Some Gigabyte GeForce RTX 30 and 40 graphics card PCBs are cracking

[nanoguy]

A hot potato: It's no secret that graphics cards have gotten larger and heavier in recent years. The problem of sagging cards is starting to bite gamers that purchase high-end custom models from companies like Gigabyte. Even though it's possible to prevent most issues using an anti-sag bracket, this shouldn't fall on the user, especially when the graphics card in question has a design flaw that makes it prone to breaking without an aftermarket accessory.

A sagging graphics card can be annoying, but it's usually an easy fix. However, a poorly-designed graphics card can also exhibit PCB cracking due to the increasingly large cooling solutions mounted on modern GPUs. As reported by Louis Rossmann, who is a known repair professional and right-to-repair advocate, several Gigabyte graphics cards have suffered physical damage in recent months, and the company doesn't appear to be interested in helping the affected customers.

Specifically, several models of RTX 30 and 40 series graphics cards from Gigabyte seem to be affected by the issue. A fellow repair professional told Rossmann they'd worked on at least a dozen cards and provided several photos to illustrate the problem. Also worth noting is that eBay seems to be littered with listings for Gigabyte cards that were refused during the RMA process (or are simply sold for parts as they are out of warranty) because they have a cracked PCB.

As you can see from the image above, the cracks aren't easily noticeable, so the owners used arrow stickers to point them out. It is very concerning that most of them seem to have developed close to one end of the PCIe connector and right next to the PCIe locking tab.

A few copper traces deliver power to certain components on the board and are routed through the area in question, so a crack will break those connections and result in a non-working graphics card. It probably doesn't help that Gigabyte has made the PCB cutout near the locking tab quite large.

The repairs involve carefully scraping off PCB layers, among other things

The repairs are quite difficult and have a typical success rate of 10 to 20 percent, but thanks to Jérémie (the independent repair professional that contacted Rossmann about the issue) there is now a guide on repair.wiki detailing the process for anyone who works as an independent repair technician or is interested in DIY fixing their broken Gigabyte card. Another problem is the cards in question will never be safe to install horizontally, but at least it's possible to get some of them working again.

You can probably prevent any cracks from occurring on your graphics card by using an anti-sag bracket, which many manufacturers (including Gigabyte) already include with some RTX 40 series cards. However, it's sad to see Gigabyte not owning up to what appears to be a design defect that can easily render some of its cards unusable for their owners. We're hoping the company will eventually decide to set up a return and exchange program as it did with the exploding PSUs of yesteryear, but we'll have to wait and see.

Permalink to story.

https://www.techspot.com/news/99030-gigabyte-rtx-30-40-series-pcbs-cracking.html

 

[takaozo]

The quality of video cards from this company ended with Gtx 1000 series.
I still have the Aorus 1060 in good condition. At Rtx 2000 cards the gaming series had underpowered coolers and serious issues with the plastic backcovers. The Rtx 2060 gaming OC I got now for a spare was so badly designed. An cooler that handled max 125w with an ock GPU drawing 165w and constant temp limit in GPU-z. 83 celsius with original cooling. After removing the backplate and swapping the 92mm fans and shroud with a custom one temps max 72-75. This new issue looks like a cheap PCB and not enough hardening compared to the card weight. There are a lot of used 3000 series here also. I was in the hunt for a newer GPU and chose Asus. The cooling it's fine but the SAP II coils very loud for me. But in life you cant have them all at once. I ramp up the fans 100 rpm more and not hearing the whine.

 

[Achaios]

I have owned three (3) Gigabyte GPU's that failed or had manufacturing defects and Gigabyte refused to honor the warranty.

1. Gigabyte NVIDIA GTS 8800
2. GTX 580 Super Overclock (Premium top of the line)
3. GTX 780Ti Gigahertz Edition (Premium)

I stopped buying Gigabyte after the third defective GPU. No other GPU I have ever owned from other AIB's has ever failed on me or had any defects.

There's a reason GIGABYTE is cheaper than everybody else & you "get what you pay for".

