EVGA expected to wind down business after exiting Motherboard Market and EVGA Taiwan Staff allegedly quitting

[leadeater]

2 hours ago, starsmine said:

The retail 4090 cooler is already overbuilt, the fact that one larger exists and was made suggest Nvidia WAS experiencing with even more wattage, however that would turn out. The leak was real, but because of an internal pivot for whatever reason, it did not come out to retail. 

That doesn't actually mean that, going with the board mounted along the bottom instead of the back simply required a wider cooler, no avoiding that. The reason why 800W GPUs were never going to be a consumer part is because you cannot get the heat away from the die using these cooler styles and airflow. Most of the really high wattage GPU solutions in servers are also liquid cooling only. Simply never was going to happen, not ever. The cooler maybe, maybe not. 800W rumor no way ever.

 

RTX 4090 GPU hotspot temperatures already hit 90C and a bigger fin stack alone won't change that. So I don't know why anyone thinks GPUs that only momentarily peak at 600W with hotspots of up to 90C could mean a 800W TDP (not peak) could have ever been possible. Even 800W peak isn't possible, thermal limits would be hitting pull back clocks making 800W never even possible.

 

2 hours ago, starsmine said:

These parts are not locked in until they are locked in. Revisions are made, Pivots happen, some parts are just for internal testing.

Some of the most egregious things MLID has said are things that would never change like this. Some things are so foundational that is not a revision change, you get things like this wrong and any excuses become invalid. GPU die's don't suddenly change from MCM to monolithic for example, it was always going to be monolithic. Number of SMs or CUs don't double or half, die sizes don't double or half, power targets do not change from 450W to 800W or vice versa. Obviously things change, likelihood and realistic predictions or changes matter. 

 

For many people it doesn't matter how much someone gets right if they deal in areas or say things that are so problematic they can no longer be treated as credible. If anyone wants to listen to any person, leaker or not, that is the choice of that person. Criticism is often earned, never forget that. Whether it is fair or balanced is a different matter.

 

Like I said people simply want rumors, without rumors there exists nothing until official release and that is boring. People want to know, want to discuss, want to hear things, even if they are wrong in the end. What they don't want however is to be mislead and lied to or feed things that were never going to be reality.

 

Also pays to remember when someone has lost their credibility and people stop listening that also means if they change, become more diligent and improve it's already too late for those that have left. Regaining that credibility is very difficult, which is why it's important to try very hard not to lose it in the first place.

 

Anyway as to the topic I would have to suspect EVGA may be closing down. Too many products across too many different areas all out of stock, pulling out of product markets, not willing to sell, needing to retire at some point. If there isn't anyone to hand the business over to, no successor, it's going to close.

 

Edit:

Also forgot to say, Nvidia is a publicly traded company. Any investments in to another company must be disclosed, this is not optional. It also has to be approved by shareholders. Even if Nvidia invest in to EVGA before they became publicly traded that situation would still be required to disclose.

[starsmine]

30 minutes ago, leadeater said:

That doesn't actually mean that, going with the board mounted along the bottom instead of the back simply required a wider cooler, no avoiding that. The reason why 800W GPUs were never going to be a consumer part is because you cannot get the heat away from the die using these cooler styles and airflow. Most of the really high wattage GPU solutions in servers are also liquid cooling only. Simply never was going to happen, not ever. The cooler maybe, maybe not. 800W rumor no well ever.

 

RTX 4090 GPU hotspot temperatures already hit 90C and a bigger fin stack alone won't change that. So I don't know why anyone thinks GPUs that only momentarily peak at 600W with hotspots of up to 90C could mean a 800W TDP (not peak) could have ever been possible. Even 800W peak isn't possible, thermal limits would be hitting pull back clocks making 800W never even possible.

As I recall every time MLID said stop believing 800W sustained rumors, every single time. Because other leakers were saying 800W and he disputed that specific rumor ever time the topic was brought up. I also never said 800W. just that Nvidia experimented with parts higher wattage than the 4090 we got from them. Given that the 4090 is not even the full AD102 die, I dont think that really is a stretch. 

30 minutes ago, leadeater said:

Some of the most egregious things MLID has said are things that would never change like this. Some things are so foundational that isn't not a revision change, you get things like this wrong and any excuses become invalid. GPU die's don't suddenly change from MCM to monolithic for example, it was always going to be monolithic. Number of SMs or CUs don't double or half, die sizes don't double or half, power targets do not change from 450W to 800W or vice versa. Obviously things change, likelihood and realistic predictions or changes matter. 

