Can’t wait for big Navi. :)

That new arctus workstation card maybe.
Or a lower tier wx gpu.

Even if this is Big Navi, Nvidia is obviously ready to launch their next-gen architecture on 7nm as well and I'm sorry AMD fans, but there's no way AMD is going to catch up yet. Big Navi might beat Big Turing, but Turing is on 12FF and is 18 months old now. When Nvidia releases Ampere (or whatever codename they give their next-gen consumer GPU architecture) they will finally take advantage of the 7nm node as well, eliminating this advantage for AMD.

It's going to be another case of "too little, too late" for AMD, I'm afraid.

techguymaxcNvidia is obviously ready to launch their next-gen architecture on 7nm as well

Since it's obvious, you wouldn't mind telling us what is the overwhelming evidence that supports this would you ?

techguymaxcon 7nm

Die size of the 2080TI is a gigantic 754 mm2, such a size at 7nm the costs will be gigantic high towards € 2000+ or so.
Nvidia will not sell that, not even to the biggest Nvidia fans.
So the die-size has to go down considerably and with it the performance. The architecture is the same as Turing. I think it will mainly consume less power and the performance for the 3080TI * will be around 15/20% higher than the 2080TI because architecture tweaks.

5700XT on 7nm has a die-size 251 mm2.
Radeon VII 331 mm2.

Vya DomusSince it's obvious, you wouldn't mind telling us what is the overwhelming evidence that supports this would you ?

His feelings of inadequacy if AMD does well. It's a sickness among Huang boys

Vya DomusSince it's obvious, you wouldn't mind telling us what is the overwhelming evidence that supports this would you ?

Umm Nvidias own promise of new products (not directly confirming Ampere of course) for GTC in March ( i still believe they will launch GA100 first) plus historical evidence of spring/summer releases and one and a half years have passed since turing launch.

LindatjeDie size of the 2080TI is a gigantic 754 mm2, such a size at 7nm the costs will be gigantic high towards € 2000+ or so.
Nvidia will not sell that, not even to the biggest Nvidia fans.
So the die-size has to go down considerably and with it the performance. The architecture is the same as Turing. I think it will mainly consume less power and the performance for the 3080TI * will be around 15/20% higher than the 2080TI because architecture tweaks.

5700XT on 7nm has a die-size 251 mm2.
Radeon VII 331 mm2.

In order to gain speed Nvidia does not have to make any tweaks other than the 7nm node. That's the scary part. They can shrink down TU102 to 300-400mm² and still be faster than AMD just with a process shrink. Besides 7nm has high yields and is very mature now. Just what Nvidia wants for big dies.

Vya DomusSince it's obvious, you wouldn't mind telling us what is the overwhelming evidence that supports this would you ?

Sure.
Fact: Nvidia is beating AMD's fastest chips (on 7nm) with 18 month old, 12nm chips right now.
Fact: The move to 7nm will allow for large improvements to both density and thermals. Think 50%+ faster than Tu102 for the Ampere equivalent.

So if AMD is just catching Tu102 after all this time, what makes you think such a product will be remotely competitive with Am102? That's illogical.

LindatjeDie size of the 2080TI is a gigantic 754 mm2, such a size at 7nm the costs will be gigantic high towards € 2000+ or so.
Nvidia will not sell that, not even to the biggest Nvidia fans.
So the die-size has to go down considerably and with it the performance. The architecture is the same as Turing. I think it will mainly consume less power and the performance for the 3080TI * will be around 15/20% higher than the 2080TI because architecture tweaks.

5700XT on 7nm has a die-size 251 mm2.
Radeon VII 331 mm2.

What makes you think Nvidia will need to design a 751 mm^2 chip on 7nm to compete with Big Navi? Such a chip would have something like 4x the compute power of Tu102. Completely unnecessary, and unprecedented.

There's an epidemic Navi driver issue right now that won't make anyone but the most hardcore of fanboys interested in AMD graphics.

TheGuruStudHis feelings of inadequacy if AMD does well. It's a sickness among Huang boys

Whatever helps you sleep better at night. I'm too old for fanboyism. I buy the best products for my workloads. AMD hasn't offered anything worth my attention in the GPU market since Hawaii. I'm sorry that offends you, perhaps if you weren't so caught up in cheerleading you might not have such an irrational, emotional response to someone offering their opinion about objective matters of engineering.

techguymaxcSure.
Fact: Nvidia is beating AMD's fastest chips (on 7nm) with 18 month old, 12nm chips right now.
Fact: The move to 7nm will allow for large improvements to both density and thermals. Think 50%+ faster than Tu102 for the Ampere equivalent.

Uhm, what's all this supposed to be ? I was expecting some inside information on how Nvidia is supposedly ready to launch anything on 7nm, instead I got some meaningless boilerplate and wishful thinking. I don't know what I was expecting.

"Fact": Nvidia is beating AMD's smallest 7nm chips with colossal 12nm chips.
"Fact" : 7nm doesn't bring any improvement to thermals, it's actually going to get worse as is the case with every node shrink which makes the thermal density increase, so I think you meant power.

