EVGA - Long-Time NVIDIA Partner - Ending Graphics Card Production

[qarium]

I highly recommend to everyone watching both GN and JTC videos, they offer slightly different perspective on the same topic since they were both invited to the meeting.

- manufacturers get to know nvidia gpu prices when leather jacket goes on stage (you know, the gpus they've ordered and already manufactured cards with)
- they don't even get the damn drivers(!) before the press does, "because muh leakz"
- leather jacket apparently thinks they should be thankful and kissing nvidia's ass for being treated like trash

Jayz put it best, calling it abusive relationship. How the hell do you run your business when dealing with a "partner" like nvidia?

I'm just kinda worried about Nintendo's relationship with nvidia now, since I'm planning to buy Switch Lite and expect them to release next nvidia-based Switch down the line (they have to, because of backward compatibility - they're using specific low-level nvidia api and running that efficiently on other gpus, while possible, could be more trouble than its worth). Microsoft tried dealing with nvidia already, during the first Xbox era, and that didn't go well. Of course big corporations like Nintendo or Microsoft can handle it, but with nvidia strongarming everyone into submission it's likely not that different. If Nintendo plans to use AI-based upscaler in their next Switch, they're screwed - they need ARM SoC with decent GPU, AI accelerator and software solution supporting it - and nvidia is the only company that has all three (FSR runs on regular shader cores and likely wouldn't be energy-efficient enough for mobile use).

yes nvidia is complete trash shit abusive relationship...

did you know AMD RDNA3 has AI FMA/Matrix-cores to calculate FSR3.0(similar to DLSS2.x) in hardware instead of shaders this gives another ~7% performance boost.

with RDNA3 nvidia lose this DLSS2.x runs on AI/Matrix cores advantage in energy-efficieny..

[Quackdoc]

i told you that this possibility is not on the table without getting money for it.
they only have the possibility of going bankrupt or drop support.

longer support times need higher sales price and the only company who can get higher sales prices in the gpu market is nvidia
thats why you get longer official support on nvidia hardware.

you also dislike the non-official support from companies like RedHat man ... then your only option is buy nvidia.



a card you paid for 4 years later is not give a company any income to finance the support. without pay extra to redhat or amd you dont get longer support. and if you want longer support for your innitial payment at the moment you bought it then the price of the gpu need to be higher to finance the longer support. the only company who is able to get a higher sales price is nvidia because of this nvidia gives you longer official support.

yes reality "sucks" but you clearly do not life in reality. you life in dreamworld....

"AMD doesn't have amf hevc support for it"

HEVC/H265 in my point of view has no relevance for the future because AV1 is the future.

as faar as i know AMF is the windows UVD video architecture why do you expect it on linux? is there a software you like who use this ?

"it's also missing critical vulkan extensions I need for features that I very much want."

i am 100% sure amd still supports 480/580/590 for vulkan extensions they only as far as i know did cut the ROCm/HIP stuff.

can you tell me what vulkan extention you miss ? or is it this one: VK_VALVE_mutable_descriptor_typehttps://www.phoronix.com/news/Vulkan-1.3.228-Released

"as well as other misc gripes that have added up over the years."

what exactly ? ... you do not need to answer i just try to unterstand what you write and mean.

"ROCm is simply the final nail, as I have said"

... you paint it black but in reality there are Git commits on Polaris support for ROCm/HIP...

just quit your believe that AMD is the only driver supply for Polaris hardware try the opensource version of ROCm for example from the ARCH packages... they do in fact support polaris.

i also have another idea for you here in this forum are many people who do upgrade to the lastest tech already
this means it is super simple to get a vega64 and just forget about the old polaris topic.
on ebay a vega64 is 250€ not a big deal if you ask me. dont see the life so serious. just keep in mind that you get money for your polaris card on ebay to. for example your old polaris card on ebay gives you 110€ this means its only 140€ difference. and its a very good upgrade stradegy to always switch to the oldest supported generation.

you have absolutely no idea what you are talking about...

first of all, I don't care what you think about hevc. I care about what I need. we don't all live in a world where what someone thinks bears relevance against reality

second of all, Nvidia is supporting their 4 year old cards perfectly fine, and so is intel, AMD is the outlier here.

third of all when RHEL comes out and says they will officially support polaris in a sufficient matter, then it will be fine. as I said, an official statement is absolutely necessary, I live in the real world, not one where a commit here and there matter.

forth polaris STILL doesn't support VK_EXT_image_drm_format_modifier on either radv or amdvlk. there were a few other extensions if I am remembering right, that other cards of the era support.

fifth

far as i know AMF is the windows UVD

It's not.

sixth there are plenty of other gripes from AMD, things like how long it took to get raytracing on linux, overall lack of compute ecosystem, a lower preformance celing, I still have power efficiency issues on all of my AMD cards on linux.

seventh, Yes A 4 year old card won't make them more money, but neither will kicking their loyal consumers who have supported them through the garbage like the radeon 200 and 300 series, the same customers who buy their crap like the HD gpus they put out for their ultra low end.

