Former Nouveau Lead Developer Joins NVIDIA, Continues Working On Open-Source Driver

[Panix]

https://en.wikipedia.org/wiki/RDNA_3#Navi_3x_dies

RX 7800 XT is Navi 32 Die

Panix there is are a few basic rule of silicon production.
1) Large there die size the higher the defect rate.
2) newer the silicon production method also higher the defect rate.

If you look at the 3 RDNA_3 Dies the Navi 31 is the most likely to be defective. Make a defective die usable you use a process to disable parts of the die.

There is another sign they are recovery.
RX 7900 GRE (4×)
RX 7900 XT (5×)
RX 7900 XTX (6×)
Note the MCD counts. Then you see the chip with heat sink removed on RX 7900 GRE different locations out of 6 MCD are populated. You see the same thing with the RX7900XT

Note RX7900 has the same MCD count as the RX7800

What do you do you have produced a stack of Navi 31 dies not good enough to be RX7900 XT or RX7900XTX and you did not leave a number space.

AMD already had what they had planned as the RX 6800 XT successor the RX 7800 XT so that statement that the RX7900 GRE is the planned RX6800 XT success is absolutely false.

I would say AMD hoped they could dump a stack of defective silicon on the Chinese market with is very lax consumer protection laws with the RX 7900 GRE. Now they are putting in other markets because opps they have had that many defective Navi 31 die that is too many for the china market to take up.

Do note it something AMD has done more than once have a opps with silicon production of something and dump the cut down version on china.

Something in the china awkwardly position in the AMD stack be it GPU or CPU is normally limit batches of rejects. Like the 8 core AM4 chips instead of 1 full CCD had 2 half CCD.

Silicon production is a horrible game. Marketing from AMD and placement in China really amd they were expect RX 7900 GRE to be a limited run of duds. Remember AMD has to pay storage fees to keep not used die or destruction fees to dispose of them.

Something you miss running the lazer on the Navi31 die to disable the functionality to turn a RX7900XT/RX7900XTX die into RX7900GRE die is not a zero cost operation and is not a 100 percent yield operation..

Something else to remember RX7000 series is the first time AMD uses MCD prior to RX7000 series everything is monolithic. So this is new tech. Lets say AMD was projecting that at worst only 1 of the connection to MCD from Navi 31 die would not work but now they have stacks of Navi31 dies where 2 MCD connections don't work.

Occam's razor say the simplest explain to RX 7900 GRE existence is the production of Navi31 dies produced more duds than expected and RX 7900 GRE exists to use cut back dies back to make them usable. Cutting the RX7900 die bake makes it conflict with Radeon RX 7800 XT. But what are going to do if you have dud chips stacking up.

Yes attempt to hide the duds in china. Then when you have more duds than what china market can take release it on the global market. AMD and Intel have done this with CPUs for decades. AMD with GPU commonly don't do this.

I thought about it and you are right - it's a difference but I still don't accept AMD using their defective material and then selling it as a 'new card' and trying to slot it in as some 7900 series. Others think it's fine and that's their/your prerogative. I wouldn't buy one. I also wouldn't buy a 4060 or 4060 8gb either. I have looked at the 4070s too - and I'm only interested in a 4070 Ti Super because, at least, you get the max. vram it can get as of now.

I don't like AMD's offerings for 2 simple reasons: 1) they are not power efficient (compared to Ada Lovelace and 2) they are not good at the software I want to run - meaning AMD doesn't sufficiently support their products - and I know about all the excuses but the facts are the facts.

As for rebadging or 'tricking' customers with these 'bait and switch' or gimping - there's ADATA with their ssds, I think others have been accused of doing these kinds of things - and the rebadge by Nvidia, I have read a bit about it but wasn't aware that AMD hasn't.

What about this, then?:

https://www.tweaktown.com/news/27738/nvidia_plays_catch_up_on_gpu_rebadging_wave_hello_to_the_geforce_gt_730m_and_geforce_710m_parts/index.html

https://www.anandtech.com/show/6570/amds-annual-gpu-rebadge-radeon-hd-8000-series-for-oems

Is it so far back, it doesn't count?

https://www.techspot.com/news/81495-amd-has-rebadged-their-budget-gpus-again-now.html

[oiaohm]

I thought about it and you are right - it's a difference but I still don't accept AMD using their defective material and then selling it as a 'new card' and trying to slot it in as some 7900 series.

