Discussion Nvidia Blackwell in Q4-2024 ?

Page 15 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Majcric

Golden Member
May 3, 2011
1,370
37
91
Looks like all the Blackwell talk have got the 4090 founders back in stock at MSRP. (at least time of writing)
 

jpiniero

Lifer
Oct 1, 2010
14,645
5,273
136
NV will still make money hand over fist.

Problem is, the actual substance behind AI might only be a handful of companies who can make their own chips. And even then, the public should lose interest once it becomes obvious that ChatGPT and it's ilk are going to be censored to uselessness.
 

Mopetar

Diamond Member
Jan 31, 2011
7,870
6,103
136
If all the big companies are censoring their own AIs that just gives you all the more reason to train your own.

"The more you buy, the more you save [freedom]!"
 

Aapje

Golden Member
Mar 21, 2022
1,395
1,885
106
And even then, the public should lose interest once it becomes obvious that ChatGPT and it's ilk are going to be censored to uselessness.

They'll just keep their level of disinformation at a level that most people will buy into it, just like the media does.

For most people, thinking for themselves is the most scary thing they can imagine anyway...after spiders.
 

Ajay

Lifer
Jan 8, 2001
15,501
7,889
136
Problem is, the actual substance behind AI might only be a handful of companies who can make their own chips. And even then, the public should lose interest once it becomes obvious that ChatGPT and it's ilk are going to be censored to uselessness.
LLMs have been around for a while - they've just gotten bigger. Most researchers now categorize current AI as 'weak AI'. It's really just more advanced ML with more depth, nodes and larger input and output (verification) databases. So, the Hype Machine is in overdrive - because it will drive stock prices and investment more than anything else. This is really all backend stuff right now and more nascent than anyone will admit. The migration towards the client side will take allot longer before it is actually more useful than traditional algorithms. Ultimately, all the building blocks are in place to make a substantial change to computing over the next 3 to 5 years. Getting in now, for those handful of companies, is key to becoming part of that future. All IMHO.



Still, wish I bought NV stock when H100 came out. I'll just add it to the list of times my excessive concerns over liquidity overrode my best investment ideas.
 

jpiniero

Lifer
Oct 1, 2010
14,645
5,273
136

Jensen is doing a keynote on June 2nd, the night before Computex. I assume this is mainly to mess with AMD over AI AI AI... but I could see some gaming products announced too.

Not even joking that I am almost expecting them to call Blackwell Gaming "RTX AI"
 

Mahboi

Senior member
Apr 4, 2024
342
580
91
Jensen is doing a keynote on June 2nd, the night before Computex. I assume this is mainly to mess with AMD over AI AI AI... but I could see some gaming products announced too.
They really must be scared somewhere in there.
First that ridiculous article trying to pan MI300X that got a serious backslap from AMD.
Second rushing Blackwell out to respond to MI300X.
Now trying to do the same thing they did with FSR 3, release DLSS 3.5 early to get people away from looking at AMD.

Something at Nvidia is stressed out. And I suspect it's Jensen and the brass.
 

MoogleW

Member
May 1, 2022
56
28
61

Jensen is doing a keynote on June 2nd, the night before Computex. I assume this is mainly to mess with AMD over AI AI AI... but I could see some gaming products announced too.

Not even joking that I am almost expecting them to call Blackwell Gaming "RTX AI"
AI has been part of RTX branding from the original presentation in 2018
 

MoogleW

Member
May 1, 2022
56
28
61
They really must be scared somewhere in there.
First that ridiculous article trying to pan MI300X that got a serious backslap from AMD.
Second rushing Blackwell out to respond to MI300X.
Now trying to do the same thing they did with FSR 3, release DLSS 3.5 early to get people away from looking at AMD.

Something at Nvidia is stressed out. And I suspect it's Jensen and the brass.
I still have no idea why people think Blackwell is rushed. Based on what? Nvidia has presented and released every two years for ages now.

