fixes the problem entirely and cards that have thermal pads/paste on the power circuit are unlikely to have the issue. I get that reading more than a single paragraph is hard, but the article doesn't end after the intro.

[-]

9okm@reddit

Yes. This post is pointless.

[-]

Dik-in-fish@reddit

Yep OP is literally just here simping for those high end rtx cards

[-]

itsforathing@reddit

That is a hot take. But the article you posted seems to blame the pcb layout more than the cooler itself. Most manufacturers will use the same pcb and slap a larger cooler on it. Which may help but if the issue is location of power traces and VRM then it should affect, for example, all ASUS cards and no XFX cards (just throwing names out at random).

[-]

curiosity6648@reddit

Yeah the basics of the issue is the PCB design on the 5070 is smaller so you need more thermal contact points to cool all that hot stuff in a tight area because it's getting hot from the stuff around it. It's actually not a "cheap cooler" issue at all.

[-]

itsforathing@reddit

The 12vhpwr cable doesn’t help either. That may be a leading factor in overheating as it seems 1-3 pins deliver the most power instead of balancing them across 2-3 8pin pcie cable pins. Having such a small power header would mean cramped power delivery at least on part of the pcb. I have a premium Asus 3080 and recently replace the paste and pads. Asus really cheaped out on the quality, less than 2 years from manufacture and less than 2 months of use and the paste and pads were dry. I got over 10 degree reduction in core temps and nearly 20degree reduction in memory temps. Plus the hotspot went from a 20 degree differential from the core to 10 degree differential. I only replaced what was there with thermalright odyssey pads, I didn’t add any more than what was already there.

[-]

the_lamou@reddit (OP)

>That may be a leading factor in overheating as it seems 1-3 pins deliver the most power instead of balancing them across 2-3 8pin pcie cable pins. No. Just no. This was found *once*, in a very questionable test. This is neither common nor established. >Asus really cheaped out on the quality, less than 2 years from manufacture and less than 2 months of use and the paste and pads were dry. You know that paste/pads dry out faster if they're not regularly heat-cycled, right?

[-]

itsforathing@reddit

My 9 year old power color r9 390 and gigabyte Rtx 2070 had better thermal pads than the modern Asus card. And the r9 390 stayed dormant longer. I’m not just speculating on the quality of Asus thermal pads, I have first hand experience with it. And a questionable test? You can’t fake the amperage. It was a card in use and the amperage was measured per wire.

[-]

the_lamou@reddit (OP)

>And a questionable test? You can’t fake the amperage. It was a card in use and the amperage was measured per wire. No, but a single result on a single card using an old known-questionable cable with a known-questionable power supply tells you absolutely nothing. A YouTuber with Der8auer's reach and alleged engineering background should have known that and tried to get multiple cables, power supplies, and cards. Which is why Falcon Northwest dunked on him so hard (despite it being largely ignored by the community because who cares about good methodology when we're all busy being outraged in a herd).

[-]

Veiny_Transistits@reddit

IIRC there was recent suspicion that accidentally bent pins on the connector were driving that Although it was a handful of anecdotes that may not hold up compared to other more thorough testing

[-]

itsforathing@reddit

I saw a video where they tested each individual wire and 1 was pulling way more amps than the gauge of wire was rated for. But there could be multiple causes for this problem. And while the few times it has happened got a lot of publicity, i don’t think it’s all that common of a problem.

[-]

the_lamou@reddit (OP)

>I’ve seen high end cards skimp on thermal pads so I’m not sure this is a msrp model vs premium model issue like you are making it out to be. I've seen it happen very rarely, but it's not a standard design feature like it is on cost-cutting brands. Like I said in the initial post, higher cost isn't a *guarantee* of quality, but *on average* a higher cost/premium card will have more thermal contacts, better fans (even with fewer cooling pads, just blowing more air over the PCB helps), better/more sensors (a big deal for proper thermal control), and a better heatsink which will pull more heat off the card and make radiant and convective cooling more likely even without thermal pads. As an example, I have a GPU VRM temp sensor on my card (which I average in with other GPU sensors in my FanControl settings). I have never seen it over 70°C, and I'm on a Gigabyte Aorous 5080 Gaming OC, a budget-to-mid-level card, just from a more reputable company than Palit and PNY.

[-]

curiosity6648@reddit

Respectfully, this falls under the "sometimes shit happens" of GPU coolers. In general, the advice of buying the cheapest model still stands. Now that PNY has been alerted to this issue, it's a relatively simple fix and I suspect you will actually see them do so in future revisions. You should never pay extra for the better coolers. You should occasionally (as seen here) avoid certain coolers.

