News EVGA motherboard owners furious over modern GPU issues — DIY users resort to taping over pins to fix RTX 50 Series problem on Z690 boards

[razor512]

Is there a way to test or get Nvidia to give info on if the RTX5000 cards have any special SMBUS functions or info that users can access?

 

[acadia11]

EVGA boards always had a good reputation, and that one was designed and released before they separated from Nvidia and reduced their overall component presence, so there wasn't much reason to think there'd be any problems. Just being EVGA is certainly not an indication that it's not capable of using a 5090 to full capability. Certainly not a difference of covered parking versus sitting out in the elements.

But being released 3 years ago is …

 

[redgarl]

EVGA PSUs and GPUs have always been rock solid. Mobos not so much.

No, I had two 1080 GTX dying on me in 10 months. I sold my third card after receiving the RMA. I never bought another EVGA product after that.

 

[thestryker]

Nvidia has nothing to do with the demise of EVGA. The CEO didn't want to do it any more and refused the sell the company, that guaranteed its downfall regardless of anything else. Rather than just publicly tell the truth and be honest with his employees, he instead decided to scapegoated Nvidia.

I suppose you have lunches with him on the regular to know this is the only motive 🙄

The reality is nvidia has been increasing margins for years which pinches AIBs and EVGA was the only manufacturer solely relying on contract manufacturing. This makes their margins lower than everyone else's and minimizes the number of SKUs they can sell. It's also widely known that nvidia has taken full control of how their GPUs can be used and locked out AIB controls at every turn. All of this adds up to video cards being a very difficult business to be in especially if you're trying to do more than sell a minimum viable product.

None of this is to say nvidia is the only reason EVGA bailed on the business, but to suggest they had nothing to do with it is foolish.

 

[artk2219]

But being released 3 years ago is …

I mean sure, but it's still LGA 1700. There isn't anything fundamentally different about LGA 1700 motherboards from 3 years ago, and LGA 1700 now. Sure Pcie 5 on some new boards, but that's not really much of a performance difference from pcie 4 with an RTX 5090, it's sub 1%.

 

[ezst036]

only issue with that is do you want someone making bios/uefi that open because that can be quite malicious

Yes.

I don't happen to own one of these EVGA boards but I have little doubt that if the UEFI for this board had been open sourced from the beginning or more specifically, had they used something like COREBOOT or LinuxBOOT, this issue may have already been resolved or now with this much scrutiny it would have been turned into a bug report and a higher profile but report at that.

Anybody with this board right now has pretty much zero chance of ever seeing a fix for this. But with an open BIOS it's plausible if not even certain that it would be fixed.

 

[das_stig]

Still find hard to believe EVGA was not taken by another company, wish it had been AMD and became their inhouse specialist branded hardware team. Like many people I paid extra to get a product from EVGA, as I trusted quality and reliability.

 

[Loadedaxe]

Still find hard to believe EVGA was not taken by another company, wish it had been AMD and became their inhouse specialist branded hardware team. Like many people I paid extra to get a product from EVGA, as I trusted quality and reliability.

You can never get a clear answer why, but the story was the CEO was tired of dealing with Nvidias grip, competing against FE cards etc.

Also, he had no interest in selling, just dropped Nvidia, went on a long extended vacation and let the show run itself on PSUs and peripherals, and it slowly filled with water and sank.

The website has been down for days and doesn't appear to be coming back. We will see.

 

[beyondlogic]

Yes.

I don't happen to own one of these EVGA boards but I have little doubt that if the UEFI for this board had been open sourced from the beginning or more specifically, had they used something like COREBOOT or LinuxBOOT, this issue may have already been resolved or now with this much scrutiny it would have been turned into a bug report and a higher profile one at that.

Anybody with this board right now has pretty much zero chance of ever seeing a fix for this. But with an open BIOS it's plausible if not even certain that it would be fixed.

