Question 'Ampere'/Next-gen gaming uarch speculation thread

[Ottonomous]

How much is the Samsung 7nm EUV process expected to provide in terms of gains?
How will the RTX components be scaled/developed?
Any major architectural enhancements expected?
Will VRAM be bumped to 16/12/12 for the top three?
Will there be further fragmentation in the lineup? (Keeping turing at cheaper prices, while offering 'beefed up RTX' options at the top?)
Will the top card be capable of >4K60, at least 90?
Would Nvidia ever consider an HBM implementation in the gaming lineup?
Will Nvidia introduce new proprietary technologies again?

Sorry if imprudent/uncalled for, just interested in the forum member's thoughts.

 

[A///]

RDNA2 hardware may be better than what 99% of customers think it will be like. AMD's software for their video cards is always a massive let down. Looks like I'm waiting for RDNA2 officially or mid-life refreshed on the 30 series.

On the one hand, this is karma for scalpers. On the other, I feel sorry for anyone who bought from scalpers and hope they get their money back.

If this were the case we would heard more from AMD at Ampere launch. The Silence is telling, that they do not have a winner at their hands. I mean not even a single "Poor Ampere" add - not that this would mean much given AMDs track record...

This is too simplistic of a view. If that's the baseline then Zen 3 is a failure as well. The people who headed marketing at AMD were either made redundant or pushed out the door over time, because they all went to NVidia and Intel, and caused issues and were let go from there too or moved onto other companies such as Nuvia. RTG's last head is now at Intel for the Xe program and is up to his old tricks.

I'd prefer to be wrong about the software.

 

[Tup3x]

RDNA2 hardware may be better than what 99% of customers think it will be like. AMD's software for their video cards is always a massive let down. Looks like I'm waiting for RDNA2 officially or mid-life refreshed on the 30 series.

On the one hand, this is karma for scalpers. On the other, I feel sorry for anyone who bought from scalpers and hope they get their money back.


This is too simplistic of a view. If that's the baseline then Zen 3 is a failure as well. The people who headed marketing at AMD were either made redundant or pushed out the door over time, because they all went to NVidia and Intel, and caused issues and were let go from there too or moved onto other companies such as Nuvia. RTG's last head is now at Intel for the Xe program and is up to his old tricks.

I'd prefer to be wrong about the software.

Their software sucks. I got a AMD video driver caused BSOD (well, picture froze but it did do the memory dump) while in Teams meeting with customer. Not only that the video had some constant flickering. There's some other (visual) issues too with Chromium browsers. Considering how old the GPU architecture is in Renoir this doesn't look good. Video engine is new though (and they got rid of every single video related setting except brightness in the driver).

 

[itsmydamnation]

Their software sucks. I got a AMD video driver caused BSOD (well, picture froze but it did do the memory dump) while in Teams meeting with customer. Not only that the video had some constant flickering. There's some other (visual) issues too with Chromium browsers. Considering how old the GPU architecture is in Renoir this doesn't look good. Video engine is new though (and they got rid of every single video related setting except brightness in the driver).

thats odd, just once or all the time?
i webex teams , MS teams and skype meeting multiple time a day every day on a renior ( lenovo slim 5, 4700u) without issue, infact its way better then my previous intel based lenovo x1 carbon.

 

[Tup3x]

thats odd, just once or all the time?
i webex teams , MS teams and skype meeting multiple time a day every day on a renior ( lenovo slim 5, 4700u) without issue, infact its way better then my previous intel based lenovo x1 carbon.

Once so far (also it was rather lengthy meeting). Especially in full screen mode seems less than ideal. Google Meet is fine though it doesn't seem to use hardware decoding.

 

[Glo.]

If this were the case we would heard more from AMD at Ampere launch. The Silence is telling, that they do not have a winner at their hands. I mean not even a single "Poor Ampere" add - not that this would mean much given AMDs track record...

Yeah, given AMD's track record lately. RDNA1, Renoir, Zen 1, Zen 2 CPUs.

It boggles my mind that people believe that this is still the same company from the times of Bulldozer, considering their latest execution on all fronts.

P.S. As I have been trying to tell everyone for months. Considering how all of Ampere rumors, and info leaked out in typical fashion of this industry: "Good news never leak out, bad news spread with the speed of light in this industry".

And considering how Ampere is an engineering flop, rushed release I would be at least slightly optimistic about AMD's releases.

 

[Dribble]

Problems with the 3080 already??

This is really useful, basically don't buy a card with the lesser capacitor. The real conclusion is the one we all knew - give any new product a little while for all the bugs and problems to become obvious before you buy.

 

[Glo.]

This is really useful, basically don't buy a card with the lesser capacitor. The real conclusion is the one we all knew - give any new product a little while for all the bugs and problems to become obvious before you buy.