[qarium]

you have absolutely no idea what you are talking about...
first of all, I don't care what you think about hevc. I care about what I need. we don't all live in a world where what someone thinks bears relevance against reality
second of all, Nvidia is supporting their 4 year old cards perfectly fine, and so is intel, AMD is the outlier here.
third of all when RHEL comes out and says they will officially support polaris in a sufficient matter, then it will be fine. as I said, an official statement is absolutely necessary, I live in the real world, not one where a commit here and there matter.

redhat does it as soon as they have a customer who pay for it. and in reality a commit here and there does in fact matter.

forth polaris STILL doesn't support VK_EXT_image_drm_format_modifier on either radv or amdvlk. there were a few other extensions if I am remembering right, that other cards of the era support.

i did write a message to bridgman about this lets wait and see. who knows maybe the vulkan support is also chanceled for polaris cards who knows

fifth It's not.
sixth there are plenty of other gripes from AMD, things like how long it took to get raytracing on linux, overall lack of compute ecosystem, a lower preformance celing, I still have power efficiency issues on all of my AMD cards on linux.
seventh, Yes A 4 year old card won't make them more money, but neither will kicking their loyal consumers who have supported them through the garbage like the radeon 200 and 300 series, the same customers who buy their crap like the HD gpus they put out for their ultra low end.

it looks like outside of the monnopole companies there is a lose lose lose situation.

but its not news that making business agaist a monopole is a hard job.

[piotrj3]

nvidia has a long history of screwing over it's partners, I have yet to hear such complaints about AMD, but that could just be because few people in comparison cared.

Few complains from video were exactly the same or even worse for AMD.

1st. AMD MSRP for their "FE" cards was even lower so even without price hiking up for production AIBs couldn't make any money. Literally when AMD announced RX6800XT for 649$, not a single AIB was offering it at that price. For nvidia there were actually AIB cards matching FE MSRP meaning AMD was even more extreme.

2nd. 5600XT bios situation anyone? When AMD released BIOS that was supposed to put GDDR6 memory on higher speeds but didnt' tell a single non exlusive AIB (Asus, MSI etcc.) that they should validate their cards for it? But for some reason only AIBs exclusive to AMD knew to validate it for it.

3rd. Unstable RDNA1 drivers made return rates of AMD gpus much higher then Nvidia which of course increases costs of servicing/warranties.

I think true reason why EVGA broken with Nvidia was that Nvidia was late to update prices of chips to market realities. Prices were going up, and all AIBs (and nvidia) was making record margins, (EVGA literally had them). Now prices go heavy down and failing to adjust prices by nvidia (as supplier of chips) is reason they go away.

People kinda forget that supply cost is not only chip, but also memory. Take for example RTX 3070 and RX6800. Even if price of RTX 3070 gpu is higher, costs of 8GB of VRAM is way lower then 16GB. And AIB has to buy both.

[Quackdoc]

Few complains from video were exactly the same or even worse for AMD.

1st. AMD MSRP for their "FE" cards was even lower so even without price hiking up for production AIBs couldn't make any money. Literally when AMD announced RX6800XT for 649$, not a single AIB was offering it at that price. For nvidia there were actually AIB cards matching FE MSRP meaning AMD was even more extreme.

2nd. 5600XT bios situation anyone? When AMD released BIOS that was supposed to put GDDR6 memory on higher speeds but didnt' tell a single non exlusive AIB (Asus, MSI etcc.) that they should validate their cards for it? But for some reason only AIBs exclusive to AMD knew to validate it for it.

3rd. Unstable RDNA1 drivers made return rates of AMD gpus much higher then Nvidia which of course increases costs of servicing/warranties.

I think true reason why EVGA broken with Nvidia was that Nvidia was late to update prices of chips to market realities. Prices were going up, and all AIBs (and nvidia) was making record margins, (EVGA literally had them). Now prices go heavy down and failing to adjust prices by nvidia (as supplier of chips) is reason they go away.

ahh interesting, seems it was in the no one really cared as much so it made less news. I knew about the 3rd point. but not the second and third.

[brucethemoose]

I'm just kinda worried about Nintendo's relationship with nvidia now, since I'm planning to buy Switch Lite and expect them to release next nvidia-based Switch down the line (they have to, because of backward compatibility - they're using specific low-level nvidia api and running that efficiently on other gpus, while possible, could be more trouble than its worth). Microsoft tried dealing with nvidia already, during the first Xbox era, and that didn't go well. Of course big corporations like Nintendo or Microsoft can handle it, but with nvidia strongarming everyone into submission it's likely not that different. If Nintendo plans to use AI-based upscaler in their next Switch, they're screwed - they need ARM SoC with decent GPU, AI accelerator and software solution supporting it - and nvidia is the only company that has all three (FSR runs on regular shader cores and likely wouldn't be energy-efficient enough for mobile use).