Evga will never release a open bios that would be company suicide and a security risk and liable to be sued. you might want it but the rest of board users probly wouldn't the end users of these boards know evga doesn't make them anymore and it's been ages realistically there lucky if any supports on going. I suspect if enough users Nvidia may actually on it and work with evga.

 

[beyondlogic]

Yes.

I don't happen to own one of these EVGA boards but I have little doubt that if the UEFI for this board had been open sourced from the beginning or more specifically, had they used something like COREBOOT or LinuxBOOT, this issue may have already been resolved or now with this much scrutiny it would have been turned into a bug report and a higher profile one at that.

Anybody with this board right now has pretty much zero chance of ever seeing a fix for this. But with an open BIOS it's plausible if not even certain that it would be fixed.

Evga will never release a open bios that would be company suicide and a security risk and liable to be sued. you might want it but the rest of board users probly wouldn't the end users of these boards know evga doesn't make them anymore and it's been ages realistically there lucky if any supports on going. I suspect if enough users Nvidia may actually on it and work with evga.

 

[spongiemaster]

I suppose you have lunches with him on the regular to know this is the only motive 🙄

The reality is nvidia has been increasing margins for years which pinches AIBs and EVGA was the only manufacturer solely relying on contract manufacturing. This makes their margins lower than everyone else's and minimizes the number of SKUs they can sell. It's also widely known that nvidia has taken full control of how their GPUs can be used and locked out AIB controls at every turn. All of this adds up to video cards being a very difficult business to be in especially if you're trying to do more than sell a minimum viable product.

None of this is to say nvidia is the only reason EVGA bailed on the business, but to suggest they had nothing to do with it is foolish.

The CEO gave multiple interviews around this announcement that gave insight into what was really going on. If you're not familiar with any of those, I'm not going to rehash everything he said. Do your own research. One of the things he said was that he wanted to spend more time with his family. You didn't need to have to eat lunch with him to know that, he told everyone. Maybe your internet wasn't working that day. Crying poor margins when GPU manufacturers had likely seen record margins through the ethereum bubble doesn't fly. GPU's were 80% of the company's revenue. There was no way that exiting the market wasn't going to kill the company. Everyone at the time could see this, but the CEO didn't? If Nvidia was such a problem, why not sell AMD GPU's and keep the company solvent? Since you're such an expert on this, please tell us what Nvidia had to do with EVGA leaving the motherboard market at the same time?

 

[tiredcrow]

Maybe it's Nvidia's fault as much as evga's?

My exact thought when reading that section, "Hold on, then why does the GPU have comm lines to it?" It's obviously not unwired in the GPU itself if bits are being sent over the pins. What is the GPU use of this?

 

[_D_D]

Is there a way to test or get Nvidia to give info on if the RTX5000 cards have any special SMBUS functions or info that users can access?

Usually I've used the Nvidia library to do that, well for Windows, Linux function is missing. IIRC my GTX 1050Ti has SMBUS connections to the GPU and PCIe, the X99 board I have only links the connections through each of the PCIe slots but the Z77 goes to chipset. I don't have a 50 series card.

Great work done by the guy(s) who worked out what the problem was.

 

[edzieba]

1) SMBus is meant to be on those traces, if the EVGA boards aren't handling it then that's on EVGA to fix (which sadly means no fix is possible, as it generally requires a board respin to implement properly). This issue has cropped up elsewhere, and is common to the point people offer services to pre-isolate the SMBus pins for use with motherboards that have SMBus isolation conflicts. "But maybe it's Nvidia's fault?!" is just lazy tribalism. If you expect I2C devices to deconflict via hopes and dreams rather than address isolation, you're gonna have a bad time, since I2C devices (e.g. fan controllers, voltage monitors, etc) are all commodity parts and demanding "no other vendor can use this random vendor's I2C address, it's my address!" is not a practical solution. The motherboard is in change of handling SMBus conflicts, and if conflict resolution is 'don't boot at all' then that's on the motherboard vendor.