There will be follow up on the video by JayZ, because it turns out that the problem is not as simple as he made it out to be in the video.

And this is worth keeping in mind, that not exactly the story is that you buy with GPU two rows of small capacitors.

 

[A///]

Yeah, given AMD's track record lately. RDNA1, Renoir, Zen 1, Zen 2 CPUs.

It boggles my mind that people believe that this is still the same company from the times of Bulldozer, considering their latest execution on all fronts.

P.S. As I have been trying to tell everyone for months. Considering how all of Ampere rumors, and info leaked out in typical fashion of this industry: "Good news never leak out, bad news spread with the speed of light in this industry".

And considering how Ampere is an engineering flop, rushed release I would be at least slightly optimistic about AMD's releases.

Never thought I'd see you be so sassy. My concerns for AMD are strictly software based. If they can compete on hardware, that's excellent. I think they will. Their software is always the big let down. I'd love to be wrong this time. Perhaps, I'm already wrong...

 

[Glo.]

Never thought I'd see you be so sassy. My concerns for AMD are strictly software based. If they can compete on hardware, that's excellent. I think they will. Their software is always the big let down. I'd love to be wrong this time. Perhaps, I'm already wrong...

Yeah. Drivers were sub-par. Thankfully, not on Linux, which is the most interesting part for me, because Im moving 100% to Linux, and Mesa on AMD side is pretty stable and very performant.

 

[insertcarehere]

Yeah, given AMD's track record lately. RDNA1, Renoir, Zen 1, Zen 2 CPUs.

It boggles my mind that people believe that this is still the same company from the times of Bulldozer, considering their latest execution on all fronts.

P.S. As I have been trying to tell everyone for months. Considering how all of Ampere rumors, and info leaked out in typical fashion of this industry: "Good news never leak out, bad news spread with the speed of light in this industry".

And considering how Ampere is an engineering flop, rushed release I would be at least slightly optimistic about AMD's releases.

Where did the "silent AMD == good AMD" narrative come from? When AMD hit paydirt with Zen 1/2, they had no problem showing off performance against the competition well before launch.

Zen 1:

in August 2016, 7 months before launch

Zen 2: https://www.anandtech.com/show/1382...live-blog-looking-ahead-starts-9am-pt-5pm-utc in January 2019, 6 months before launch

 

[Glo.]

Where did the "silent AMD == good AMD" narrative come from?

And how does it work for GPUs in AMD/ATi history?

P.S. Where did the "Silent AMD = bad AMD" narrative came from?

 

[Heartbreaker]

Why is this becoming an AMD thread?

 

[lightmanek]

Where did the "silent AMD == good AMD" narrative come from? When AMD hit paydirt with Zen 1/2, they had no problem showing off performance against the competition well before launch.

Zen 1:

in August 2016, 7 months before launch

Zen 2: https://www.anandtech.com/show/1382...live-blog-looking-ahead-starts-9am-pt-5pm-utc in January 2019, 6 months before launch

And even then, where they were loosing significantly to competition, loosing investors, working against credit rating companies under constant pressure to show something positive off, they still underplayed their progress by claiminmg 40% IPC improvement ...
Yes, when you are behind you need to keep some sort of public intrest not to loose partners, investors and clients, therefore you are more willing to send early samples, leak information or even do a pre-launch demos showing yourself in a best light possible.
Look at Intel and how much they stared leaking roadmaps and products information lately, there are reasons for that, and one of them is strong competion, not only from AMD.

 

[DooKey]

Why is this becoming an AMD thread?

Look at the usual suspects infecting the thread...

 

[Mopetar]

You haven't demonstrated that Steam Data is unreliable. Throwing FUD at it, doesn't make it unreliable, and as I pointed out in the post you quoted, Overall Steam results concur with the other sources.

Steam is handy because we get to see more granular results, but they still line up with other less granular sources.

Again the burden to show they're reliable rests on you. That's why researchers publish their methodology. It's not someone else's job to prove you wrong. It's your job to show that there's no good reason to believe that you're incorrect.

But it doesn't take much to demonstrate why it's unreliable. It's not a random sample of the population. The statistics that allow us to determine with reasonable accuracy the likely value for a population while only looking at a small subset of it require that the sample be random. Steam survey isn't.

You could still use it, but without knowing the methodology, response rate, etc. you cannot even start to determine how confident you should be in the results or what the possible margin of error might be.

Here's a good historical example of exactly why this is a problem and just how wrong a survey can be: https://www.math.upenn.edu/~deturck/m170/wk4/lecture/case1.html

 

[Grooveriding]

The issue with the caps is real and confirmed by EVGA. They obviously knew about this prior to launch, these other AIBs and nvidia should have worked to not release any boards like this. Could have been too late for AIBs who got finished cards earlier than others I guess. They should all be revising their boards, once the stock situation improves, a lot of people will just avoid the boards with the lower quality power delivery.