Nintendo chose Nvidia Tegra because it was cheap and available. But the next gen Tegra chips (Xavier, Orin, Atlan) are all bigger-die automotive/embedded chips. They are too expensive for Nintendo, and probably too power hungry.

I think a custom design is out if the question. Nintendo is cheap, and 7nm or less Nvidia is not.

They *could* use the Tegra X2. That would be incredibly stingy of Nintendo, but I would not put it past them.



Also, AMD and Apple have ML hardware and low level APIs, but not in anything available. Hence I think the real match is Intel. Once the ARC IGPs come out, they will have everything Nintendo needs in a single package.

[stormcrow]

second of all, Nvidia is supporting their 4 year old cards perfectly fine, and so is intel, AMD is the outlier here.

Hardly. I've got pretty ancient AMD cards that are still functional going back to the AGP bus days. They aren't great on modern software strictly because of how processor heavy modern UIs have become, but they're still fully functional with driver support. Not all of that support is directly from AMD, of course, however they have from time to time dumped their support documentation to the wider world to allow continued functionality of that hardware. Whereas Intel generally has also been pretty good in supporting old video hardware, Nvidia has been the world's worst since the Bad Old Days (when you were lucky to go an hour without X freezing regardless of your graphics adapter). The only way you get Nvidia full hardware support on any given board is through their proprietary drivers and/or support libraries, which won't last indefinitely. If that doesn't matter to you, then it doesn't matter. Keep buying what floats your boat.

Say what you will about bugs in AMD drivers (which may or may not actually be in the drivers nor even AMD's fault even if they are given that the whole stack other than the firmware is open source), but at least the basic support is there. You can't say that about Nvidia drivers & keeping current with recent software and kernel versions. Note I'm purposely ignoring nouveau because I said "full hardware support". I honor what the people behind nouveau are trying to do, but they will remain hamstrung by Nvidia's "our way or the highway" mentality which is unlikely to change even with that bare minimum driver code drop a few months ago.

[Quackdoc]

Hardly. I've got pretty ancient AMD cards that are still functional going back to the AGP bus days. They aren't great on modern software strictly because of how processor heavy modern UIs have become, but they're still fully functional with driver support. Not all of that support is directly from AMD, of course, however they have from time to time dumped their support documentation to the wider world to allow continued functionality of that hardware. Whereas Intel generally has also been pretty good in supporting old video hardware, Nvidia has been the world's worst since the Bad Old Days (when you were lucky to go an hour without X freezing regardless of your graphics adapter). The only way you get Nvidia full hardware support on any given board is through their proprietary drivers and/or support libraries, which won't last indefinitely. If that doesn't matter to you, then it doesn't matter. Keep buying what floats your boat.

Say what you will about bugs in AMD drivers (which may or may not actually be in the drivers nor even AMD's fault even if they are given that the whole stack other than the firmware is open source), but at least the basic support is there. You can't say that about Nvidia drivers & keeping current with recent software and kernel versions. Note I'm purposely ignoring nouveau because I said "full hardware support". I honor what the people behind nouveau are trying to do, but they will remain hamstrung by Nvidia's "our way or the highway" mentality which is unlikely to change even with that bare minimum driver code drop a few months ago.

im not asking for the odd bug patch to keep the card usable, im asking for official support on a card thats only about 4 years old.

[fitzie]

the gpu 3rd party aftermarket serves no purpose except for the rgbs and overclockers, who often just put on yet another cooling system. when you can do 4k/hdr/120hz with a gpu inside the cpu, will anyone except some esports players get an discrete gpu? That day is coming soon. Maybe evga is cutting their losses, maybe they want to switch to amd and this is the first step they have to do legally due to contracts with nvidia. Either way, there will be two or maybe three more generations of gpus before no one bothers to buy them anymore.

Of course, I say this as someone who still buys nics and sata cards, and has external dacs, because I have needz, but still

[theriddick]

They said repeatedly that they are not interested in partnering with AMD or Intel. Maybe that could change, but as of right now, they are cutting away 78% of their business.

I think ALLOT of people don't realise that AMD is still only barely %30 of the discrete GPU market.
NVIDIA is STILL the king! I don't like it either but lets face it, they offer the stuff gamers want, functional RT and DLSS that CAN beat native outputs.
Many may disagree with me but AMD is still catching up with NVIDIA's feature set even thought IMO AMD has better silicon roadmap, that isn't good enough by itself!

Still waiting for AMD RT performance and stability (basically a no show in RADV MESA stable atm) and also still no HDMI-2.1 support while NVIDIA has it in open and closed drivers!

PS. 6800xt owner here. I'm not talking out of hole in my head.