One of the realities is that a 7900 TX is a defective 7800 XTX. This is just how silicon binning works.

https://en.wikipedia.org/wiki/GeForce_30_series#Desktop

Nvidia has tones of examples that clearly show that they will stick different silicon under the same Retail name. The reality is Nvidia in AMD boat leading to AMD release the RX7900 GRE would have just branded it a RX 7800 and sold it at a higher price(nvidia done this back in the 10 series).

and I'm only interested in a 4070 Ti Super because, at least, you get the max. vram it can get as of now.

At this point the 4070 ti Super is only one bit of silicon but as you can see by 10,20,30 series before it over time its likely to come multi silicon.

As for rebadging or 'tricking' customers with these 'bait and switch' or gimping - there's ADATA with their ssds, I think others have been accused of doing these kinds of things - and the rebadge by Nvidia, I have read a bit about it but wasn't aware that AMD hasn't.

ADATA and Nvidia do the same thing very commonly. We have X brand but we then ship multi different silicon under the same brand so that a customer bought X brand one year and it perform Y and they buy the same X brand the next year and it performs Z because they are different silicon.

I am not saying AMD branding old silicon designs with a new brand is on the up and up but there is one reality this is not what ADATA and Nvidia is doing. You buy a card with X brand on it from AMD it will have same silicon die on side no matter the time you buy it. AMD mandates policy different silicon different consumer brand name.

Panix there are two different behaviors.
1) branding old silicon as new.
2) stuffing multi different silicons including possible different generations under the same retail brand.

AMD only guilty of behavior 1. The 7900GRE is what happens when AMD has not left enough space in the numbers to allow on the production failures to put to market. 7900GRE still a different retail branding.
Nvidia is guilty of both 1 and 2.
ADATA is guilty of 2. ADATA has never done 1.

I class 2 as worse because consumer buys the item it works perfectly then decide they will buy the same one again but now when the use the same brand name again they get something that behaves vastly different. 2 also makes it when a person reviews the product and provided buy with advice not trust worthy due to the issue action 2 causes.

The reality is does not matter if it AMD/Intel/Nvidia/any major silicon chip supplier all of them will sell you defective material because they have to so they can make profit due to how hard it is to get silicon production to produce decent perfect yield. The important point is how they brand this stuff so they don't confuse consumers.

Action 2 is really 'bait and switch' . ADATA SSD were very bad reviewers testing with one make of SSD controller but the consumers were getting a completely different SSD controller under the same retail brand with complete different performance.

AMD actions by at least Australian law where I am is not legally bate and switch.
Action 1 if it illegal in Australia would fall under misleading claims.
Action 2 ADATA/Nvidia branding the different silicon under the same name is possible bate and switch and this comes under false claims if illegal Australia.

Panix both are bad for consumers in different ways most countries laws class them as different things. So it does not pay to mix them with each other otherwise you are not being clear to your regulator that they should do anything.

Yes the RX600 and the HD8000 series have a problem that should not be allowed. Reviews are upset by what AMD does and talks about it because it causes them more work. But there is worse like the GeForce RTX 3060 Ti where there is 3 different silicons and reviewers normally only put one in the charts. This where the bate and switch with Nvidia comes in. You are advertised X but you are getting Y in the Nvidia case.

[tenchrio]

So, what? The 1030 and gt 710 are old cards - they're not even gaming cards. The 7900 GRE is in the generation's upper tier cards. Also, my point is, fools like you make excuses for AMD.

The other guy is ignoring these points, too - he's also an AMD fanboy but at least his argument is mostly tech-related. You have bias and you're a tool who is obnoxious. Go find a corner.

No the point is you are absolutely too foolish to understand the difference, you still don't get it. Also now all of a sudden gaming is relevant You truly are a hypocrite. The fact they are old cards means that open source drivers are the best bet to keep them usable but due to circumstances created by Nvidia there is a problem with where you basically need to be lucky enough to have 1 variant over the other without ever having had the choice. The cards have the exact same name but different chips and capabilities which stop people from using them fully with open source drivers (it really isn't complicated unless you are insanely dumb ).

We aren't ignoring your points, we are calling them what they are: irrelevant (and also poorly constructed, having near 7800XT performance doesn't mean it has to be called a 7800 card, the branding names are marketing anyway such a childish thing to bring up).

But then you come barging in here again, like the biased obnoxious child that you are and bring up a completely unrelated AMD card and go "bUt AmD iS bAd tOo". And the second anyone even slightly disagrees with you, you cry and bring up incredible bad research with poor reasoning and start calling everyone AMD fanboys, fools and biased (while demonstrating how little you actually know on the subject) because the truth is that you are just an AMD hater that can't live with the fact that Nvidia is being criticized on something so you need to bring up AMD for your own delusional satisfaction, regardless if it is on topic or not and incapable of admitting there is a difference so you result to name calling.