Not sure which one you are talking about here. Ampere did double its CUs though, well, ish. It doubled its CU per SM, so if all you have is the SM, you are doing to get the CU count wrong. If that is not what you are refering to, then im just lost. 

30 minutes ago, leadeater said:

 

For many people it doesn't matter how much someone gets right if they deal in areas or say things that are so problematic they can no longer be treated as credible. If anyone wants to listen to any person, leaker or not, that is the choice of that person. Criticism is often earned, never forget that. Whether it is fair or balanced is a different matter.

 

Like I said people simply want rumors, without rumors there exists nothing until official release and that is boring. People want to know, want to discuss, want to hear things, even if they are wrong in the end. What they don't want however is to be mislead and lied to or feed things that were never going to be reality.

 

Also pays to remember when someone has lost their credibility and people stop listening that also means if they change, become more diligent and improve it's already too late for those that have left. Regaining that credibility is very difficult, which is why it's important to try very hard not to lose it in the first place.

Not denying that, MLID needs to be criticized often. Just often the criticism I see for MLID here is just people not having basic understanding of Analytic confidence rather than a fair criticism, such as the mentioned spreadsheet. And the critism comes from people misunderstanding what people are getting out of the leaks. Its fair to push back and say hey, its not gospel to anyone who treats it as such, but thats not the take I see. 
 

30 minutes ago, leadeater said:

Anyway as to the topic I would have to suspect EVGA may be closing down. Too many products across too many different areas all out of stock, pulling out of product markets, not willing to sell, needing to retire at some point. If there isn't anyone to hand the business over to, no successor, it's going to close.

What markets are they even in anymore?

They fire sold ALL their mice last year. Their AIOs never took off. Does not look like new egg has any of their keyboards, just 3rd party marketplace sellers. 
Is it just PSUs?
The audio card I dont think had more then a couple of runs.

[leadeater]

1 minute ago, starsmine said:

What markets are they even in anymore?

Well I thought motherboards and PSUs but based on availability apparently not 

 

They also have keyboards, mice, capture cards and AIOs.

 

Ages ago they did sound cards but that became irrelevant, sound cards I mean.

 

4 minutes ago, starsmine said:

As I recall every time MLID said stop believing 800W sustained rumors, every single time.

Not every single time since MLID was one of the major sources of that rumor and is quoted in multiple articles. Sure what he said may have changed but I know full well he was a source, the topic was really active on this forum and I tried my best to caution against believing it.

 

9 minutes ago, starsmine said:

Not sure which one you are talking about here. Ampere did double its CUs though, well, ish. It doubled its CU per SM, so if all you have is the SM, you are doing to get the CU count wrong. If that is not what you are refering to, then im just lost. 

This isn't about generation changes, this is about rumors for a specific architecture and details like that change to that degree which has happened. I honestly don't remember if MLID was the one or participated in it but I know there was an instance like this but I don't remember well enough, too long ago. Just an example of things that just don't change that much. Doesn't matter the point was to be careful about leniency for changes and revisions and how far that should realistically be taken.

 

You don't get something wrong due to a revision from Nvidia making a semi-truck then it being "revised" to a bicycle. That's not a revision change. 

 

16 minutes ago, starsmine said:

Not sure which one you are talking about here. Ampere did double its CUs though, well, ish. It doubled its CU per SM, so if all you have is the SM, you are doing to get the CU count wrong. If that is not what you are refering to, then im just lost. 

FYI

 

CU = AMD

SM = Nvidia

[LAwLz]

I too can make very accurate predictions if I am allowed to come up with excuses for what I said didn't turn out to be true (they changed their plans!) as well as make multiple guesses on things like process nodes, and then say "I was right!" when one of them turns out to be true.

 

A while ago, someone took 100+ claims made by MLID and then checked if they turned out to be true or not. 

54% of his claims about AMD processors turned out to be true and 66% for Intel claims. However, those numbers do not include the videos he deleted, and it included very vague claims like "the next-gen processor will have improved cache" (without elaborating on how it would be better) and "LGA1700 will support more than one generation of processors". You know, stuff that has been true for the last 5+ generations and probably will be true for the next 5 as well.