Message about EVGA GeForce RTX 3080 POSCAPs - EVGA Forums

Hi all, Recently there has been some discussion about the EVGA GeForce RTX 3080 series. During our mass production QC testing we discovered a full 6 POSCAPs solution cannot pass the real world applications testing. It took almost a week of R&D effort to find the cause and reduce the PO...

forums.evga.com

Relieved because I had an MSI card in my cart and skipped on it for EVGA or FE. Hope big piles of stock next week and I can finally get one of these.

 

[Heartbreaker]

But it doesn't take much to demonstrate why it's unreliable. It's not a random sample of the population. The statistics that allow us to determine with reasonable accuracy the likely value for a population while only looking at a small subset of it require that the sample be random. Steam survey isn't.

Where are you getting the idea that it isn't random?

 

[CakeMonster]

Have any of the reviews so far done some in depth testing of core and vram OC separately?

 

[A///]

No. With EVGA doing a likely manual swap of the caps on whatever has been manufactured, I doubt we'll see anything. The Asus TUF card, the one shown off a week ago now is the only card with pure MLCC. If you go by Igor's theory, it should not be a problematic card. However, my own intuition tells me people should wait to see if that is the only source of problems or if it's a multi component failure.

Someone suggested on ocn that Nvidia signs off on all AIB partner designs. If true, I'm curious why they didn't flag any of these.

 

[Markfw]

No. With EVGA doing a likely manual swap of the caps on whatever has been manufactured, I doubt we'll see anything. The Asus TUF card, the one shown off a week ago now is the only card with pure MLCC. If you go by Igor's theory, it should not be a problematic card. However, my own intuition tells me people should wait to see if that is the only source of problems or if it's a multi component failure.

Someone suggested on ocn that Nvidia signs off on all AIB partner designs. If true, I'm curious why they didn't flag any of these.

I thik its interesting that EVGA is the only one that SHIPPED nothing but fixed hardware, hence why they were late. Thats why I only get EVGA cards now.

 

[A///]

I thik its interesting that EVGA is the only one that SHIPPED nothing but fixed hardware, hence why they were late. Thats why I only get EVGA cards now.

They've always been a pretty decent company IMO. A lot of good AIB partners have come and gone. Powercolor reached out to a YouTube tech influencer and said to hold off on the review because they were encountering trouble with the hardware themselves.

There's a lot of wild theories about this debacle. If NVidia approved these designs, what does that say about NVidia? Especially after their entire botched launch of their own FE cards among other things. We're not at Fermi or FX levels of disappointment. Yet.

 

[KompuKare]

Powercolor reached out to a YouTube tech influencer and said to hold off on the review because they were encountering trouble with the hardware themselves.

Kudos on Powercolor. Pity they didn't make s public statement, although like the relationship between Intel and motherboard makers, the AIBs might be reluctant to say anything without Nvidia's approval.

We're not at Fermi or FX levels of disappointment. Yet.

Surely the infamous "golden standard" for GPU problems was when Nvidia sold the ticking timebomb millions of parts with faulty solder bumps. Of course, that mostly manifested itself 1-2 years later.

 

[JujuFish]

They've always been a pretty decent company IMO.

EVGA screwed me over. Have no desire to ever buy from them again. Fought with me for 4months to honor their warranty, made me pay shipping, and sent me a dead on arrival card, which they then refused to replace.

 

[ozzy702]

Software/stability and top tier performance has always been what has brought me back to NVIDIA time and time again for my main box. My kids get AMD cards. I would love for AMD to invest heavily in their driver/software teams because it does seem like their hardware is solid and usually underutilized/optimized for by drivers/software. Hopefully NVIDIA and AIBs figure out this 3000 series mess, Samsung can make some progress on "8nm" and availability and quality improve over the next few months.

We'll see. If AMD actually pulls off a win with RDNA2 maybe I'll roll the dice, cross my fingers, throw salt over my shoulder, offer up some sacrifices to the PC gods and try my luck with Big Navi... NVIDIA has screwed the pooch twice in a row now, and left themselves open to competition.

 

[A///]

Kudos on Powercolor. Pity they didn't make s public statement, although like the relationship between Intel and motherboard makers, the AIBs might be reluctant to say anything without Nvidia's approval.

Jay hinted at that. I bet you they've asked AIB partners not to say anything with threat of canceling their contract if they do. We're going to hear more about this and possibly other problems in the coming weeks.

Surely the infamous "golden standard" for GPU problems was when Nvidia sold the ticking timebomb millions of parts with faulty solder bumps. Of course, that mostly manifested itself 1-2 years later.

Laptops? Initially I thought you were talking about the VRM failures but those resulted in melting cards and cards smoked out by a mini-fire.