Despite oiaohm treating you with respect, the same I once did, you feel a need to name call him too but you don't deserve it because you always name call and can never admit you are in the wrong, it is just laughable, you are Dunning-Kruger incarnate. It shows how much you hate AMD that you must bash them and keep trying to defend that your unrelated hate for the 7900GRE is exactly the same as our criticism of Nvidia splitting silicon like that which combined with their denying of automatic reclock for a lot of architectures that are EOL or near EOL makes for a very confusing list of what is and isn't fully supported by the open source drivers which is something Nvidia can fix by giving us literally anything (firmware, source cod etc) to enable automatic reclocking for those architectures, it really is an Nvidia only issue.

You still don't understand the naming, performance and all that other shit you cry about the 7900GRE is irrelevant, even children understand: you consistently know what you are getting with that card even if it is a cut down 7900 XT/XTX with the performance equivalence of a 7800 XT.
Let me break it down in the simplest terms possible:
7900GRE VS 7900XT VS 7800XT:

• Small but noticeable difference in name, noticeable on the box even
• Difference in pricing
• You know what you get even if it isn't great value compared to others

GT710M VS GT710M :

• No difference in name
• One requires automatic reclock, the others does not (which until Nvidia gives us a solution, heavily impacts the usage of open source drivers)
• You don't know which you have until you check it

[oiaohm]

• No difference in name
• One requires automatic reclock, the others does not (which until Nvidia gives us a solution, heavily impacts the usage of open source drivers)
• You don't know which you have until you check it

GT710M are even worse. Lot of ODM shipping GT710M used the same box for both forms. So nothing on the outside the the packaging you could have read in the store would have told you any difference. And the soc inside is from two completely different design generations.

Even up to the 40 and 30 series of Nvidia cards not all Nvidia ODM put on the outside of the box what nvidia chip/chips are inside card and this can even apply to the specification sheet from the ODM.

Yes GeForce RTX 3060 Ti has 3 possible chips inside. Yes you can have 4 boxes absolute identical that contain the 3 different chipped cards yes there is 4 cards here.

https://en.wikipedia.org/wiki/GeForce_30_series#Desktop One behaves very differently out of the 4 possible cards that have the GeForce RTX 3060 Ti because it only has 9.5GT/s of memory instead of the other 3 that have 14Gt/s. Yes the GA104-202 chip GeForce RTX 3060 Ti can be populated with 2 types of ram.

Yes GeForce RTX 3060 Ti is 4 cards per ODM 3 different soc chips and 2 different memory types. Yes this is cards released from Dec 2, 2020 to Oct 2022. So if you bought early Geforce RTX 3060Ti then bought another one a year and a bit latter it was different silicon. You might have hit the one with the slower transfer speed and wonder why this card did not behave quite right.

GT710M is an old example. We have example that less than 4 years old.

GeForce RTX 4060 Ti in the 40 series is not too bad because 1 bit of silicon is the 4G card and the other bit of silicon is the 16G card. Yes it another 2 bits of silicon one sales name.

If Nvidia sticks to form as Geforce 40 series ages we are going to have more cards that have multi silicon under the same retail sales name.

Yes it not fun to have to tell a person who bought a GeForce RTX 3060 Ti that they bought the 1 version of it that absolute suxes due to poor memory transfer speed and this is why their card cannot perform like the on-line reviews of the GeForce RTX 3060 Ti because the reviewers are using 1 of the other 3 other card options. These poor users are truly getting bated and switched.

There is example like this in 20 series and 10 series as well I know of. I don't know why we tolerate Nvidia doing these bate and switches.

GT710M is a really bad historic case because the silicon in the 2 different forms are from 2 completely different generations of Nvidia silicon design. But that does not make the ones where they are inside the same generation of silicon not bate and switch when there are quite major differences in performance.

[Panix]

No the point is you are absolutely too foolish to understand the difference, you still don't get it. Also now all of a sudden gaming is relevant You truly are a hypocrite. The fact they are old cards means that open source drivers are the best bet to keep them usable but due to circumstances created by Nvidia there is a problem with where you basically need to be lucky enough to have 1 variant over the other without ever having had the choice. The cards have the exact same name but different chips and capabilities which stop people from using them fully with open source drivers (it really isn't complicated unless you are insanely dumb ).

We aren't ignoring your points, we are calling them what they are: irrelevant (and also poorly constructed, having near 7800XT performance doesn't mean it has to be called a 7800 card, the branding names are marketing anyway such a childish thing to bring up).

