DrMrLordX
Lifer
- Apr 27, 2000
- 21,791
- 11,133
- 136
Two words: Heat Density.
You think heat density on Navi10 is worse than Vega20? Or the same?
Two words: Heat Density.
Should be the same for all 10 nm products.You think heat density on Navi10 is worse than Vega20? Or the same?
Should be the same for all 10 nm products.
Yes, 7 nm...Er, 7nm?
Anyway this does not explain why Radeon VII has OC headroom while 5700XT apparently has none. Unless the 5700XT's cooler is just that bad.
https://www.computerbase.de/2019-07/radeon-rx-5700-xt-test/4/
What will you say now?
And lastly. I never claimed that it has higher per ALU performance. I said that it has the same performance per ALU as Turing. And that it MAY have higher performance per ALU, versus Turing, but that remained to be confirmed. Now it is.
Navi 10 full design was respun to increase clocks to hell.
In no world any GPU can lose performance per ALU at higher clocks. That is not how this works.Finally a source. Interesting. Now how things work in civil society is that this does not erase my two sources of TPU and AT, rather it compliments them and we pool the knowledge together.
Based on three different sources showing frequency, it seems to show that Navi is equal or a tad higher perf per shader at lower clocks but begins to lose the edge at higher clocks. Notice the 1.5ghz there, compared to 1.8ghz± at AT and TPU both where Turing wins. Perhaps Navi frequeny scalability is a bit worse at these higher clocks. It'll be interesting to see aftermarket OC battles where perhaps we can get stable 2ghz battles. Thanks for the link.
Possibly not. Haven't seen any. Its funny but this makes me thinking, how this will affect smaller GPUs, with lower TDPs, like Small Navi, which I am interested in. There is quite a large downfall of efficiency with RX 5700 XT compared to standard RX 5700. I wonder how would full GPU perform in power consumption if it would be equally clocked to RX 5700 and what would be the performance difference between those two.Anyone done a 5700XT on water review yet?
Thats because GPU performance is not only affected by pure GPU core clocks. But in no world, when you take out all of variables on performance, GPU can lose ALU performance at higher clocks. Its not how this works. If you will forget about those variables, which affect GPU performance, then yes, it can lose performance per ALU. But not, OBJECTIVELY.That's exactly how it can work. GPU performance doesn't scale 1:1 with overclocking.
Are you really gonna make me show you overclocks between 2 different GPUs with varying performance gains? Really?
This. In the first place. It may tell a lot.We need more to know, and aftermarket coolers.
The main issue is AMD needs 7nm to be “about equivalent”.
One: it's a blower card so all the air is going out the back. A blower will actually keep your PC cooler.
Two: The temperature of the GPU is irrelevant. What matters is how many watts of energy are being dumped into your case.
While aslo complete gutting computte performance, I mean that used to be the strong point before and now Navis compute is way behind NV. They went from one extreme to the other and still only match NV on peformance / watt with a node advanatge. It's fine for now and AMD can simply hope Ampere really is a GV100 replacement only and NV will switch to 7nm in 2021 as rumors say.
Has anyone got any info for the new features? Particularly the reduction of input latency? I've been waiting for what I assumed was a more robust review by AT, but I get the feeling what's up is the final version.
Somewhat regrettable, however, is Anti-Lag only works with Direct X 11 and Direct X 9, and last but not least with the new Navi GPUs. While there are still plenty of DX11 titles, in the future these will probably be rarer when developers switch to DX12 or volcano. The benefits of Anti-Lag will likely decrease slowly if AMD does not provide support for these new APIs.
BTW, Navi could definitely match or even beat GTX 2080 Ti for 300W – after all, 5700 proves Navi, when clocked right, can match Turing PPW.
Er, 7nm?
Anyway this does not explain why Radeon VII has OC headroom while 5700XT apparently has none. Unless the 5700XT's cooler is just that bad.
Dat's a spicy meatball!
I can think of a reason!
The Radeon VII clocks quite a bit less. So with the 5700XT they might have pushed the limit further just so it can beat the RTX 2070.
5700XT boost clock is only ~100 MHz higher. Not sure what clocks 5700XT actually gets during gameplay, but I've seen my undervolted-but-otherwise-stock Radeon VII hit 1900 MHz playing The Division 2 . . . and when I OCed it to 1940MHz, it would boost over 2000 MHz sometimes.
2086 MHz on the stock blower is not that bad. Would like to see where it can go under water.
FinFET processes are more efficient in lower frequencies, but seems to need increased power after a certain point is reached. We saw that in a way with Ivy Bridge.
It is because 5700XT hits temperature problems(I thinknthisbis mentioned in the TPU review) and downclocks a lot. Those average frame rates aren't painting the whole picture and we will need better cooling for similar looking graph like those of Nvidia. Then I am pretry sure that it will be similar performance as it is with 1500mhz.I'm not talking about "losing ALU performance at higher clocks"? Huh?
I'm not saying Navi loses performance as it overclocks, just that perhaps it doesn't get quite as much extra gain per clock. We know different GPUs gain more from overclocks than others (irrefutable), and perhaps Navi is less than Turing.
At 1.5GHz Navi => Turing
At ~1.85Ghz Turing => Navi
An analysis based on the 3 sources we have both presented anyway. We need more to know, and aftermarket coolers to test higher clocks.