Rumor: NVIDIA GeForce Ampere to be fabbed at 10nm, all cards RTX ?

Denial

2020-03-12 16:46

I don't get it. This guy comes out of no where and just starts posting this stuff on twitter and we believe him, why? You write:

(who has a bit of a reputation for these things)

but what reputation does he have? He has no prior leak history, let alone one that's been shown to be accurate. His account was just created in January. His opinion should weigh just about as much as anyone else on any forum. Why are all the sites looking at this guys post with credibility? I'm not saying he's wrong or right, I just don't get why every single tech news site is looking at this guy like he's god's gift to leaks. And while I think it's fine to post it as a rumor because others are and I understand the need to want to capture that audience - I think at least part of the post should explain why this is just a rumor and not verifying his credibility when as far as I can tell he has none.

#5768861

Astyanax

2020-03-12 16:49

3. Ignores the reality of how RTX works, you can't scale RTX up without scaling up raster too.

#5768863

Kaarme

2020-03-12 16:51

According to this 3060 would have the exact same amount of CUDA cores as 2060 (non-super) and even the same 6GB of memory. I hope these particular rumours aren't the truth.

#5768864

Denial

2020-03-12 16:53

Astyanax:

3. Ignores the reality of how RTX works, you can't scale RTX up without scaling up raster too.

I'm not sure that's entirely accurate. Surely you can scale it that way but you can't tell me there is zero way for them to improve the implementation without increasing the number of RT/SMs. RT seems to touch multiple parts of the chip - there has to be room for optimizations outside of just increasing the number of cores.

#5768866

Hilbert Hagedoorn

Administrator

2020-03-12 16:55

Denial:

I think at least part of the post should explain why this is just a rumor and not verifying his credibility when as far as I can tell he has none.

Actually I was still editing this story and included precisely that.

#5768874

Denial

2020-03-12 17:08

Hilbert Hagedoorn:

Actually I was still editing this story and included precisely that.

Sweet - yeah I don't blame you for posting these - I'm sorry if It came off that way or as condescending. I understand the need to generate revenue and why a breaking story like this in other places needs to be posted here, even if you may disagree with it. I just wish more tech journalists would err on the side of caution with these stories and I don't understand the particular fascination everyone has with this guy. These are the types of rumors that tend to come back two years from now when people go "look nvidia lied they said it would be x% more performance!" Like just the other day a member was posting April Fools slides as proof AMD claimed 4x PERF/W and lied about it because they didn't get it. He later admitted his mistake so it's fine but I see this happening a lot with rumors being taken as fact.

#5768883

Astyanax

2020-03-12 17:30

Denial:

I'm not sure that's entirely accurate. Surely you can scale it that way but you can't tell me there is zero way for them to improve the implementation without increasing the number of RT/SMs. RT seems to touch multiple parts of the chip - there has to be room for optimizations outside of just increasing the number of cores.

you have as many fp32 units as you do int32 units as they are one and the same, just able to do either task in parallel, RT scales on Int count, BVH doesn't scale RT performance directly. You can't raise RTX performance without having a non linear but similar increase to Rasterisation.

#5768891

Hilbert Hagedoorn

Administrator

2020-03-12 17:35

Denial:

Sweet - yeah I don't blame you for posting these - I'm sorry if It came off that way or as condescending.

Not at all and in fact, you bring up a fair point. I was seriously doubting posting this one. It's a balance and judgment call at times as to what to post and what not. But I hope I inserted enough disclaimers now lol.

#5768899

Ricepudding

2020-03-12 18:11

Hilbert Hagedoorn:

Not at all and in fact, you bring up a fair point. I was seriously doubting posting this one. It's a balance and judgment call at times as to what to post and what not. But I hope I inserted enough disclaimers now lol.

I think posting these, specially if you do put "rumor" at the front is fine, it gives us all something to discuss and at the end we can what rumours were right... could be good to go back to these and be like hey this person was right about x y and z so maybe they know more inside stuff than we thought. They should always be taken with a grain of salt, but it is very interesting to see these come out. Though if this rumor is true i might be a little sad as it seems to be weaker than i expected it to be, was hoping for more of a 50-60% jump, not the barely 40% being shown here

#5768927

Reddoguk

2020-03-12 19:35

We can all speculate on 3XXX and he might be right or close to being right. It all makes sense when you think about it. 7nm is already taken so they have to go with 10nm. We'll see soon enough.

#5768932

Denial

2020-03-12 19:41

Reddoguk:

We can all speculate on 3XXX and he might be right or close to being right. It all makes sense when you think about it. 7nm is already taken so they have to go with 10nm. We'll see soon enough.

He could be right but 7nm being "already taken" doesn't mean much.. what if it's already taken (partially obviously) by Nvidia? These contracts are signed way in advanced. Jens just said a few months ago they are continuing production with TSMC: https://www.tweaktown.com/news/69424/tsmc-handle-nvidias-next-gen-7nm-ampere-gpus/index.html

During GTC 2019 in China, NVIDIA founder and CEO Jensen Huang addressed questions from the tech press over who will be building out most ogf the new 7nm GPUs in 2020 and beyond. Huang said that most of the 7nm GPU production will be done by TSMC while Samsung will only handling a small portion of 7nm GPU production for NVIDIA.

In fact he reiterated that in an investor call on Feb 13th:

Well, I think we're in pretty good shape on supply. We surely won't have ample supply, it is true that the industry is tight. And the combination of supporting multiple processes, multiple fabs across our partner TSMC, we've got a lot of different factories and a lot of different -- several different nodes of process qualified. I think we're in good shape. And so, we just have to watch it closely, and we're working very closely with all of our customers and forecasting and, of course, that gives us better visibility as well. But all of us have to do a better job forecasting and we're working very closely between our customers and our foundry partners, TSMC.

Just seems weird to me how the CEO is telling in investors and people one thing but random guy is saying it's something else.

#5768945

alanm

2020-03-12 20:23

Remember, this is the guy who posted the fake Big Navi specs leak by Hynix who called him out and said it was false.

#5768946

pharma

2020-03-12 20:23

This information comes from Twitter user CorgiKitty, who is relatively new to the leaks scene, so we might need some additional salt this time.

https://www.notebookcheck.net/No-7-nm-for-Nvidia-s-upcoming-Ampere-gaming-GPUs.457329.0.html

#5768953

Denial

2020-03-12 20:28

alanm:

Remember, this is the guy who posted the fake Big Navi specs leak by Hynix who called him out and said it was false.

Nah it was a different guy, @CyberCatPunk - but just another reason not to believe these people.

#5768997

Reddoguk

2020-03-12 23:56

If Nvidia don't come out with a reply to these rumors pretty quickly then can we assume it is in fact correct? That's 2 or 3 news articles saying no 7nm for Amp so i thought Nvidia would respond to at least fake allegations if it was so.

#5769010

nevcairiel

2020-03-13 00:54

Reddoguk:

If Nvidia don't come out with a reply to these rumors pretty quickly then can we assume it is in fact correct? That's 2 or 3 news articles saying no 7nm for Amp so i thought Nvidia would respond to at least fake allegations if it was so.

2 or 3 articles all based on the same guys statements with no proof and no history of being correct? Not sure why this is even given as much attention. What makes these rumors any better then the last ones that said 7nm and 70%+ performance increases? Slow tech news week I guess with the entire world busy fighting viruses. As a company, your best bet is to not respond to rumors, or people will just make up more to get more information out of you.

#5769046

Astyanax

2020-03-13 03:44