But then you come barging in here again, like the biased obnoxious child that you are and bring up a completely unrelated AMD card and go "bUt AmD iS bAd tOo". And the second anyone even slightly disagrees with you, you cry and bring up incredible bad research with poor reasoning and start calling everyone AMD fanboys, fools and biased (while demonstrating how little you actually know on the subject) because the truth is that you are just an AMD hater that can't live with the fact that Nvidia is being criticized on something so you need to bring up AMD for your own delusional satisfaction, regardless if it is on topic or not and incapable of admitting there is a difference so you result to name calling.

Despite oiaohm treating you with respect, the same I once did, you feel a need to name call him too but you don't deserve it because you always name call and can never admit you are in the wrong, it is just laughable, you are Dunning-Kruger incarnate. It shows how much you hate AMD that you must bash them and keep trying to defend that your unrelated hate for the 7900GRE is exactly the same as our criticism of Nvidia splitting silicon like that which combined with their denying of automatic reclock for a lot of architectures that are EOL or near EOL makes for a very confusing list of what is and isn't fully supported by the open source drivers which is something Nvidia can fix by giving us literally anything (firmware, source cod etc) to enable automatic reclocking for those architectures, it really is an Nvidia only issue.

You still don't understand the naming, performance and all that other shit you cry about the 7900GRE is irrelevant, even children understand: you consistently know what you are getting with that card even if it is a cut down 7900 XT/XTX with the performance equivalence of a 7800 XT.
Let me break it down in the simplest terms possible:[/LIST]

Blah, blah, blah, more bs from you so that you can insult - as that is all you do. I agree with oiaohm and prefer discussing with him because he's above insulting unlike you.

I do understand the chip difference but regardless, it doesn't change the fact the 7900 GRE was named under the 7900 series and is heavily gimped - your average consumer probably won't have any idea and might be assuming near-7900 XT performance - which it DOES NOT come close.

The reason I didn't care much about the 710 and 1030 fiasco is because Nvidia is already not really compliant or known for supporting open source anyway - they're already hated for this - so, the chip thing isn't very surprising from them. I shopped for a GT 1030 a while ago when I had an old sff Dell - so, I was aware of what they did with it. Since, I was looking for used card and refused to buy new - partly because it was overly expensive for a small form factor gpu, partly because of what Nvidia DID - so, I was not contributing to Nvidia revenues and profit!

It's basically a gpu to get video, imho - so, it's pretty pathetic they'd cripple that one and to save money further, use the wrong tech/chip with it. But, that's Nvidia? So, yes, the fact it's not a gaming chip - I think it is RELEVANT! More ppl will buy a gaming card so more ppl will likely be buying the 7900 'Rabbit card' - which is not really a 7900 series 'under the hood' - with defective chips, also - we all agree, right? The semantics of arguing it's 'still a 7900' notwithstanding. I already posted a number of ppl discussing it that proven without doubt, it's only a 7900 in name - since, there's a bunch of gimping done because it was silicon they could still savage.

I don't hate AMD - well, maybe somewhat - because they refuse to support their hardware and neglect the software side - they don't support programs in the productivity sphere - so, yeah, they are like Nvidia in that they put profit and shortcuts above a good hardware/good software stack. Their processors are good - if I was buying a new PC or build, I'd definitely pick AM5 today.

https://www.techradar.com/computing/gpu/amd-radeon-rx-7900-gre

https://www.techpowerup.com/311822/amd-radeon-rx-7900-gre-asic-smaller-than-navi-31-slightly-larger-than-navi-21

https://www.tomshardware.com/pc-components/gpus/amd-radeon-rx-7900-gre-review/7

Several links detail the tech differences. I'd give the gpu a hard pass - most ppl just care about gaming performance and the price.

[oiaohm]

I do understand the chip difference but regardless, it doesn't change the fact the 7900 GRE was named under the 7900 series and is heavily gimped - your average consumer probably won't have any idea and might be assuming near-7900 XT performance - which it DOES NOT come close.

Also the 7900GRE AMD has priced as if it a updated 7800TX. So the price tag of the 7900GRE is not close to the 7900XT either.

RX 7900GRE is 599.99 USD card.
RX 7900TX is $936.71 USD card
$936.71 USD card
This is you amazon PowerColor Red Devil Hell hounds.
Had to go to a lower fighter class card to get
PowerColor Fighter AMD Radeon RX 7800 XT is USD $489.99 The fighter is a 100 USD cheaper heatsink than the Red Devils. Recommend retail has 50USD difference between the RX7900GRE and the RX 7800XT of the same type.