2) EVGA did not just dump Nvidia as a supplier and switch to AMD or Intel, they stopped all hardware production for things unrelated to GPUs (e.g. PSUs). All other Nvidia OEMs shrugged their shoulders and continue manufacturing without issue. The problem was clearly with EVGA's hardware division being unsustainable as a business, and Nvidia made for a convenient scapegoat.

3) Open-sourcing UEFI (or installing Coreboot, which is an existing open EFI implementation) likely would not solve the issue, as the problem is proper SMBus isolation not being implemented. At best you could apply a brute force 'fix' to kill SMBus to all peripheral devices, but only if
a) The motherboard traces even allow for this (i.e. different paths for PCIe ports and all other devices) as SMBus is required for the system to operate, so you can't just turn it off completely.
b) You are willing to give up on the things that SMBus on PCIe does (e.g. monitoring card voltages and fan speeds independently of vendor software).

 

[Loadedaxe]

@edzieba makes a solid case, and I agree that proper SMBus isolation is fundamentally the responsibility of the motherboard vendor......especially when integrating commodity I2C components with fixed addresses. That being said, there’s still a conversation to be had around cross vendor coordination. While expecting Nvidia to accommodate every potential SMBus routing implementation might be unrealistic, designing hardware that minimizes the chance of bus address collisions could improve overall system stability.

As far as EVGA goes, I would agree with @edzieba , their exit wasn't a simple pivot to another GPU vendor, it was a strategic withdrawal from all hardware sectors, including PSUs. If you want to argue that, all we can go by is what we see, if you want to base your judgement off what the CEO discussed in interviews, well there your naive. One thing we do know, no matter why EVGA exited, it doesn't look good for EVGA customers, including me, I own a few PSUs.

Still, the dynamics between EVGA and Nvidia weren’t trivial. It’s reasonable to suspect that strained supplier relations accelerated what was already an unsustainable business trajectory for EVGA.

As for UEFI or Coreboot? .... open-sourcing the firmware won't magically correct signal integrity issues or implement proper bus isolation. However, maybe more transparency can empower developers to create firmware level workarounds, perhaps via selective SMBus device disabling or remapping where feasible? maybe, I don't know.

Admittedly, that would be contingent on the board having discrete trace routing and sufficient granularity in the firmware’s power and communication management stack.

It is a thought.

 

[DS426]

This is a great example of what it means when something that's "in support" vs. something that isn't. Sure, a lot of products -- especially good quality ones -- will long outlast their end-of-support or EoL date, but issues like this can come up that likely won't get resolved by the manufacturer or software vendor.
As far as I'm concerned, nVidia killed EVGA, so are we really going to get furious with them (EVGA) over their dead product lines? EVGA doesn't have in-house programmers at this point, so we all have to be realistic and see it for what it is.

 

[Loadedaxe]

This is a great example of what it means when something that's "in support" vs. something that isn't. Sure, a lot of products -- especially good quality ones -- will long outlast their end-of-support or EoL date, but issues like this can come up that likely won't get resolved by the manufacturer or software vendor.
As far as I'm concerned, nVidia killed EVGA, so are we really going to get furious with them (EVGA) over their dead product lines? EVGA doesn't have in-house programmers at this point, so we all have to be realistic and see it for what it is.

I agree the distinction between supported and unsupported hardware is an important one. While high quality products can remain functional long past their EoL, the lack of firmware or driver updates means issues like SMBus conflicts or compatibility gaps often go unaddressed. That’s just the reality of relying on legacy hardware.

As for EVGA, their exit from GPU manufacturing was a complex situation. While Nvidia's business practices may have played a role, other Nvidia board partners are still operating without issue, suggesting that EVGA’s struggles were multifaceted.

If EVGA had been in a stronger position financially and operationally, they likely would have made a smarter decision to work out their issues with Nvidia rather than exiting hardware production entirely. The fact that they opted to shut down instead speaks to deeper internal challenges beyond just vendor relations.

 

[evokethunder]

I'm gonna be honest, this is 100% Nvidia's fault.

While it's true that the vast majority of mobo manufacturers don't connect the smbus pins, it is not outside the realm of reason.