By the way, here is another Excel spreadsheet that someone made for other leakers. AdoredTV, another leaker I absolutely hate because he is an idiot who makes things up all the time for clicks, got 37% accuracy, and even then they were pretty nice to him because they listed things like "Nvidia will continue to make money a lot of money from GPUs" as a claim he got correct. 

 

 

I really don't get how anyone can still take either of them seriously. 

[dogwitch]

26 minutes ago, LAwLz said:

I too can make very accurate predictions if I am allowed to come up with excuses for what I said didn't turn out to be true (they changed their plans!) as well as make multiple guesses on things like process nodes, and then say "I was right!" when one of them turns out to be true.

 

A while ago, someone took 100+ claims made by MLID and then checked if they turned out to be true or not. 

54% of his claims about AMD processors turned out to be true and 66% for Intel claims. However, those numbers do not include the videos he deleted, and it included very vague claims like "the next-gen processor will have improved cache" (without elaborating on how it would be better) and "LGA1700 will support more than one generation of processors". You know, stuff that has been true for the last 5+ generations and probably will be true for the next 5 as well.

By the way, here is another Excel spreadsheet that someone made for other leakers. AdoredTV, another leaker I absolutely hate because he is an idiot who makes things up all the time for clicks, got 37% accuracy, and even then they were pretty nice to him because they listed things like "Nvidia will continue to make money a lot of money from GPUs" as a claim he got correct. 

 

 

I really don't get how anyone can still take either of them seriously. 

fanboying.. its one hell of a drug.

 

off topic.

the 800 watt gpus was the server/hpc stuff.

seeing general i think on 3k series and up.

its from the same line. failed hpc dies get turn into consumer gpus.

 

[leadeater]

1 hour ago, dogwitch said:

failed hpc dies get turn into consumer gpus.

They actually don't, the big HPC dies only feature HBM memory controllers and the architecture variant has no video capability at all so they can only function in a compute capacity. A30 are cutdown failed A100 for example. There are also datacenter only architectures like Hopper which will never have consumer or even workstations cards.

 

There are of course lower end datacenter products that share dies between workstation and consumer but those don't really come under Nvidia's designation of HPC GPUs. It's mostly a technicality but Nvidia only labels the Gx100 products as HPC and nothing else.

[LogicalDrm]

Just moderation note. This thread is on the edge of derail which will lead to removal of derailing comments and/or lock. Please keep discussion of validity of some sources as well as general tech discussion to minimum.

[starsmine]

6 hours ago, leadeater said:

FYI

 

CU = AMD

SM = Nvidia

Boy was that a 4am addled brain mistake. I kept thinking CU was a weird way to shorten cuda core but failed to realize why.

[AlwaysFSX]

19 hours ago, leadeater said:

Truth is people want rumors otherwise there is nothing to see or discuss. Due to that people will choose or be far more lenient on mistakes or misinformation because the official public information is nothing at all. How far each person is willing to go on the leniency is obviously different but my standard is if it's not reliable enough, creditable enough or likely to be true to be reported on by general tech media then it's not up to standard and probably wrong.

 

For example how far did that whole 800W 40 series GPU thing go. It was obvious BS, I said it was yet so many wanted argue the point in spite of direct existing evidence that 800W GPU in a PCIE formfactor and cooler isn't viable and nobody is or will ever do that. 600W TDP GPUs are proprietary server formfactor only like SXM and will only ever be at that power level. And who was one of the major sources to push that obviously false information, yea exactly. It was fake the whole time and anyone worth listening to knew it.

Remind me, was there a time on internet forums where people had a little more common sense when it came to the rumormill? I remember reading WCCFTech articles and thinking it's BS and the calls from people on here to ban them as a news source for how varied and untrustworthy they are.

Yet here we are with MLID, lol.

 

I'd keep track of his 'information' for the 50 series myself and see the percentage that turns out to be accurate vs spitballing but I think I'll lose whatever remaining brain cells I even have left.

17 hours ago, starsmine said:

the spreadsheet link was not made well at all. The spreadsheet was wrong half the time aswell. 

MLID says X.

Spreadsheet denotes he said X.

X was false.

Seems like it's working right to me?

[Deadpool2onBlu-Ray]

EVGA is definitely winding down for sure.