 

[p51d007]

Ahhh the good ole days when I could solder on PCB's without my tri-focals!
That and I still have a large stash of lead-tin solder which I think is better than
the stuff today.

 

[Adhmuz]

It seems like a bad design choice to have such a large gap where the locking tab is, as this tab essentially acts as a fulcrum once the card is installed in a case horizontally exacerbating the issue are the bigger and heavier cards.

At first I'd have thought this design is to counter the over heat-sinked motherboards with M.2 slots encroaching on the PCIE slots in combination with ever bigger GPU coolers, I've had this exact problem in the past where getting the GPU to seat fully was nearly impossible on a first gen Asus Threadripper motherboard and an EVGA GTX 1080ti which wasn't even that outlandish a design.

 

[yRaz]

Really a shame, this seems like an easily avoidable engineering problem. I guess this is also an unforseen consequence of the massive coolers needed for the absurd power consumption numbers.

I knownita Gigabytes engineering team that's mostly at fault for this, but they're working around nVidia's crazy power requirements

 

[texasrattler]

Its a design flaw and can be proved. Someone should just sue them. Should be a fairly easy case.
Sometimes it's the only way to get a company to do what's right otherwise they will continue doing business as usual.

 

[neeyik]

I knownita Gigabytes engineering team that's mostly at fault for this, but they're working around nVidia's crazy power requirements

If the power draw was such an issue, other vendors would be struggling with the same problem and so far, there don't seem to be many reports (if any at all) coming from owners of Asus, MSI, ASRock, PNY, et al cards. This is entirely Gigabyte's fault -- be it a design or manufacturing flaw.

 

[MrBlkfx1]

I have a Gigabyte 3080Ti and 4090, and I have been fortunate not to run into any issues. It sucks that this is an issue in the first place, and on top of that, they aren't honoring RMAs. I will think twice about buying a Gigabyte card in the future. The last thing I want to worry about is my $1800 GPU failing because of a PCB crack.

 

[MaestroIT]

I was unsure which 4080 to get (Gigabyte as my 2070 super I had or something else), at the end I went with PNY which runs cooler and is cheaper, and it is great...

 

[godrilla]

I have owned three (3) Gigabyte GPU's that failed or had manufacturing defects and Gigabyte refused to honor the warranty.

1. Gigabyte NVIDIA GTS 8800
2. GTX 580 Super Overclock (Premium top of the line)
3. GTX 780Ti Gigahertz Edition (Premium)

I stopped buying Gigabyte after the third defective GPU. No other GPU I have ever owned from other AIB's has ever failed on me or had any defects.

There's a reason GIGABYTE is cheaper than everybody else & you "get what you pay for".

The only Gigabyte hardware I owned was the 1070 extreme gamer. Compared to Zotac, MSI, ASUS, EVGA I fealt the quality was down there with XFX. imo.

 

[takaozo]

Ahhh the good ole days when I could solder on PCB's without my tri-focals!
That and I still have a large stash of lead-tin solder which I think is better than
the stuff today.

I still have a lot of lead-tin alloy (rosin core lead free also) from my youth days, from 0.35 to 2mm.
Same issue here years had left a mark on my vision, glasses only for near distance for now.
I'm in the middle of tuning my Audigy RX card right now, gathering parts. Want to change the op-amp's and put a separate power supply for the 3.3v and 12v line on pcie slot.

 

[toooooot]

I have gigabyte card. But at least it is mountain in such way that its rear is attached to the top of the case. I should be safe there.

 

[VaRmeNsI]

So if their PSU doesn't blow up your system, the backdoor motherboard spyware doesn't get your system hacked, your graphics card might crack...

A new low for gigabyte...

 

[yRaz]

If the power draw was such an issue, other vendors would be struggling with the same problem and so far, there don't seem to be many reports (if any at all) coming from owners of Asus, MSI, ASRock, PNY, et al cards. This is entirely Gigabyte's fault -- be it a design or manufacturing flaw.