Just some details.

techguymaxcSuch a chip would have something like 4x the compute power of Tu102. Completely unnecessary, and unprecedented.

Not only that your "4x" figure is impossible based on current technology projections it's also impossible based on physical limitations. 7nm is 60% denser, so a 7nm Turing based chip of the same size would have roughly 1.6x times the compute power, to get to your mythical "4x" number the SMs would need to become a mindbogglingly 2.5x times faster. What's really unprecedented is your extremely poor estimation and it might just be the most ridiculous thing I read on this forum as of late. Talk about matters of objective engineering ...

I must say that at least you tried.

TomorrowThat's the scary part. They can shrink down TU102 to 300-400mm²

What's scary is that your math is just as bad as above. No, they can't do that, TU102 on 7nm would be about 470mm^2 anything below that is mathematically and physically not possible. 470 mm^2 is still relatively big, comparable to the Pascal predecessor, GP102. And this 60% denser business is the most optimistic figure by the way.

Tomorrowand still be faster than AMD just with a process shrink

And they'd still use chips that are bigger and a lot more costly. This stuff ain't magic, you make a bigger chip on the same node and it's likely going to be faster. Whoop de doo !

techguymaxcEven if this is Big Navi, Nvidia is obviously ready to launch their next-gen architecture on 7nm as well and I'm sorry AMD fans, but there's no way AMD is going to catch up yet. Big Navi might beat Big Turing, but Turing is on 12FF and is 18 months old now. When Nvidia releases Ampere (or whatever codename they give their next-gen consumer GPU architecture) they will finally take advantage of the 7nm node as well, eliminating this advantage for AMD.

It's going to be another case of "too little, too late" for AMD, I'm afraid.

And still way over priced.

I llike that they keep the ATI name throughout :D
I knows names and labels don't mean much in all seriousness, but I still like the feeling :p

Vya DomusUhm, what's all this supposed to be ? I was expecting some inside information on how Nvidia is supposedly ready to launch anything on 7nm, instead I got some meaningless boilerplate and wishful thinking. I don't know what I was expecting.

"Fact": Nvidia is beating AMD's smallest 7nm chips with colossal 12nm chips.
"Fact" : 7nm doesn't bring any improvement to thermals, it's actually going to get worse as is the case with every node shrink which makes the thermal density increase, so I think you meant power.

Just some details.



Not only that your "4x" figure is impossible based on current technology projections it's also impossible based on physical limitations. 7nm is 60% denser, so a 7nm Turing based chip of the same size would have roughly 1.6x times the compute power, to get to your mythical "4x" number the SMs would need to become a mindbogglingly 2.5x times faster. What's really unprecedented is your extremely poor estimation and it might just be the most ridiculous thing I read on this forum as of late. Talk about matters of objective engineering ...

I must say that at least you tried.



What's scary is that your math is just as bad as above. No, they can't do that, TU102 on 7nm would be about 470mm^2 anything below that is mathematically and physically not possible. 470 mm^2 is still relatively big, comparable to the Pascal predecessor, GP102. And this 60% denser business is the most optimistic figure by the way.



And they'd still use chips that are bigger and a lot more costly. This stuff ain't magic, you make a bigger chip on the same node and it's likely going to be faster. Whoop de doo !

TSMC 12FF process node transistor density: 33.8 MTr/mm^2
TSMC 7nm+ EUV process node transistor density: 115.8 MTr/mn^2

www.techcenturion.com/7nm-10nm-14nm-fabrication

If you want to quibble over a few % off 4x scaling have at it, but I’m not interested.

Even if Nvidia has new cards waiting in response, it still means better cards for less money. A 5600 or 5700 would be more than enough for me, but at this point I’m waiting it out to see the next exchange. I sure hope nvidia brings new cards, and does it soon.

LindatjeDie size of the 2080TI is a gigantic 754 mm2, such a size at 7nm the costs will be gigantic high towards € 2000+ or so.
Nvidia will not sell that, not even to the biggest Nvidia fans.
So the die-size has to go down considerably and with it the performance. The architecture is the same as Turing. I think it will mainly consume less power and the performance for the 3080TI * will be around 15/20% higher than the 2080TI because architecture tweaks.

5700XT on 7nm has a die-size 251 mm2.
Radeon VII 331 mm2.

nvidia can cut CUDA core count by half and retain 90% to 95% performance. and they can do it without any IPC increase to the architecture. they have done it in the past. i see many people expect nvidia will be in trouble on 7nm because even with die shrink "turing" will still going to be very big. but most often people did not count what other improvement nvidia can do to turing even without changing the architecture too much

TheGuruStudHis feelings of inadequacy if AMD does well. It's a sickness among Huang boys

the same sickness AMD boy had if they see progress on nvidia.

Bring em on.