As for DLSS3.5, its not like fsr3 is not a response to DLSS 3 (makes more sense to me to call that the panic response but you do you) in the first place, thats how progress works. Ray reconstruction is so different from anything AMD offers even now that what you are saying makes no sense to me.

To be honest, I eally don't get the overall idea behind how incompetent Nvidia is assumed to be but not usually proven to be. Kind of how Ada Lovelace was also rumored to be a late release and Nvidia was said to be struggling to get MCM woking at all, unlike AMD or even Intel for that matter
 
Reactions: xpea

Mahboi

Senior member
Apr 4, 2024
342
580
91
I still have no idea why people think Blackwell is rushed. Based on what? Nvidia has presented and released every two years for ages now.
Worded that wrong.
The Blackwell announcement is rushed. MI300X comes out Jan 2024, Blackwell gets announced late Q1, while the release is probably for more than 6 months later and volume in 2025.
They aren't rushing Blackwell's development, but they're definitely aggressively trying to get people's attention away from MI300X.
As for DLSS3.5, its not like fsr3 is not a response to DLSS 3 (makes more sense to me to call that the panic response but you do you) in the first place, thats how progress works.
Panic response? A year later?
Or you mean the announcement was ridiculously early?
Ray reconstruction is so different from anything AMD offers even now that what you are saying makes no sense to me.
Oh no, we're starting to enter into the NV mind worms.
"rrrrrrrrrrrrray rrrrrrrrrreconstruction" is a fat joke where NV took a handmade denoiser and replaced it with an AI denoiser.
It's not magical, it's not highly efficient (actually like most AI stuff it is highly INefficient) and broadly speaking is a nice to have, at best.
I'm not going to entertain a conversation that starts with "ray reconstruction is so different from anything AMD offers". Might as well argue philosophy with a cow.
To be honest, I eally don't get the overall idea behind how incompetent Nvidia is assumed to be but not usually proven to be. Kind of how Ada Lovelace was also rumored to be a late release and Nvidia was said to be struggling to get MCM woking at all, unlike AMD or even Intel for that matter
No idea where you got that fantasy and I said nothing of the sort.

Seems to me like you're spending too much time listening to Nvidia propaganda.
 
Reactions: Tlh97

MoogleW

Member
May 1, 2022
56
28
61
Worded that wrong.
The Blackwell announcement is rushed. MI300X comes out Jan 2024, Blackwell gets announced late Q1, while the release is probably for more than 6 months later and volume in 2025.
They aren't rushing Blackwell's development, but they're definitely aggressively trying to get people's attention away from MI300X.

Panic response? A year later?
Or you mean the announcement was ridiculously early?

Oh no, we're starting to enter into the NV mind worms.
"rrrrrrrrrrrrray rrrrrrrrrreconstruction" is a fat joke where NV took a handmade denoiser and replaced it with an AI denoiser.
It's not magical, it's not highly efficient (actually like most AI stuff it is highly INefficient) and broadly speaking is a nice to have, at best.
I'm not going to entertain a conversation that starts with "ray reconstruction is so different from anything AMD offers". Might as well argue philosophy with a cow.

No idea where you got that fantasy and I said nothing of the sort.

Seems to me like you're spending too much time listening to Nvidia propaganda.
Its not a fantasy, its everywhere. 'The next gen gaming is delayed, but RDNA5 is a year early'. '4090 will b slower and use 600W despite having more SM than RDNA3'. We'll be lucky to get 30% out of next gen. Happens all the time.

I do find it weird that Nvidia goes HVM or both client and desktops at same time despite such a far apart announcements, happened 2 years ago as well with Hopper, if I am not mistaken.

As for fsr3, it took many months before it actually appeared anywhere, much more likely to be a panic respons than ray reconstruction. What's the AMD equivalent? Make it make sense. Its not magical, tech is magic only if you don't understand it. It doesn't have to be, it just have to wok well enough. Just like supe resolution DLSS2 compared to the stuIts not magical, tech is magic only if you don't understand it. It doesn't have to be.Its not magical, tech is magic only if you don't understand it. It doesn't have to be. that existed beIts not magical, tech is magic only if you don't understand it. It doesn't have to be.ore it, Bicubic, checkerboard, etc.

