WYP
News Guru
Last edited:
AMD's rumoured Navi 21 graphics cards may feature 80 CUs - 2x stronger than a RX 5700
- Thread starter WYP
- Start date
NeverBackDown
AMD Enthusiast
Beat it in temps and power consumption too.
AngryGoldfish
Old N Gold
Are you saying that because it's out of place for a higher powered card to draw more power and output more heat?
NeverBackDown
AMD Enthusiast
Double everything about the 5700xt. It will very much be power-hungry and hot. It doesn't sell well.
The only time people stop caring about performance is if AMD draws more power and runs hotter. It's a convenient excuse to stay Nvidia.
I do doubt the expectations this rumour gives though. Take into account Ray Tracing and there's not much room left and that means it's dense as f* which means hot. Which means lower clocks... etc etc.
Think we had similar discussion in a different thread. Its still a feature, which makes it marketable, but its use is very limited now given its adoption.
I wish AMD had simply come out and said "we don't think RT is mature enough to use yet, so for now we will focus on raw performance"
But saying that, perhaps if RT is brought to market by AMD, we will see more developers adopt it, thus making games more immersive. Next gen as you say.
AlienALX
Well-known member
To be fair AMD (well, ATI) were in the opposite position with the 5000 series and really wore out the high temp and power argument.
Thing is they were right, and it was a big deal.
Thing is since then it's been completely reversed. 6000 series, 7000 series, 290/x and so on. Fury needed an aio ffs.
It's become less of a big deal in the media though. Why? Probably because if they did what they did with Fermi AMD as a GPU company wouldn't exist any more.
But after that Vega I'm staying well away. Yeah my games ran good, but man what a price I had to pay. Not so much the energy costs but in human sweat.
This won't get better when they start adding more. The 5700XT is already too hot and loud when pushed hard. What's going to happen with a core twice as dense? The mind boggles.
Normally I'd be up for it as I have water cooling. But with a threadripper I now need to be careful not to overdo the GPU.
Thing is they were right, and it was a big deal.
Thing is since then it's been completely reversed. 6000 series, 7000 series, 290/x and so on. Fury needed an aio ffs.
It's become less of a big deal in the media though. Why? Probably because if they did what they did with Fermi AMD as a GPU company wouldn't exist any more.
But after that Vega I'm staying well away. Yeah my games ran good, but man what a price I had to pay. Not so much the energy costs but in human sweat.
This won't get better when they start adding more. The 5700XT is already too hot and loud when pushed hard. What's going to happen with a core twice as dense? The mind boggles.
Normally I'd be up for it as I have water cooling. But with a threadripper I now need to be careful not to overdo the GPU.
AMD is currently on par with NVidia in energy efficiency terms (RTX 2070 Super consumes about 10% more power than the 5700XT when both are stock). Admittedly that's with a node advantage but in therms of raw power efficiency there shouldn't be much concern.
Of course, that node advantage does mean AMDs dies are denser so dissipating each Watt of heat energy becomes a little bit harder, but GPU coolers have improved a lot.
Of course, that node advantage does mean AMDs dies are denser so dissipating each Watt of heat energy becomes a little bit harder, but GPU coolers have improved a lot.
AngryGoldfish
Old N Gold
We don't know for sure, but it's likely that RDNA2 will offer higher IPC. And the 7nm+ process will likely improve efficiency further. So while you could be as right, I think the next Navi cards will offer either double the performance at less than double the TDP, or will offer more than double the performance at double the TDP. While I could eat my words, as AMD have disappointed greatly in the past, it stands to reason that if the next Navi cards are based on a new architecture (obviously not a 'from the ground up' redesign), they could bring greater performance per watt over current Navi. Which means either AMD could go all out and give us a 300W behemoth that easily beats the 2080Ti (and would match the 3080 at a higher TDP but lower price), or a 250-275W sweet spot card that just beats the 2080Ti at the same TDP and at a much lower price.
I thought they did.
"I think ray tracing is an important technology, it's something we're working on as well from both a hardware and software standpoint. I think the important thing though -- and that's why we talk so much about the development community -- is that technology for technology's sake is okay, but technology done together with partners and really getting the development community fully engaged is really important." - Lisa Su
“Utilization of ray tracing games will not proceed unless we can offer ray tracing in all product ranges from low end to high end,” - David Wang
NeverBackDown
AMD Enthusiast
You're not wrong, but just going off past history I don't suspect much improvement to IPC. I suspect smaller inefficiencies they will address but the main underlying feature would be RayTracing support. That's why I am saying what I said.
I'd love to be wrong. I'd love AMD to surprise the heck out of everyone. I just don't hold much faith in that notion right now. It's kinda like expecting a Ryzen rebirth in the GPU space. I don't think we are there yet.
I'd love to be wrong. I'd love AMD to surprise the heck out of everyone. I just don't hold much faith in that notion right now. It's kinda like expecting a Ryzen rebirth in the GPU space. I don't think we are there yet.
Dicehunter
Resident Newb
AMD need to start using the chiplet approach like they did with Ryzen, I think that would net some large performance gains every gen.
NeverBackDown
AMD Enthusiast
That doesn't really suit GPU architectures though. Latency is a GPU's worst nightmare. That's the easiest and biggest performance loss for GPUs.
Lads, 2070 Super is rated at 215W but nobody loses their mind when 2080ti is so much larger.
Also, AMD is intentionally binning the Navi cards loosely to improve yields. Look at the 5600 XT for reference, it's a 160W card with the same shaders as 5700 and outperforms 2060 usually.
Which is a 160W+ card.
Or look at it this way, Navi was actually aimed much lower on the stack but because of the pricing reasons they turned up the clocks outside of its greatest efficiency band to match Nvidia's performance. Does it matter that it's new 7nm Vs a mature 12? Nope. What all matters is cost (from yields and pricing) and performance.
And AMD already basically tested tsmc's 7nm on the Radeon 7 so that they can judge cost of the 7nm, obviously they can do something much larger like they did but they had to wait for the process to mature and raytracing of course. Although it is unconfirmed AMD isn't simply slapping on RT cores and calling it a day. What they're gonna do for RT idk really but,
High end dies are always better binned and it's 7nm EUV AND a evolution of Navi so it's not unreasonable to expect of AMD to deliver similar TDP just that because the die sizes are smaller the coolers may be taxed more compared to 2080ti but then again 2080ti don't exactly run cool either.
Also 2080ti is after all much larger than the TU106 yet doesn't eat more than twice. It is simple better binned.
Similar threads