[-]

curiosity6648@reddit

And here it is! I guess my memory was pretty good! https://youtu.be/H7lnBCFnBok?si=nwAn8TlJ--w6G0DZ As you can see, sometimes cooler designs are just duds. Not a cost related thing.

[-]

the_lamou@reddit (OP)

But this report isn't just "here's one model with an unfortunate design." It's two, both from budget brands. As a point of comparison, my Gigabyte has a VRM temp sensor and I never see it get anywhere near that high. I think I might have an IR camera somewhere, so maybe I'll triple-check, but mine seems to cool well.

[-]

curiosity6648@reddit

Gigabyte? You mean the worst GPU maker? Ok so you really got no clue....

[-]

the_lamou@reddit (OP)

It was the first 5080 I was able to grab and will be replaced as soon as I bother restarting my hunt for a 5090, but even with that it's so far from the worst GPU manufacturer. I mean, PNY exists. So does Palit. And Asus, while their stuff is typically high quality, massively screwed the quality:performance ratio. But also, I'm not sure why I'm even bothering to respond to someone that actually purchased a 5060ti.

[-]

VruKatai@reddit

The best advice is to not buy a 5xxx card, period. Between this article covering every single tier and ever manufacturer, articles about leaking thermal gel, terrible drivers, getting less generational uplift for more money and the missing ROPs, this was a *terrible* launch. These are going to be one series of cards to *definitely* not buy in the used market. All this on top of the not great 4xxx launch that barely saw uplift from the 3xxx series. I *almost* pulled the trigger on a 5070 ti but just couldn't get past the ridiculous pricing. These last two gens make me feel like my EVGA 3080 12gb Hybrid at msrp was one of the best purchases in my 30+ years of building. If people want solid advice, either grab an AMD card or grab a 3xxx series card used for really good pricing.

[-]

the_lamou@reddit (OP)

>These are going to be one series of cards to *definitely* not buy in the used market. This is going to be the pattern going forward, until there's a massive breakthrough in lithography or superconducting materials. We've basically hit the limit of what we can do on a mass market scale with existing technology, so the days of 50% gen on gen increases are essentially over. AMD is still managing it because they were significantly behind NVIDIA to begin with and only focused on the bottom tier this gen (basically just repackaging last-gens high-end cards into cheaper price tiers plus starting to catch up on ray tracing).

[-]

Infamous_Campaign687@reddit

The 4090 and 4080 and solid uplift over their 30-series counterparts. The 4080 had about 50% which has been fairly typical for the 80-series. It even increased VRAM by 60%. For the 50-series only the 5090 has been a decent upgrade and that only by brute forcing. Unimpressive.

[-]

DirtyDag@reddit

>most if not all RTX 50-series AIB partner cards are allegedly prone to high-temperature hotspots in the power delivery area OP, it doesn’t sound like it’s exclusive to budget brand.

[-]

the_lamou@reddit (OP)

"Prone-to" in this case is "they all use similar PCBs so it's possible they all have this weakness," but since the problem goes away with cooling on the power circuits, it's a hypothetical. As I've mentioned elsewhere, I have temp monitoring on my GPU power distribution circuits and while it's warmer than other parts, it's nowhere near what they found in the article despite being an OCed 5080.

[-]

itsforathing@reddit

Or budget models within a brand.

[-]

DrKrFfXx@reddit

Higher end cooling solutions are certainly worth something. But they tend to be heavily overpriced. But there is some contradiction in the community that's for sure, they would put mid end CPU's under costly 360 radiatators just for looks, but scream in disbilief if a card with proper cooling goes for 200€ more than the "MSRP" models, citing "they'll perform the same" as the reason of the outrage.

[-]

curiosity6648@reddit

Ok so the difference is: A 360mm AIO is not for looks always. An Intel core I9 14900k performs noticably better on a 360mm AIO than on any air cooler and can be overclocked farther. Now on a 9800x3d, yeah I recommend 240mm or an air cooler, you don't need more unless it's for looks. Now with gpus, I'm not gonna recommend spending $200 for a better cooler. That's stupid. Get a better GPU model with the money (or ram, CPU, etc). Spend it somewhere with a bigger performance benefit. And really, in the modern day even on high end gpus like the 5090 it's less enticing to spend on coolers. Back in 2016, if you wanted to get a 1080ti with a hybrid cooler or exotic cooling, I can for that. There were overclocking gains to be had, you could flash an XOC bios and crank that puppy up from its stock 250w to say 350w. There was also the fact blower cards were hot, loud, and barely adequate. Today? The cheapest 5090 can easily maintain 600w of power. The most expensive 5090 can only draw 600w of power. There is a noise difference maybe, but that's it. You're no longer seeing performance gains like was once doable on the high end cards.

