• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."
  • Community Question: What makes a good motherboard?

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

Page 205 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Mopetar

Diamond Member
Jan 31, 2011
5,236
1,850
136
Well it is only FE cards so it doesn't stop them from getting the AIB cards that most people are more interested in anyhow, but it also makes you wonder if Nvidia would lead on the AIB partners not to send cards to Hardware Unboxed as well. That starts to cross over into what could get Nvidia into legal trouble, but good luck taking an industry giant to court as a small YouTube channel.
 

CakeMonster

Senior member
Nov 22, 2012
993
79
91
I would love for reviews to be released 2-3 months after each new GPU and game is released, so that we have an idea of actual performance for the product's lifetime rather than with unfinished drivers and day 0 version of the game.

There's absolutely no market demand for that so we'll never have it outside of reviewers updating their review rigs and re-testing everything when the next products arrive.
 

Mopetar

Diamond Member
Jan 31, 2011
5,236
1,850
136
I would love for reviews to be released 2-3 months after each new GPU and game is released, so that we have an idea of actual performance for the product's lifetime rather than with unfinished drivers and day 0 version of the game.

There's absolutely no market demand for that so we'll never have it outside of reviewers updating their review rigs and re-testing everything when the next products arrive.
I'm not sure what you're talking about since there are plenty of reviews that look at older cards and how performance has changed over time. Here's one showing the 5700XT and 2070 Super over half a year after their launch.

You just have to go looking for that kind of content if you really want it since it doesn't get passed around or as publicized on forums as launch reviews do.
 

CakeMonster

Senior member
Nov 22, 2012
993
79
91
That is really hit or miss though, and you can't be sure to find it for the games or hardware you specifically are interested in. Its more a general criticism that if you want to have any idea how some kind of hardware or game has matured, you typically will need to look for the data in articles or reviews focused on something different. 6xxx vs 3xxx series general performance numbers over many game titles in march will be hard to get from any kind of review expect for reviews of new games that are released then. Performance of CP2077 in march (with hopefully tons of fixes) for those that want to play the game then will be hard to figure out unless there is an excuse for some site to do a 6xxx vs 3xxx re-test then. We are too often stuck with early numbers which might not reflect performance properly since both hardware and games are usually enjoyed over months if not years, and most of the change in performance happen with driver and patches in the few weeks or months after release.
 

Stuka87

Diamond Member
Dec 10, 2010
5,294
1,075
136
Well it is only FE cards so it doesn't stop them from getting the AIB cards that most people are more interested in anyhow, but it also makes you wonder if Nvidia would lead on the AIB partners not to send cards to Hardware Unboxed as well. That starts to cross over into what could get Nvidia into legal trouble, but good luck taking an industry giant to court as a small YouTube channel.
Except that nVidia prevents AIB cards from being reviewed until AFTER FE cards. And in the world of YouTube, if you don't have a video up at the very moment you are allowed, you arent going to get views. This is nVidia's way of trying to force just not HWUB, but every other reviewer out there to review products the way nVidia wants them too.

Now nVidia did eventually walk back the content of that email, but this is not the last we have heard of this, just like its not the first (Lets not forget what they did to [H]).
 
  • Like
Reactions: coercitiv

uzzi38

Golden Member
Oct 16, 2019
1,305
2,425
96
Well it is only FE cards so it doesn't stop them from getting the AIB cards that most people are more interested in anyhow, but it also makes you wonder if Nvidia would lead on the AIB partners not to send cards to Hardware Unboxed as well. That starts to cross over into what could get Nvidia into legal trouble, but good luck taking an industry giant to court as a small YouTube channel.
Jay talked about this in his video. AIBs send a list of everyone they're seeding cards to Nvidia who have to approve it. So Nvidia could also block AIBs sending cards to HUB easily.
 
  • Like
Reactions: Glo.

jpiniero

Diamond Member
Oct 1, 2010
8,699
1,561
126

Videocardz says the A6000 is now available. $4650. Fully enabled GA102 with 48 GB GDDR6.
 
  • Like
Reactions: lightmanek

jpiniero

Diamond Member
Oct 1, 2010
8,699
1,561
126

Now MSI makes an 3090 board that rips off the Fermi reference design.
 
  • Haha
Reactions: Mopetar and ozzy702

Stuka87

Diamond Member
Dec 10, 2010
5,294
1,075
136

Now MSI makes an 3090 board that rips off the Fermi reference design.
Is it really ripping it off if its a cooler they made back then?

However, a blower on a 400W card? No thanks.
 

CP5670

Diamond Member
Jun 24, 2004
4,607
107
106
I think those blowers actually do keep the cards cool, but make tons of noise. The Quadros have always used them so they must be reliable. I guess the noise doesn't matter if the card is in a server room.
 
  • Like
Reactions: VirtualLarry

Bouowmx

Golden Member
Nov 13, 2016
1,088
481
146
The next (GeForce) architecture might not be Hopper (MCM), but another intermediate architecture still on 5 nm (monolithic).
Could be that the next product family is Ampere on Samsung 5 nm. This situation draws parallels to Maxwell-Pascal-Volta, in which Volta was to succeed Maxwell, but Volta became compute-only, and Pascal was inserted as Maxwell on a new process.
 

GodisanAtheist

Platinum Member
Nov 16, 2006
2,679
1,171
136
I was always a little stunned that we were going for MCM given (as far as I know) basically none of the challenges revolving around MCM GPUs for gaming purposes has actually been resolved as of yet.

I can see MCM becoming a big deal in the compute/ML side of things, but gaming?

It would honestly be shocking if NV/AMD pulled it off in 2022.
 

uzzi38

Golden Member
Oct 16, 2019
1,305
2,425
96
The next (GeForce) architecture might not be Hopper (MCM), but another intermediate architecture still on 5 nm (monolithic).
Could be that the next product family is Ampere on Samsung 5 nm. This situation draws parallels to Maxwell-Pascal-Volta, in which Volta was to succeed Maxwell, but Volta became compute-only, and Pascal was inserted as Maxwell on a new process.
Hopper isn't even a GPU uArch ahaha. It's more accurate to call it a chonker Tegra.
 
  • Like
Reactions: Glo.

VirtualLarry

No Lifer
Aug 25, 2001
50,040
5,856
126
Re: 3060 non-ti
The card will feature a 192-bit bus interface and at 16 Gbps, delivering bandwidth of up to 384 GB/s. The GeForce RTX 3060 could end up being priced at $299 US which will make it a formidable graphics card.
Do I hear "mining card"? I think that I do!

Edit: With the 12GB VRAM buffer, it should be able to hash the Grin/CuckooWHATEVER algos that require the "larger" VRAM frame-buffers.
 

GodisanAtheist

Platinum Member
Nov 16, 2006
2,679
1,171
136
The rumors on RAM with NV cards is giving me a headache.

Have we ever historically had a situation where a lower tier graphics card had more onboard ram of the same type as a higher tier card (i.e. 3060 12gb GDDR6 vs 3060ti/3070 8Gb)?

I know there is precident for things like 4GB Fiji vs 8GB Grenada on account of one being HBM and the other GDDR5.

I don't think that has ever been the case even with staggered releases...
 

ozzy702

Golden Member
Nov 1, 2011
1,059
444
136
The rumors on RAM with NV cards is giving me a headache.

Have we ever historically had a situation where a lower tier graphics card had more onboard ram of the same type as a higher tier card (i.e. 3060 12gb GDDR6 vs 3060ti/3070 8Gb)?

I know there is precident for things like 4GB Fiji vs 8GB Grenada on account of one being HBM and the other GDDR5.

I don't think that has ever been the case even with staggered releases...
It's strange for sure. NVIDIA got caught with their pants down, that's all there is to it.
 
Last edited:
  • Like
Reactions: maddie

ASK THE COMMUNITY