If Nvidia is connecting smbus on cards that don't support it, that is on them completely. Nvidia can't just go off the assumption that manufacturers "likely won't" provide smbus on the board. This is a classic example of a big company cutting corners to save money and blaming the little guy for it.

Nvidia should have disabled those pins on the GPU if they were not going to make use of it.

I'll give an example. Let's say your job is to replace valves on water tanks. 90% of the time, the tanks are drained before you arrive to change the valve, but draining isn't a requirement because let's just say that it is also in your job description even though you usually don't have to do the draining.

Imagine you arrive at a site and take the valve off without checking and water starts pouring out and draining somewhere it isn't supposed to. Are you gonna blame the client for not draining it (even though that's just a courtesy, and not required?) it's your responsibility to check that it's empty.

It's the same thing here with Nvidia. They just assumed the pins wouldn't be there on the board, so they didn't bother doing their job correctly. Linus Sebastian has said before that Nvidia is a greedy <Mod Edit> company, and he wasn't wrong. This is just more evidence of that.

 

[evermorex76]

But being released 3 years ago is …

It's still a high-end board, which can carry a high-end processor that could meet the requirements to not be a bottleneck with a 5090. There isn't THAT big a difference between the Alder Lake and Raptor Lake series and current processors.

 

[evermorex76]

SMBus is meant to be on those traces, if the EVGA boards aren't handling it then that's on EVGA to fix (which sadly means no fix is possible, as it generally requires a board respin to implement properly).

You're assuming that EVGA did NOT implement it properly on the boards, rather than the possibility that Nvidia is the one that didn't implement it properly on the GPU. I can't see any reason that EITHER of them would have even included the pins and not have designed it properly, or why board makers would have gone to the trouble of putting in the traces to connect the pins if it wasn't done properly, so without real testing by someone that can determine where the problem lies, your assumption is no more valid than anyone else's and sounds like the same tribalism. The fact that this issue occurs not just on EVGA boards but on other brands and even systems from OEMs like HP and Dell, and with devices like RAID controllers and NICs further complicates the situation. If EVGA isn't the only one that has weird "conflict resolution" then is EVGA actually doing something wrong?

 

[edzieba]

You're assuming that EVGA did NOT implement it properly on the boards, rather than the possibility that Nvidia is the one that didn't implement it properly on the GPU.

The motherboard is the component that manages the SMBus. The worst a device (like a GPU or anything else on the bus) can do is share an I2C ID with another component, but since I2C components sharing an ID is commonplace (since I2C components are commodity parts, and two unrelated vendors can both buy the same I2C chip) it is on the motherboard to be able to handle that gracefully.

It's like a piece of software crashing because two users have the same name, and then blaming the users for not choosing unique names.

 

[thestryker]

If Nvidia was such a problem, why not sell AMD GPU's and keep the company solvent?

How do we know this would have helped? They would be competing in a much smaller market without guaranteed sales. They'd also be starting a new relationship with GPU supplier and if AMD does things like nvidia they wouldn't necessarily have had priority access.

Since you're such an expert on this, please tell us what Nvidia had to do with EVGA leaving the motherboard market at the same time?

EVGA solely relied on contract manufacturing for video cards and motherboards. The vast majority of their volume came from video cards which makes the contracts cost more without them. The chances of their motherboard business being viable after bailing on video cards is zero.

Crying poor margins when GPU manufacturers had likely seen record margins through the ethereum bubble doesn't fly.

How exactly do AIBs benefit from retailer markups? Everyone sold more cards that's for sure, but that doesn't mean everyone made more money on every card sold. The margins on the products don't change unless they raise the prices and EVGA didn't when selling direct until well into things (and who knows what volume direct sales are and what retail contracts look like). The crypto boom also led to overproduction by both AMD and nvidia and in turn over buying by AIBs (I don't know how much of a problem this was for EVGA specifically).

GPU's were 80% of the company's revenue. There was no way that exiting the market wasn't going to kill the company. Everyone at the time could see this, but the CEO didn't?