RX 7900GRE is performance from the benchmarks is on majority of the time slightly better than you RX7800. You would say the RX7900GRE is priced for the performance it in fact gives.

https://gamersnexus.net/gpus/amd-radeon-rx-7900-gre-gpu-review-benchmarks-vs-rx-7900-xt-7800-xt-rtx-4070-super

The gamersnexus write up is important.

7900 XT 17% faster than the 7900 GRE. But remember you are paying 1/3 less for the GRE. Performance per dollar 7900GRE is better buying than the 7900TX.
7900 GRE a 13% faster than the RX7800TX. At recommend retail price on performance vs price the 7900GRE is better buying than the RX7800TX. Now lot of cases you are not buying at recommend retail price so RX7800TX in some cases could be better buying in performance vs price.

Its kind of surprising for how much is cut out of the 7900GRE that is only a 17% performance loss from the RX7900XT back to it.. Think there is over a 20% between the RX7900TX and the RX7900XTX

In reality the 7900GRE performs good enough to be a historic non TX/XTX card. It would have been a lot worse if AMD had just label it RX7900 and called it done. Yes the card has the performance for that.

You are not thinking that its ~20% performance difference between the TX and the XTX with AMD and that it should be 20-30% back to the not titled card. We have not seen AMD releasing non titled cards in recent years.

Consumers would be more confused if AMD had just released a non titled card after all these years because people would over presume how much untitled card performs.. They created a new title the GRE.

Panix being near china I have seen different AMD cards for only the china market that don't have TX/XTX on them that are the 20-30% slower to the TX versions from older generations of course normally only in small batches(the duds). We have got use to not seeing the duds in markets outside china. Something really gone wrong with the 31 die production has to have had a lot of duds for the GRE to be coming into the Weston market. I say they suspected this case when they started making the GRE card this is why the GRE marking from the start instead of the normal nothing marking.

Yes the navi 31 die is the first time AMD has attempted to mass produce a multi chip GPU.

Yes the 7900GRE performance is lot closer to the 7900TX than one would suspect reading the specification sheet for what has been cut out. Also AMD priced as if RX7900GRE was 30% slower and turned out not to be. .

[Panix]

Also the 7900GRE AMD has priced as if it a updated 7800TX. So the price tag of the 7900GRE is not close to the 7900XT either.

RX 7900GRE is 599.99 USD card.
RX 7900TX is $936.71 USD card
$936.71 USD card
This is you amazon PowerColor Red Devil Hell hounds.
Had to go to a lower fighter class card to get
PowerColor Fighter AMD Radeon RX 7800 XT is USD $489.99 The fighter is a 100 USD cheaper heatsink than the Red Devils. Recommend retail has 50USD difference between the RX7900GRE and the RX 7800XT of the same type.

RX 7900GRE is performance from the benchmarks is on majority of the time slightly better than you RX7800. You would say the RX7900GRE is priced for the performance it in fact gives.

https://gamersnexus.net/gpus/amd-radeon-rx-7900-gre-gpu-review-benchmarks-vs-rx-7900-xt-7800-xt-rtx-4070-super

The gamersnexus write up is important.

7900 XT 17% faster than the 7900 GRE. But remember you are paying 1/3 less for the GRE. Performance per dollar 7900GRE is better buying than the 7900TX.
7900 GRE a 13% faster than the RX7800TX. At recommend retail price on performance vs price the 7900GRE is better buying than the RX7800TX. Now lot of cases you are not buying at recommend retail price so RX7800TX in some cases could be better buying in performance vs price.

Its kind of surprising for how much is cut out of the 7900GRE that is only a 17% performance loss from the RX7900XT back to it.. Think there is over a 20% between the RX7900TX and the RX7900XTX

In reality the 7900GRE performs good enough to be a historic non TX/XTX card. It would have been a lot worse if AMD had just label it RX7900 and called it done. Yes the card has the performance for that.

You are not thinking that its ~20% performance difference between the TX and the XTX with AMD and that it should be 20-30% back to the not titled card. We have not seen AMD releasing non titled cards in recent years.

Consumers would be more confused if AMD had just released a non titled card after all these years because people would over presume how much untitled card performs.. They created a new title the GRE.

Panix being near china I have seen different AMD cards for only the china market that don't have TX/XTX on them that are the 20-30% slower to the TX versions from older generations of course normally only in small batches(the duds). We have got use to not seeing the duds in markets outside china. Something really gone wrong with the 31 die production has to have had a lot of duds for the GRE to be coming into the Weston market. I say they suspected this case when they started making the GRE card this is why the GRE marking from the start instead of the normal nothing marking.

