Speculation: GPU Performance Comparisons of 2020 *Spawn*

Status
Not open for further replies.
Idk chief a lot of procurements rely on SPEC scores so it probably is.

Without system configuration information, a SPEC score is pretty much meaningless.

Furthermore, Milan and Rocket Lake are targeting very different markets -- I assume you have some kind of performance preview for Milan, correct?
A ~15% IPC improvement of Zen 2 over Zen 3 may not be enough to sufficiently dethrone Intel when it comes to desktop gaming performance leadership. Sunny Cove/Willow Cove already offers a ~15% IPC uplift over Skylake, roughly matching the IPC gain of Zen 3. The only thing holding Intel back is its broken 10nm process and its decision to stick to 14nm++ with RKL.
 
navi 10 clocked at 2230Mhz and fast dram could be just 20% faster, so PS5 and XBSX should be both near RTX 2080 Super land, still other 20% left for 2080 Ti or RTX 3070...i don´t count on much IPC gain in RDNA 2 so they would need 60cu aka navi 22 for that.
 
Zen3 will offer around 15% IPC improvement but also 3-400Mhz clock boost. I've heard AMD internal challenge is to ship a 5Ghz boost desktop SKU. However, no idea if they will succeed...
 
PS5 probably around 2070/2070s. Theoretically a 3070 is double the performannce.



Yes a narrow 36cu part. A 80CU variant wont be screaming along at 2.3ghz range of speeds.
You really think RTX 3070 will double 2070S? PS5 closer to 2070S and XBSX closer to 2080S ok...so PS5 close to navi 10, will see. Anyway, 2080 Ti is 16TFlop at real gaming clocks of 1830Mhz, 5700XT is 9.5TFlop at real gaming clocks of around 1890Mhz, while performance difference is currently 43% 1440p and 50% 2160p, so its clear navis TFlops are more performant than Turings in gaming real world (RT off), let alone Amperes which seems like 2/3rd Turings in gaming performance wise.
 
You really think RTX 3070 will double 2070S? PS5 closer to 2070S and XBSX closer to 2080S ok...so PS5 close to navi 10, will see. Anyway, 2080 Ti is 16TFlop at real gaming clocks of 1830Mhz, 5700XT is 9.5TFlop at real gaming clocks of around 1890Mhz, while performance difference is currently 43% 1440p and 50% 2160p, so its clear navis TFlops are more performant than Turings in gaming real world (RT off), let alone Amperes which seems like 2/3rd Turings in gaming performance wise.

Or maybe the limiting factors are not always tflops...
 
So far all estimations of 30x0 are based on nVidias own numbers, and Digital Foundry who were only allowed to use a few titles specifically selected by nVidia (gee, wonder why?).
nVidia knows that lacking better information, tech sites and forum warriors will run with what crumbs of information they have, but it really doesn’t make sense to get Very Invested in how these upcoming cards will perform relative to each other since we don’t have good performance data for any of them.
 
So far all estimations of 30x0 are based on nVidias own numbers, and Digital Foundry who were only allowed to use a few titles specifically selected by nVidia (gee, wonder why?).
nVidia knows that lacking better information, tech sites and forum warriors will run with what crumbs of information they have, but it really doesn’t make sense to get Very Invested in how these upcoming cards will perform relative to each other since we don’t have good performance data for any of them.
3080 is genuinely an >okay product if you don't mind housefires.
 
3080 is genuinely an >okay product if you don't mind housefires.

I can deal with a housefire. I cant however write my own driver in case Big Navi turns out like 5700XT with software issues for the first 7 months. Nobody cares if its 3% faster than 3080 if it still has shitty drivers

AMD fine wine must mean you get yout moneys worth 2 years after launch
 
Status
Not open for further replies.
Back
Top