I'm not sure what relevance this has to anything, but I'm sure he did know it likely would. I'd be surprised if anyone in the company didn't know they were winding down. They still had a fair amount of stock to sell so no CEO in that position is going to come out and say "we're going out of business". It should have been obvious at a public level when the new high end PSUs came out to no fanfare and a 3 year warranty.

The CEO gave multiple interviews around this announcement that gave insight into what was really going on. If you're not familiar with any of those, I'm not going to rehash everything he said. Do your own research. One of the things he said was that he wanted to spend more time with his family.

I guess two counts as multiple since it is more than one. I enjoy "do your own research" line always rolled out by those in the know. In your world I guess wanting to spend more time with the family is something that exists in a vacuum. It couldn't possibly be part of the reasoning instead must be the whole. It couldn't possibly be cultivated by a lack of enjoyment in the work caused by the realities of the business.

 

[Loadedaxe]

I'm gonna be honest, this is 100% Nvidia's fault.

While it's true that the vast majority of mobo manufacturers don't connect the smbus pins, it is not outside the realm of reason.

If Nvidia is connecting smbus on cards that don't support it, that is on them completely. Nvidia can't just go off the assumption that manufacturers "likely won't" provide smbus on the board. This is a classic example of a big company cutting corners to save money and blaming the little guy for it.

Nvidia should have disabled those pins on the GPU if they were not going to make use of it.

I'll give an example. Let's say your job is to replace valves on water tanks. 90% of the time, the tanks are drained before you arrive to change the valve, but draining isn't a requirement because let's just say that it is also in your job description even though you usually don't have to do the draining.

Imagine you arrive at a site and take the valve off without checking and water starts pouring out and draining somewhere it isn't supposed to. Are you gonna blame the client for not draining it (even though that's just a courtesy, and not required?) it's your responsibility to check that it's empty.

It's the same thing here with Nvidia. They just assumed the pins wouldn't be there on the board, so they didn't bother doing their job correctly. Linus Sebastian has said before that Nvidia is a greedy <Mod Edit> company, and he wasn't wrong. This is just more evidence of that.

The motherboard manufacturer controls the SMBus and how it's implemented on their boards. They decide whether to connect those pins and how the communication functions between components.

In this case, the issue seems to stem from a mismatch in expectations, Nvidia assumed most manufacturers wouldn't connect SMBus, while some actually did. If motherboard makers had clearer standards or ensured consistency in whether SMBus was enabled, it might have avoided this situation altogether.

I think this is more a matter of poor communication between manufacturers really. But at the end of the day the board manufacturers are responsible for the implementation of the SMBus.

I get that some people trust Linus’s opinions on tech, but when it comes to ethics in business decisions, I don’t think he’s the most reliable source. He’s great at breaking down technical issues for the most part, but his views on companies often feel more like personal grievances than objective analysis.

While Nvidia has made questionable moves, labeling them as purely greedy might be an oversimplification. There’s always more to the story, and business decisions, especially in the tech world, are rarely black and white.

 

[edzieba]

EVGA solely relied on contract manufacturing for video cards and motherboards. The vast majority of their volume came from video cards which makes the contracts cost more without them. The chances of their motherboard business being viable after bailing on video cards is zero.

If EVGA chose an unsustainable business model then that's down to EVGA, not one of their suppliers.

Plus, plenty of other GPU vendors are also using contract manufacturing, with the vast majority having boards manufactured and assembled by either Foxconn or Pegatron.

 

[thestryker]

If EVGA chose an unsustainable business model then that's down to EVGA, not one of their suppliers.

Right... because this couldn't possibly have anything to do with it at all:

Plus, plenty of other GPU vendors are also using contract manufacturing, with the vast majority having boards manufactured and assembled by either Foxconn or Pegatron.

Can you name a single video card manufacturer who doesn't have their own factories? I'm not aware of any, but maybe you are.

note: I'm not saying they make every part just that they have their own factories for at the very least final manufacture