Yes the navi 31 die is the first time AMD has attempted to mass produce a multi chip GPU.

Yes the 7900GRE performance is lot closer to the 7900TX than one would suspect reading the specification sheet for what has been cut out. Also AMD priced as if RX7900GRE was 30% slower and turned out not to be. .

Wrong again, Mr. Oiaohm.... I know you are trying but this is more evidence that AMD just doesn't care about the consumer and is pricing their hardware way too high - and they're just like Nvidia. Everything I've read indicates that the 7900 GRE is geared towards being a 2k or 1440p card. It's going against the 4070 and 4070 Super - one could argue it's a 'better pick' than the 4070 but that can't be said with a 4070 Super comparison.

In my country, the 4070 Super is only $40 more. The cheapest 7900 GRE is $750 and the cheapest 4070 Super is $790.

https://www.techspot.com/review/2826-geforce-rtx-4070-super-vs-radeon-7900-gre/

vs 4070:

https://www.tomshardware.com/pc-components/gpus/rtx-4070-vs-rx-7900-gre-faceoff-which-mainstream-graphics-card-is-better

The 4070 is a bit cheaper, here. Thus, when you compare tasks other than gaming, the 4070 Super is a much better choice. The only disadv. it has is lacking vram in comparison.

This is a typical 'loss' for AMD though - with their gpus. It's the typical pattern.

As for the 7800 xt - $680.

https://tech4gamers.com/rx-7800-xt-vs-rx-7900-gre/

https://www.youtube.com/watch?v=4mFssoxlJHk

It seems like there's around a 7-8 fps gain for the 7900 gre. Anyway, I don't care too much about these 2 cards - they seem mediocre to me - worse power efficiency than the 4070 series cards and weaker, overall performance.

[mrg666]

Wrong again, Mr. Oiaohm.... I know you are trying but this is more evidence that AMD just doesn't care about the consumer and is pricing their hardware way too high - and they're just like Nvidia. Everything I've read indicates that the 7900 GRE is geared towards being a 2k or 1440p card. It's going against the 4070 and 4070 Super - one could argue it's a 'better pick' than the 4070 but that can't be said with a 4070 Super comparison.

In my country, the 4070 Super is only $40 more. The cheapest 7900 GRE is $750 and the cheapest 4070 Super is $790.

https://www.techspot.com/review/2826-geforce-rtx-4070-super-vs-radeon-7900-gre/

vs 4070:

https://www.tomshardware.com/pc-components/gpus/rtx-4070-vs-rx-7900-gre-faceoff-which-mainstream-graphics-card-is-better

The 4070 is a bit cheaper, here. Thus, when you compare tasks other than gaming, the 4070 Super is a much better choice. The only disadv. it has is lacking vram in comparison.

This is a typical 'loss' for AMD though - with their gpus. It's the typical pattern.

As for the 7800 xt - $680.

https://tech4gamers.com/rx-7800-xt-vs-rx-7900-gre/

https://www.youtube.com/watch?v=4mFssoxlJHk

It seems like there's around a 7-8 fps gain for the 7900 gre. Anyway, I don't care too much about these 2 cards - they seem mediocre to me - worse power efficiency than the 4070 series cards and weaker, overall performance.

Well, the videos you linked are for Windows. As a Linux enthusiast, AMD is less trouble and as good (maybe better) performance on Linux. I would not buy Nvidia for Linux as it was a very frustrating experience for me. Just my 2c though, that is your money.

[tenchrio]

Blah, blah, blah, more bs from you so that you can insult - as that is all you do. I agree with oiaohm and prefer discussing with him because he's above insulting unlike you.

LOL so you can give but can't take, typical hyprocite.
And a reminder that you started with the insults once again with your "reading comprehension" and "this site is full of delusional AMD fanboys so no chance of that happening" as you probably already knew you were wrong deep inside. And the best part is now you admit it !! You admit it is different but have to continue arguing about the name of a card while it is unrelated to the topic that was discussed previously, but hey progress I guess.

I do understand the chip difference but regardless, it doesn't change the fact the 7900 GRE was named under the 7900 series and is heavily gimped - your average consumer probably won't have any idea and might be assuming near-7900 XT performance - which it DOES NOT come close.