Yeah, hurr durr propaganda, Nvidia sucks. But that kinda thing makes this thread less fun than the RDNA 4 thread. I honestly enjoy going through it.
 
Last edited:

Mahboi

Senior member
Apr 4, 2024
342
580
91
Its not a fantasy, its everywhere. 'The next gen gaming is delayed, but RDNA5 is a year early'.
RDNA 5 is not a year early.
'4090 will b slower and use 600W despite having more SM than RDNA3'.
RDNA 3 tops at 96 CUs.
4090 has 120 SMs and can go to 600W.

And the last time anyone said RDNA 3 would beat Lovelace was over 2 years ago, so maybe mend that record and stop repeating yourself?
There isn't a person that's saying RDNA 3 as is is beating Lovelace since it has the wattage problem that it has. RDNA 3.5 could probably do it but it'll only go into APUs that top at around 40CUs, so who cares what it could do?
We'll be lucky to get 30% out of next gen. Happens all the time.
What "all the time" are you even on about?
As for fsr3, it took many months before it actually appeared anywhere, much more likely to be a panic respons than ray reconstruction.
Oh my Lord. Where to begin.

One, FSR 3 was announced a whopping entire year before, so there is definitely no panic there.
Two, it came out in 2 test games pretty much immediately, almost unexpectedly, and didn't start being commonplace in more games until several more months because AMD used these two games as a beta. It came out with unfinished frame pacing and several VRR problems, and that got fixed over the next 2-3 months. That's called an open beta. Nothing panicky about it.
Three, FSR 3 responded to nothing specific at the time. DLSS 3 had come an entire year prior. Whether they responded with an open beta in September or a final product at the end of December didn't change anything, they were a year behind either way.

"Ray Reconstruction", that meme, was announced and released out of nowhere a month after the FSR 3 beta/first outing. It was unfinished as well, which is not common from Nvidia. AMD doesn't care about releasing unfinished software, as their "finewine" reputation says. They just fix the thing over time. Nvidia likes their image of all their stuff being perfect and The Best no matter what. It's strange that Nvidia would release RR in an unfinished state. It's strange that they would release it at all, when there were no promises to keep, no announcements prior, nobody expecting it. And it's strange that they would release it when people started asking interesting questions, like:
- why "can't" DLSS 3 work on Ampere but FSR 3 can, and isn't doing any worse than on Lovelace?
- why is DLSS 3 needing that "bespoke Lovelace hardware" but FSR 3 runs on every single card from AMD or Nvidia?

So before they had to answer those questions, they released their AI denoiser early to keep people looking away from their lies, and to not give AMD any attention. Same as they did with Blackwell announcement as MI300X starts getting bought here and there.
What's the AMD equivalent? Make it make sense.
Literally any current high quality denoiser. As I said, AI denoising has nothing extraordinary.
I'm not the one that needs to start making sense here.
All you are doing is hur durr propaganda, Nvidia sucks.
Literally responded to everything you said, despite you making little to no sense in that tangled narrative of yours. While you write stuff like:
"What's the AMD equivalent to denoising" and I respond...denoising.
But that's what to expect when this thread is so inferior to the RDNA 4 thread. I honestly enjoy goin through it.
Good that you enjoy yourself, if only I could say the same talking to you.
 
Reactions: Tlh97

jpiniero

Lifer
Oct 1, 2010
14,645
5,273
136
Worded that wrong.
The Blackwell announcement is rushed. MI300X comes out Jan 2024, Blackwell gets announced late Q1, while the release is probably for more than 6 months later and volume in 2025.
They aren't rushing Blackwell's development, but they're definitely aggressively trying to get people's attention away from MI300X.

This is true but NV's been announcing their server products way early for some time now.
 