They also make AMD graphics cards and this issue isn't present on them(that we know of). The biggest difference I see between the cards is the TDP and the size of the cooler. Frankly, I think running traces through the PCIe hook is just poor design. I know gigabyte is getting all the attention right now but I have a feeling they aren't the only ones.

Heck, maybe the only reason Gigabyte is getting all the attention is because they're rejecting RMAs while other board partners are replacing the cards. This could be a customer service issue more than an engineering problem unique to Gigabyte

 

[apexz99]

It should be noted that the included anti sag bracket that shipped with the Gigabyte RTX 4000 series card has very limited compatibility with motherboards and cases from different manufacturers. I have reached out to Gigabyte for response on whether using the anti sag bracket is purely optional or they are fine with using bracket from other manufacturer. So far there is no response.

 

[whateversa]

Please continue to name and shame these companies. It seems that these more expensive 'premium' brands are not premium at all - they are worse than the cheaper brands when it comes to support. I've had Zotac GPUs for quite a few years now (1070, 3070) with no problems.

 

[neeyik]

They also make AMD graphics cards and this issue isn't present on them(that we know of). The biggest difference I see between the cards is the TDP and the size of the cooler. Frankly, I think running traces through the PCIe hook is just poor design. I know gigabyte is getting all the attention right now but I have a feeling they aren't the only ones.

Heck, maybe the only reason Gigabyte is getting all the attention is because they're rejecting RMAs while other board partners are replacing the cards. This could be a customer service issue more than an engineering problem unique to Gigabyte

Gigabyte doesn't make the largest/heaviest 4090s out there (for example, the Asus Strix OC is 0.5 kg heavier than Gigabyte's Gaming OC model) and there are Radeon RX 7900 XTX SKUs that are heavier than 4090s (e.g. Asus, again, with its TUF OC model is 2.14 kg). Rossmann confirms this himself, in his video, so the cooler mass/size isn't the direct cause here.

Neither, for that matter, is the TDP -- the Radeon HD 7950 X2, R9 290 X2, R9 295 X2, and R9 380 X2 all had maximum heat values in the range of 500 to 580 W. Which raises an interesting question as to why modern GPUs seem to require larger cooling systems than some cards from years ago, but that's an entirely different discussion.

And don't forget that Gigabyte 3070 & 3080 cards were also cracking -- these are all lighter than any 4090 on the market. When one considers the sheer number of these cards that have been sold over the past 2.5 years, an endemic cracking problem across all vendors would have raised its head by now, regardless of RMA policy.

 

[yRaz]

Gigabyte doesn't make the largest/heaviest 4090s out there (for example, the Asus Strix OC is 0.5 kg heavier than Gigabyte's Gaming OC model) and there are Radeon RX 7900 XTX SKUs that are heavier than 4090s (e.g. Asus, again, with its TUF OC model is 2.14 kg). Rossmann confirms this himself, in his video, so the cooler mass/size isn't the direct cause here.

Neither, for that matter, is the TDP -- the Radeon HD 7950 X2, R9 290 X2, R9 295 X2, and R9 380 X2 all had maximum heat values in the range of 500 to 580 W. Which raises an interesting question as to why modern GPUs seem to require larger cooling systems than some cards from years ago, but that's an entirely different discussion.

And don't forget that Gigabyte 3070 & 3080 cards were also cracking -- these are all lighter than any 4090 on the market. When one considers the sheer number of these cards that have been sold over the past 2.5 years, an endemic cracking problem across all vendors would have raised its head by now, regardless of RMA policy.

It actually is interesting that you bring up the TDP and coolor size. I don't have an answer for that.

I'm more curious about what's actually going on with the card than the brand surrounding it. I wish I had more to say about the cards but I really don't know right now

 

[neeyik]

I'm more curious about what's actually going on with the card than the brand surrounding it. I wish I had more to say about the cards but I really don't know right now

One possibility is the PCB manufacturer(s) that's been used for these particular models is/are at fault (assuming this component is outsourced) and Gigabyte just hasn't been thorough enough with its quality testing on the batches provided. The boards are routinely sampled and then imaged using x-rays to ensure traces, contacts, components, etc are all correctly positioned but whether the same tests would show up any defects in the fiberglass layers is unclear, and or even if they did, would the testers be tasked with routinely checking for them?