That was never relevant to this thread, you just randomly brought it up to thread hijack because once again you have to hate on AMD, regardless of the topic, regardless of the news being discussed, you have for some reason a burning desire to bring AMD up and to bash it. The thread here was about the open source nouveau drivers of Nvidia and how particularly confusing it is getting with the automatic re-clocking what is and isn't supported under Nouveau (all thanks to Nvidia being actively in the way of the open source driver) . The card is clearly named differently, how about you using that reading comprehension of yours to read the name.

The reason I didn't care much about the 710 and 1030 fiasco is because Nvidia is already not really compliant or known for supporting open source anyway - they're already hated for this - so, the chip thing isn't very surprising from them.

Except that this discussion right here is about Nvidia slowly given us this compliance at least somewhat for Turing and newer. Which leaves us with a weird spot of Nvidia GPUs that have automatic reclock but don't use the new GSP firmware. We need it for Kepler because it is EOL, we want it for Maxwell as it is reaching EOL hence why it is relevant in this thread about the open source nvidia driver.

I shopped for a GT 1030 a while ago when I had an old sff Dell - so, I was aware of what they did with it. Since, I was looking for used card and refused to buy new - partly because it was overly expensive for a small form factor gpu, partly because of what Nvidia DID - so, I was not contributing to Nvidia revenues and profit!

Okay, cool story.

It's basically a gpu to get video, imho - so, it's pretty pathetic they'd cripple that one and to save money further, use the wrong tech/chip with it. But, that's Nvidia? So, yes, the fact it's not a gaming chip - I think it is RELEVANT! More ppl will buy a gaming card so more ppl will likely be buying the 7900 'Rabbit card' - which is not really a 7900 series 'under the hood' - with defective chips, also - we all agree, right? The semantics of arguing it's 'still a 7900' notwithstanding. I already posted a number of ppl discussing it that proven without doubt, it's only a 7900 in name - since, there's a bunch of gimping done because it was silicon they could still savage.

Oh my god Argumentum ad populum, what you think the earth is also flat and Scientology is real because a bunch of people are telling you that is the case too?
It always amazes me how idiots try to flock together in the face of facts and go "wE cAn'T aLl Be WrOnG!".

As me and oiaohm have been saying the 7900GRE uses the same silicon as the 7900XT but a cut down version to salvage chips that came short.
The xda-developers you quoted even says so that "on paper it is extremely close to the 7900XT" (how about you read your sources for once), just less cores, less buswidth/bandwith and less VRAM. The 7800XT uses NAVI 32 , the 7900 GRE And 7900XT both use NAVI 31 and with Nvidia this naming behavior is also the same the RTX 4070 and 4070 TI both use AD104.... on launch as Nvidia starting March 2024 also began to ship RTX 4070 with a cutdown AD103 (RTX 4080 chips) and just like before you won't know if yours has AD103 or AD104 until you plug it in or take it apart entirely (the effects are also hard to measure as reviewers would have to get lucky and receive both an AD103 and AD104 and again they aren't sold as separate models, they are both put inside of the same RTX 4070 box).

But if you hate the 7900GRE because its performance is less than the other cards that it shares its die with (for some reason) wait until I hit you with this bombshell, Nvidia's 4000 mobile GPUs always use the silicon for one tier lower. So the RTX 4090 (and Dragon) uses AD102 but the RTX 4090 mobile uses AD103, the same as the RTX 4080, the RTX 4080 Mobile however uses AD104, the same one as the RTX 4070 and 4070 TI and this goes on until AD107 where finally the RTX 4050 and RTX 4050 Mobile are together at last (as are the RTX 4060 and 4060 Mobile. Well sort of as just like the RTX 4070 with AD103, nvidia started shipping RTX 4060's with AD106, again you can't know if yours does or not). [s]My god you must be livid!! How can Nvidia do this absolute crime of saying it is a 4080 (mobile) but use the 4070 silicon![/s]
And mind you this wasn't the case last gen where the corresponding mobile GPU did have the same silicon so the RTX 3070 had GA104, same as the RTX 3070 mobile.

Also originally there was a RTX 4080 12GB and if you look at the spec sheet like the AD104 silicon instead of AD103 like the RTX 4080 16GB, the 7680 shading units, the 240... TMUs, the 80 ROPS.... wait a minute that is identical to the RTX 4070 TI specsheet! Nvidia actually tried to sell a "4080" which in reality had 4070 silicon and way lower specs than the 4080 16GB, they then "unlaunched" this card and relaunched it as the 4070 TI which makes more sense since the RTX 4070 also uses AD104. Mind you that historically every xx70 and TI card had a corresponding silicon with 104 (3070 (TI)=> both GA104, 1070 (TI) => both GP104, 970 => GM204, Turing being weird and making the RTX 2070 TU106 but the RTX 2070 Super TU104). So uh Nvidia actually attempted and does what you accuse AMD of (wrongfully since the name stems less from performance and more from the silicon as it has done for ages and man finding out that stuff about the 4000 series mobile GPUs would not have been so hilarious if it wasn't for your ridiculous opinion).