Reactions: MoogleW

MoogleW

Member
May 1, 2022
56
28
61
RDNA 5 is not a year early.

RDNA 3 tops at 96 CUs.
4090 has 120 SMs and can go to 600W.

And the last time anyone said RDNA 3 would beat Lovelace was over 2 years ago, so maybe mend that record and stop repeating yourself?
There isn't a person that's saying RDNA 3 as is is beating Lovelace since it has the wattage problem that it has. RDNA 3.5 could probably do it but it'll only go into APUs that top at around 40CUs, so who cares what it could do?

What "all the time" are you even on about?

Oh my Lord. Where to begin.

One, FSR 3 was announced a whopping entire year before, so there is definitely no panic there.
Two, it came out in 2 test games pretty much immediately, almost unexpectedly, and didn't start being commonplace in more games until several more months because AMD used these two games as a beta. It came out with unfinished frame pacing and several VRR problems, and that got fixed over the next 2-3 months. That's called an open beta. Nothing panicky about it.
Three, FSR 3 responded to nothing specific at the time. DLSS 3 had come an entire year prior. Whether they responded with an open beta in September or a final product at the end of December didn't change anything, they were a year behind either way.

"Ray Reconstruction", that meme, was announced and released out of nowhere a month after the FSR 3 beta/first outing. It was unfinished as well, which is not common from Nvidia. AMD doesn't care about releasing unfinished software, as their "finewine" reputation says. They just fix the thing over time. Nvidia likes their image of all their stuff being perfect and The Best no matter what. It's strange that Nvidia would release RR in an unfinished state. It's strange that they would release it at all, when there were no promises to keep, no announcements prior, nobody expecting it. And it's strange that they would release it when people started asking interesting questions, like:
- why "can't" DLSS 3 work on Ampere but FSR 3 can, and isn't doing any worse than on Lovelace?
- why is DLSS 3 needing that "bespoke Lovelace hardware" but FSR 3 runs on every single card from AMD or Nvidia?

So before they had to answer those questions, they released their AI denoiser early to keep people looking away from their lies, and to not give AMD any attention. Same as they did with Blackwell announcement as MI300X starts getting bought here and there.

Literally any current high quality denoiser. As I said, AI denoising has nothing extraordinary.
I'm not the one that needs to start making sense here.

Literally responded to everything you said, despite you making little to no sense in that tangled narrative of yours. While you write stuff like:
"What's the AMD equivalent to denoising" and I respond...denoising.

Good that you enjoy yourself, if only I could say the same talking to you.
What tangled narrative? Which AMD denoising? Hand crafted denoising techniques have existed before and online you can see the ones AMD and Nvidia show off. DLSS 3.5 is just a continuation of Nvidia research to offload onto AI as usual and asserting 5 times that it exists because of FSR3 which is nothing like it is not a good case for your assertion

RDNA 5 if it comes it in 2025 is early. AMD is not in the current habit of releasing a gen after 1 year and the rumors that say its coming out i 2025 pretty much boil it down to RDNA 4 being a stop gap. So that means AMD is then rumored to do two gens in 2 years. Pretty impressive if true, but harder to believe for me than that Nvidia is needing to rush out Blackwell for consumers to lanch in 2024, after 2 years since they launched Ada Lovelace, despite them consistently delivering a new gen every 2 years or less. So if people insist that there will be 2 AMD gens in 2 years but Nvidia needs to delay one GPU gen to 3 years at the same time, its pretty weird. But same exact rumors happened 2 years ago. rtx 40 was rumored late and rx 7000 was rumored early.

"Same as they did with Blackwell announcement as MI300X starts getting bought here and there." Looking at AMD sales forecast and what Nvidia does all the time, you need something more substantial to convince me that Nvidia continuing their modus operandi of announce HPC now, release months later just before announcing gaming is any different from 2 years ago which I already stated they did the exact same thing. Maybe if the sales were triple what I am expecting from AMD or there were MLPerformance benchmarks showing any need to panic I might be more convinced.