 

[yRaz]

One possibility is the PCB manufacturer(s) that's been used for these particular models is/are at fault (assuming this component is outsourced) and Gigabyte just hasn't been thorough enough with its quality testing on the batches provided. The boards are routinely sampled and then imaged using x-rays to ensure traces, contacts, components, etc are all correctly positioned but whether the same tests would show up any defects in the fiberglass layers is unclear, and or even if they did, would the testers be tasked with routinely checking for them?

I have a feeling that the people designing the PCB's aren't accustom to making them structural. I also have a feeling that the people designing nVidia's PCBs for Gigabyte are a different team than those designing AMD's cards.

This is a really interesting issue. While it's unfortunate for the people involved with broken cards, I really can't wait to know more. My brain can't wait to learn more.

 

[neeyik]

I have a feeling that the people designing the PCB's aren't accustom to making them structural. I also have a feeling that the people designing nVidia's PCBs for Gigabyte are a different team than those designing AMD's cards.

I suspect it's more down to which sub-contractor Gigabyte has been using for the PCBs, as there is no shortage of ultra-cheap vendors for such work. Upon first hearing about the cracking problem, I assumed that it was only for the really big cards, ones weight over 2.2 kg, and that it was probably a design failure. However, once I learned that OEM 3070s were cracking, then it became clear that it wasn't a cooler weight or design issue -- it was a PCB manufacturing one.

In the Gamers Nexus Gigabyte factory tour, it's stated that the company goes through about 400,000 PCBs for graphics cards and motherboards per month, so it's unlikely that Gigabyte themselves makes them (probably even using multiple sources).

So really, the onus is on Gigabyte to sample the boards, to an appropriate degree of sample count, frequency, and test types to ensure that the boards can withstand the loads they'll experience. This is especially true for OEM products, as the products will go through multiple distributions before reaching the end user.

Of course, manufacturing on this scale is always going to result in some faulty products reaching the end user. I recently had to return a Zotac card, barely two days after getting it, because a fan/RGB header completely failed. Cracking PCBs is no different -- it's just one of the many components that go into making a graphics card and as odd as the failure is, it's not a significant issue, given the number of reported failures.

However, Gigabyte's handling of the problem is an issue. It appears to readily replace cards with faulty VRMS or DRAM modules, but to not accept that the PCB shouldn't be cracking or treat the failure in the same manner as any other component, seems very short-sighted.

 

[yRaz]

However, Gigabyte's handling of the problem is an issue. It appears to readily replace cards with faulty VRMS or DRAM modules, but to not accept that the PCB shouldn't be cracking or treat the failure in the same manner as any other component, seems very short-sighted.

One reason I don't think this problem is just Gigabyte's problem is because of how many partners contract with the same PCB manufactures. I'm not saying this is a good thing, but I really think the only reason we're hearing about Gigabyte right now is because of their customer service issues.

I'm honestly not trying to make excuses for Gigabyte at this point, I really think they could have handled this on the customer service side. I've been following tech for nearly 30 years, this is one of the most interesting issues I've heard of. It was one thing when electrolytic caps were going bad on motherboards or power supplies popping in the mid 2000's, but PCB damage? It's a really interesting issue and I hate to say that because people are losing lots of money because of this, but I can't wait to see a deeper dive into what's happening. Louis Rosseman already did a good job, but there are also company politics behind this and I think we're in for a really interesting rabbithole as this issue moves forward

 

[hwertz]

Anti-sag bracket? Damn that's a new one on me. I have not bought any of these multi-100W GPUs so I guess nothing I've bought is heavy enough. Sounds like they REALLY need to provide this bracket with all their GPUs if they are getting cracks!