And still: all of this was off topic as the original topic talked about was the state of nouveau drivers and how the GT710M users are in a coin flip predicament if they can use nouveau or must use the nvidia legacy drivers as the card is no longer supported by Nvidia, your terribly informed rant was never related and as usual the premise of it is your own delusion and a far cry from reality for people who actually do have knowledge on GPU manufacturing, naming and silicon.

I don't hate AMD - well, maybe somewhat -

No, you absolute do, you posted in a thread of which the article was already deleted for over a day and despite it being about a performance fix (on APUs but hey the article was deleted, how could you know) you immediately went to bash AMD for it (hell even if it was for GPU, wouldn't that be a good thing? That AMD fixes something that you find important?).That is clearly indicative of an unhealthy obsession. Here is the kicker; when Intel fixed an actual similar issue for their dedicated GPUs this month, you didn't randomly start posting there bashing Intel (or AMD, which wouldn't make sense but hey you are doing it here), despite that article still being up (and thus still accessible through the front page and most importantly readable).

because they refuse to support their hardware and neglect the software side - they don't support programs in the productivity sphere -

Oh god here we go again, I think we had this song and dance like what 4 times already?
I shared the benchmarks of Davinci Resolve and Premiere Pro before, the AMD cards punch well above their price in that department (compared to Nvidia).
So no you are wrong, per usual. AMD is behind with HIP-RT but that is less of "don't support" and more of "in development" and I have already giving you a dozen 3D use case where HIP-RT and Optix performance do not apply. AMD has also developed the current Blender USD Hydra plugin that is shipped with Blender that can be used as a render delegate for more advanced render engines like Dreamworks Moonray or Pixar's Renderman and it is usable for both Nvidia and Intel and as said multiple times before their own render engine that can also be used for Nvidia and Intel, so they do support the productivity sphere and their support of that sphere even benefits Nvidia and Intel (the things you know if you are actually in the productivity sphere).

so, yeah, they are like Nvidia in that they put profit and shortcuts above a good hardware/good software stack. Their processors are good - if I was buying a new PC or build, I'd definitely pick AM5 today.

...what? If there is one thing you need to hand to Nvidia it is that CUDA is a great (albeit be it closed source) software stack, being exceptionally well optimized compared to OpenCL and having good documentation. CUDA being that good is what helped Nvidia achieve its current dominance (not to mention the insane performance/value aspect of the Pascal line during a time when people were taking bets when AMD would be declared bankrupt), but ROCM Is slowly and steadily getting better and in some cases has proven to give a better value already which is great for the consumer, stronger competition leads to better products, just look at the CPU space before and after first gen Ryzen.

Several links detail the tech differences. I'd give the gpu a hard pass - most ppl just care about gaming performance and the price.

There is also a difference between the 7900XT and 7900XTX, and the 4070 and 4070 TI doesn't mean the name does not make sense.
And okay, sure I think most people are of that opinion on that card. Doesn't mean you need to randomly rant about it (or how about you do it in the Phoronix review about it, well you actually already did do that but clearly you need to get it out of your system).

[oiaohm]

The 4070 is a bit cheaper, here. Thus, when you compare tasks other than gaming, the 4070 Super is a much better choice. The only disadv. it has is lacking vram in comparison.

The RX7800TX is a 16GB card RX7900RE 16G card. and the 4070Super is 12G card. Guess what running ram consumes power. The power difference between 4070 Super and RX7800TX and RX7900GRE is mostly ram.

4070 Super is not cheaper everywhere. That 4G of ram difference can be the difference if a compute workload works well or not. 16 G ram card is a lot more expensive from Nvidia.

Linux users like me build systems I run for at least a decade. AMD I will have driver updates for that time frame. Nvidia not so much. So do more Nvidia matter if in 5 years time I am going to have replace the Nvidia card because it drivers don't want to work with the current X11/Wayland solution of Linux by then or by the AMD card that going to be good for the 10+ years.

This is a factor of long term performance is driver support. It common for AMD developers to find something wrong current driver and see that the old driver had the same problem on Linux and fix that has well.

Me building a system to run for a decade I am compare the 7900GRE against the 7800TX the 4070 super from Nvidia does not come into it.