[-]

the_lamou@reddit (OP)

>You're no longer seeing performance gains like was once doable on the high end cards. You are on 5080s, which may not be the *highest*-end but is still a high-end card. Pushing clock and men speeds definitely creates a noticeable heat burden and will absolutely throttle with poor cooling. It's the difference between being able to maintain +400/+2,000 without needing to seriously undervolt or having to undervolt.

[-]

DrKrFfXx@reddit

>There were overclocking gains to be had, you could flash an XOC bios and crank that puppy up from its stock 250w to say 350w. There was also the fact blower cards were hot, loud, and barely adequate. You're describing exactly the low end 5080s (360w) vs higher end 5080s (450w) > An Intel core I9 14900k That's why I explicitly said mid end chips.

[-]

curiosity6648@reddit

On a 5080, you should spend more and get a 4090/5090. Now, I admit in the current market that's less than feasible, but it's one exception. It's also the case now that if you got a 5080 Tuf or other low end model it's not using a blower fan. It's actually a pretty adequate cooler. Look at the 3080 and 3090 cards that were dual slot we were pushing 350-450w though, even the lowest end 5080 coolers are "overkill".

[-]

DrKrFfXx@reddit

>On a 5080, you should spend more and get a 4090/5090. Even in a world where MSRP existed, you would be spending "1200" for a good 5080, and 2000+ for a basic 5090. It's just not a matter of "make the effor", they are worlds apart still. Real world pricing is 1400 vs 3000+ I'm not seeing the Shadow, Ventus, Palit, Inno 3X, PNY, and a few more models as being overkill really. Tuf isn't really lowend either. It's mid to high end compared to other models.

[-]

dzone25@reddit

I'm always someone who says go with the brand with the best customer service & do your own research RE the brands that people say they should buy. I think that's what this Reddit usually does?

[-]

the_lamou@reddit (OP)

Used to be. Lately it's mostly been "any price higher than what I think I should pay is a scam!"

[-]

SliceOfBliss@reddit

Tbh, me and a friend used to briefly work at a repair shop, and the most common brands/models to get in were Ventus, Dual and Challenger (we got 5 Ventus on a weekend)...you get what you pay for, but at the same time, nowadays since most stuff is expensive, getting a MSRP GPU without asking is a "deal". Also, it enters into equation how much is the difference in price from a basic model to a superior/premium, Nitro+ is always the most expensive, same as ROG Strix, and if they cost like +100 dollars more, then it might not be worth it.

[-]

SuperPork1@reddit

I wonder how many of those broken models are due to those models being some of the most popular choices

[-]

SliceOfBliss@reddit

The most popular models from NVIDIA in my country are Asus TUF and Gigabyte Windforce/Aero, as for AMD, it's kinda hard to find but Gigabyte and XFX are present, and few times AsRock.

[-]

curiosity6648@reddit

To be fair, cheaper cards are more likely to be in prebuilts which as we know the prebuilt and enthusiast market is different and may also have something to do with the failure rates...

[-]

chrisz2012@reddit

Sometimes you can get better budget cards or bad high-end cards. Gamers Nexus showed in the past some video card makers had insufficient cooling of VRMs and specifically the mosfets. I ran an RX 6600 Eagle from Gigabyte for 3-years worked just fine. Eagle cards are supposed to be Gigabyte's more entry-level cards. My RX 7800 XT from XFX is a 210 SWFT entry-level card. Unless someone is overclocking the beefier coolers probably don't make a difference. Also higher-end cards can be factory overclocked and will take advantage of a better cooler because they're designed with running faster than a stock card. For the average gamer I think just buying a well-reviewed card from the masses is best. If you see something with 5,000 reviews 4.5 Stars and read the reviews that the card is stable. It will likely be that way for a long time to come. Also reading professional reviews if they exist for your exact make and model of card helps too.

[-]

Background-Rise-8668@reddit

What are you smoking dude? Stop consuming rage bait clickbait YouTubers. They ALL have nvidia gpus, and your outrage helped pay it for them.