I am just rejecting some ideas we don’t share, like how they can’t even get a GPU gen out the door in the usual time (or are too greedy to do so). I guess you don't like arguing with me to convince me that's the case. Tough cookies.
 
Last edited:

Mahboi

Senior member
Apr 4, 2024
342
580
91
What tangled narrative?
Are you roleplaying Solid Snake and repeating my question, or just not worth the time I spend responding to you?
Which AMD denoising?
Any high quality denoiser.
Hand crafted denoising techniques have existed before and online you can see the ones AMD and Nvidia show off.
"Hand crafted denoising techniques"
That is the most ridiculously convoluted way of saying "denoising".
AI denoising offers a makeshift, low cost solution that demands a ton of computational power on use. Like everything AI.
Spend less on dev time, run ML on the thing and then use inference at a much higher computational cost than an algorithm.

It's all there is under that fancy name "ray reconstruction" that Nvidia sold to you.
DLSS 3.5 is just a continuation of Nvidia research to offload onto AI as usual and asserting 5 times that it exists because of FSR3 which is nothing like it is not a good case for your assertion
Ok you are apparently unable to follow so I will make it simple:

FSR 3 comes out
Works on non-Lovelace cards
People with brains may start asking why
Nvidia releases DLSS 3.5 early so people talk about something else
People with brains notice that it's unfinished and that's not typical of Nvidia

People with brains noticed the pattern, but apparently even when it's explained to you, you can't see it. Wonder why.
RDNA 5 if it comes it in 2025 is early. AMD is not in the current habit of releasing a gen after 1 year and the rumors that say its coming out i 2025 pretty much boil it down to RDNA 4 being a stop gap. So that means AMD is then rumored to do two gens in 2 years. Pretty impressive if true, but harder to believe for me than that Nvidia is needing to rush out Blackwell for consumers to lanch in 2024, after 2 years since they launched Ada Lovelace, despite them consistently delivering a new gen every 2 years or less.
It's not 2025.
So if people insist that there will be 2 AMD gens in 2 years but Nvidia needs to delay one GPU gen to 3 years at the same time, its pretty weird.


But same exact rumors happened 2 years ago. rtx 40 was rumored late and rx 7000 was rumored early.

"Same as they did with Blackwell announcement as MI300X starts getting bought here and there." Looking at AMD sales forecast and what Nvidia does all the time, you need something more substantial to convince me
You need something substantial to enter your head alright.
that Nvidia continuing their modus operandi of announce HPC now, release months later just before announcing gaming is any different from 2 years ago which I already stated they did the exact same thing.
What did I just read?
Maybe if the sales were triple what I am expecting from AMD or there were MLPerformance benchmarks showing any need to panic I might be more convinced.
"muh benchmark thing is the only thing that matters"

Sounds like you'd enjoy Userbenchmark.

Also there was that very memorable bit of Nvidia trying to take the piss out of MI300X in a ridiculous "article" and getting their mouth slapped back shut ultra hard by AMD within 2 days. That was something. I guess you didn't pay attention.
I am just rejecting some ideas we don’t share
Like English?
, like how they can’t even get a GPU gen out the door in the usual time (or are too greedy to do so). I guess you don't like arguing with me to convince me that's the case. Tough cookies.
I don't really care to convince you, you can spend your entire money in Nvidia cards for all I care.
Unlike what Daddy Jensen told you, you are not very important by virtue of being an Nvidia fan.
Tough cookies as you say.



using multiple quotes to make multiple insults?
Take a couple days off.


esquared
Anandtech Forum Director
 
Last edited by a moderator:
Reactions: Tlh97
Jul 27, 2020
16,475
10,500
106
If you had any doubt what the focus of Blackwell is going to be...
Yeah Jensen would jam up a GPU in everyone's arse if he could and make them pay for it for the rest of their lives..

"But your life will be so fulfilling and beautiful with it! There is no reason to live without it! Buy now and live forever!"
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |