• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The (Ada) Lovelace architecture Thread.

Wrinkly said:
The exact spec, no. The possibility of Lovelace before Hopper due to Ampere being on 8nm and eating a lot of power sounds valid.
Click to expand...

Yeah I can see there being a shrink before a new architecture but a lot of comments on here were a lot more about it actually being what has been stated even though we should all know better than that.
 
opethdisciple said:
For me it's quite evident that performance has hit a wall now against power consumption. These GPU's are at the edge of acceptable power consumption if you ask me.

750w - 850w is probably as high as they can go for a consumer based gaming GPU.

I cannot see them making more powerful cards with increased power requirements. It would be crazy.
Click to expand...

my 3090 at 410w has hit that wall.

It's pumping out so much heat in the case that my other components, especially the x570 pcb is getting close to their maximum safe temp range. (Keep in mind I also have a 5950x overclocked at 250w).

If I add any more heat some of these components will start causing system crashes. And I have a very airflow focused case, there is just a limit to how much heat can be dealt with. If I did add more heat, I would now have no choice but to move to custom water cooling
 
Wrinkly said:
The exact spec, no. The possibility of Lovelace before Hopper due to Ampere being on 8nm and eating a lot of power sounds valid.
Click to expand...

the specs while sounding plausible. I belive they have come from one of two things:

1) someone either just making up crap or

2) an Nvidia employee is speculating with specs - I say speculation because in fact the only people at Nvidia who would know the exact details of future products are the President/Vice Presidents who would never leak anything. The people at Nvidia who leak stuff is low level employees like engineers, however they also don't know the full spec because Nvidia uses a special environment where none of these employees ever get to work on or see the entire product, they work in seperate teams designing single sections of the gpu/card in an emulation environment.

Also the specs could just one of many designs they are testing
 
Grim5 said:
my 3090 at 410w has hit that wall.

It's pumping out so much heat in the case that my other components, especially the x570 pcb is getting close to their maximum safe temp range. (Keep in mind I also have a 5950x overclocked at 250w).

If I add any more heat some of these components will start causing system crashes. And I have a very airflow focused case, there is just a limit to how much heat can be dealt with. If I did add more heat, I would now have no choice but to move to custom water cooling
Click to expand...

Exactly. Power consumption and heat go hand in hand. Increased power consumption is possibly acceptable, but the levels of heat these cards will produce will make them not really feasible for mass production.

You don't expect the end user to have any special heat dissipating abilities at home.
 
Grim5 said:
a 120tflop card is almost equal to 4x 3090. Control with rayvtracing maxed out at 8k and no dlss runs at 10fps on the 3090, so the 120tflop card would do 40fps... at8k, never mind 4k
Click to expand...
Yes but I said 'vaguely demanding' meaning more demanding than 'Control' which doesn't even bother to render half the destruction in RT reflections. I'm assuming games will actually start to use RT to create images such as water and reflections that are a significant advance on what we have today rather than to a similar standard to how it's traditionally been rendered with a gimped non-RT version to showcase how 'great' RT is. A fourfold increase in performance will be good but it's just a step in the right direction as we can expect the level of game design and complexity to continue to increase as it always has.
 
Gregster said:
I can see CPUs becoming the major bottleneck soon. With the current CPUs, there is a bottleneck at 1080P and even signs at 2160P, chuck in a beast of a card and that 4K game is now held back because of the CPU.
Click to expand...
At 4K that would depend on what fps is your target. I always lock mine to stay either at 57fps (like in cyberpunk) and 60fps for other games that need 60 to run properly like Mortal Kombat 11. My 3600 never gets fully utilised trying to hit 60fps.

That said obviously there is more to it than that and I may benefit from better minimums etc with a 5600x which has better IPC or some other intel cpu, but for now I am happy for another year at least.

Want see what Intel and AMD bring with their next gen offerings and decide if I will go to a new platform or just grab a discounted 5900X and be done with it it for the next 5 years.
 
D.P. said:
If Hopper was ever going to be MCM design ghen it was never destined for consumers Graphics.
Click to expand...

More people not understanding the MCM approach which is mostly likely to be employed for consumer graphics. Everyone seems to be locked in a vision of Zen like chiplets.
 
TNA said:
At 4K that would depend on what fps is your target. I always lock mine to stay either at 57fps (like in cyberpunk) and 60fps for other games that need 60 to run properly like Mortal Kombat 11. My 3600 never gets fully utilised trying to hit 60fps.

That said obviously there is more to it than that and I may benefit from better minimums etc with a 5600x which has better IPC or some other intel cpu, but for now I am happy for another year at least.

Want see what Intel and AMD bring with their next gen offerings and decide if I will go to a new platform or just grab a discounted 5900X and be done with it it for the next 5 years.
Click to expand...

hea talking like 120fps+
 
Rroff said:
More people not understanding the MCM approach which is mostly likely to be employed for consumer graphics. Everyone seems to be locked in a vision of Zen like chiplets.
Click to expand...

How does it work? A quick look seems to show separate GPU modules and an IO die.
 
You must log in or register to reply here.
Back
Top Bottom