It's better that they moved than stay on another GCN Revision, GCN has ran its course

renz496nvidia can cut CUDA core count by half and retain 90% to 95% performance. and they can do it without any IPC increase to the architecture. they have done it in the past. i see many people expect nvidia will be in trouble on 7nm because even with die shrink "turing" will still going to be very big. but most often people did not count what other improvement nvidia can do to turing even without changing the architecture too much


the same sickness AMD boy had if they see progress on nvidia.

Wheres your system specs?

techguymaxcEven if this is Big Navi, Nvidia is obviously ready to launch their next-gen architecture on 7nm as well and I'm sorry AMD fans, but there's no way AMD is going to catch up yet. Big Navi might beat Big Turing, but Turing is on 12FF and is 18 months old now. When Nvidia releases Ampere (or whatever codename they give their next-gen consumer GPU architecture) they will finally take advantage of the 7nm node as well, eliminating this advantage for AMD.

It's going to be another case of "too little, too late" for AMD, I'm afraid.

renz496nvidia can cut CUDA core count by half and retain 90% to 95% performance.

Not all cores are created equal.

Lindatjethe architecture is the same as Turing.

really ?
cause I never seen anything about ampere actually confirmed,let alone the architecture whitepaper released.

Lindatje(...) because architecture tweaks.

:laugh:

Vya Domus"Fact": Nvidia is beating AMD's smallest 7nm chips with colossal 12nm chips.

And yet still has better performance per watt despite having chips twice the size running on an older process. Not bad.

Vya Domus"Fact" : 7nm doesn't bring any improvement to thermals, it's actually going to get worse as is the case with every node shrink which makes the thermal density increase, so I think you meant power.

That's for CPU's. GPU's are much larger and thus thermal density is much less of an issue. No to mention that unlike CPU's they are direct die cooled instead of including an IHS layer that further complicates heat transfer. We are not talking about packing tens of billions of transistiors on 75mm² chiplet here. And on the lower end cards that do use smaller dies it's also not a problem due to their much smaller power consuption.

Vya DomusWhat's scary is that your math is just as bad as above. No, they can't do that, TU102 on 7nm would be about 470mm^2 anything below that is mathematically and physically not possible. 470 mm^2 is still relatively big, comparable to the Pascal predecessor, GP102. And this 60% denser business is the most optimistic figure by the way.

That's for the highest end chip. RTX 2080 is 545mm². Even if we assume a conservative 40% area reduction from moving to 7nm that would produce a ~390mm² card. And thus far AMD can't even beat 2080. Sure they will with big navi but remember that that card wiill likely also be ~500mm² at 7nm. Essentially two times 5700 size.

So in essence AMD will beat a 545mm² current card or 390mm² die shirked card with their 500mm² card. But gets beaten by next gen Ampere within months most likely.

that "smallest" chip in action

I hope this is the "Big Navi" I been looking to see what it can do. Just got some eagerness to maybe purchase something new :)

Do you, guys, realize that you are arguing over a non-existent product?
Rumors of "Big Navi" have been debunked some time ago, and "Arcturus" is a made-up name that someone came up with almost 2 years ago and everyone still uses it (even though it's not a thing).
Having a 225W RX5700XT is already a good-enough sign that consumer cards are near the upper TDP cap, so the most we could actually hope for is a 48CU Navi card, maybe on 7nm+ process.
Also, the majority of consumer GPU market segments are covered already, and as with Polaris, there's no need to rush for that 0.01% that'll give up their 2080Ti in favor of marginally faster AMD flagship.

silentbogoDo you, guys, realize that you are arguing over a non-existent product?
Rumors of "Big Navi" have been debunked some time ago, and "Arcturus" is a made-up name that someone came up with almost 2 years ago and everyone still uses it (even though it's not a thing).
Having a 225W RX5700XT is already a good-enough sign that consumer cards are near the upper TDP cap, so the most we could actually hope for is a 48CU Navi card, maybe on 7nm+ process.
Also, the majority of consumer GPU market segments are covered already, and as with Polaris, there's no need to rush for that 0.01% that'll give up their 2080Ti in favor of marginally faster AMD flagship.

Non-existant product that passed certification. Yeah seems really non-existant to me for AMD submit something that does not exist.
I would like to read those debunkings about big navi. Who said it and when. Even Lisa Su said you can expect a 4K capable card this year. And i doubt she meant running 4K 30 or 4K low.
That's as official as it gets. We will learn more next month during AMD's Financial briefing.

Codenames are irrelevant and often wrong. So what if Arcturus is not a name they will use? The product (MI100) still exists.

The only reason why we have a 225W RDNA right now is that it's the first generation. I have no doubt AMD will improve power efficiency with RDNA 2.0. Problably lower clocks and fitting a 80 CU card inside 300W limit is not a big deal. Plus IPC increase from architectural improvements.

techguymaxcSo if AMD is just catching Tu102 after all this time, what makes you think such a product will be remotely competitive with Am102? That's illogical.

Fortunately, some illogical things happen from time to time. Right, Ryzen? We can hope!