-
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.
AMD Navi 23 ‘NVIDIA Killer’ GPU Rumored to Support Hardware Ray Tracing, Coming Next Year
- Thread starter Gregster
- Start date
- Status
Soldato
- Joined
- 20 Apr 2004
- Posts
- 4,413
- Location
- Rugby
Radeon VII is GCN ?
So in that slide they are comparing a 17nm chip with a 7nm chip so ofc the PPW will be better. That or the PPW gains from 7nm GNC to the 'new' 7nm RNDA1 where overblown abit
Your napkin maths is, how should i put this nicely, D+ at best hambug.
5700XT power per watt lead vs Radeon VII is
21% ( 1080P )
14% ( 1440P )
9% ( 4K)
15% overall.
If you get same uplift on the RDNA2 expect 15% across the board.
The Radeon VII is not RDNA and it (at least on that slide) is not GCN. GCN on that slide could be anything from Polaris or older but its not the Radeon VII, that was never intended to be a gaming card, its a bastardized Radeon MI card rebranded as an act of desperation to fill a gap between Polaris and Navi 10.
They are not comparable.
I'll put it this way: the Radeon VII has 60 CU's (3840 Shaders) @ 1800Mhz, the 5700XT has 40 CU's (2560 Shaders) @ 1900Mhz
They are = in gaming performance, its about a 40% IPC uplift to RDNA.
RDNA3, yes but a bit later than that. Q4 2021 maybe.
They are = in gaming performance, its about a 40% IPC uplift to RDNA.
RDNA3, yes but a bit later than that. Q4 2021 maybe.
In the GCN comparison vs Navi 10 is as follows. https://d1io3yog0oux5.cloudfront.ne...gy+Overview_AMD+FAD+2020_For+Distribution.pdf This is the full presentation. On a previous slide they break down the 50% perf/watt gain and point to end note RX-362 where it says Testing done by AMD performance labs on June 4 2019. Systems were tested with Intel i7-5930K CPU @3.5Ghz (6 core) with 16GB GDDR4 @ 2133 Mhz using an Asus X99-E Motherboard running Windows 10 Enterprise 64 bit (Ver: 1809, build 17763.053). Using the following graphics cards. Navi 10 (driver 19.30_1905161434 (CL# 1784070)) with 40 compute units, versus a Vega 64 (driver 19.4.1) with 40 compute units enabled.
As it so happens the actual gain over Vega 64 was 63% perf/watt in the computerbase.de test suite and 59% perf/watt in the Techpower up test suite.
EDIT: The RDNA vs RDNA2 is RX-325 which just says Testing done by AMD performance labs on 06/01/19* using the Division 2 @ 25x14 Ultra settings. Performance may vary based on use of latest drivers.
* Guessing this is M/D/Y
Soldato
- Joined
- 16 Aug 2009
- Posts
- 8,197
Optimistic, I can see it competing with the 3070 though right in time for Nvidia to drop the 3070ti 16gb that was
Soldato
- Joined
- 20 Apr 2004
- Posts
- 4,413
- Location
- Rugby
apart from A: Vega 56/64 was marketed as a gaming card, B: the Radeon VII was marketed as a gaming card then suddenly a content creators card in its short life, C: there are Very little differences between Vega 10 and Vega 20 die's bar the node, Vega 20 had some extra transistors to allow for greater clocks but its pretty much a die shrink of Vega 10
Right, thank's
I think you're overestimating the 3070.
The thing is, Nvidia said the 3070 is as fast as the 2080TI, they also said the 3080 is twice as fast as the 2080, it isn't, not even close to that.
The 3070 has 5888 Shaders, the 3080 8704, the 3080 has 48% more shaders, its 32% faster than the 2080TI.
I guarantee the 3070 is no where near as fast as the 2080TI, its a 2080 Super. Its not much faster than a 5700XT.
The 2080TI is 35% faster than the 5700XT, it has 40 CU's (2560 shaders) 6800XT 60 CU's rumoured (3840 Shaders) +50%, scaling to 45%, that alone, without any IPC gain or higher clock speeds a 60 CU 5700XT like card is 5 to 10% faster than a 2080TI, that's a significant margin above the 3070.
Those financial Analyst slides were interesting, about 25% of the performance gains were from Vega /GCN to Navi were from the fab. process improvements from moving to 7nm.
75% of 50 (the claimed performance gain percentage) is 37.5%. So if we assume no fab. process improvement from RDNA to RDNA 2, +37.5% seems like a reasonable figure to expect for RDNA 2 vs the 5700xt, if AMD achieve similar performance improvements in other areas as they did with the first navi series.
5700 XT +37.5% performance would put it just ahead of the RTX 2080 TI.
75% of 50 (the claimed performance gain percentage) is 37.5%. So if we assume no fab. process improvement from RDNA to RDNA 2, +37.5% seems like a reasonable figure to expect for RDNA 2 vs the 5700xt, if AMD achieve similar performance improvements in other areas as they did with the first navi series.
5700 XT +37.5% performance would put it just ahead of the RTX 2080 TI.
Last edited:
Associate
- Joined
- 20 Jun 2016
- Posts
- 1,384
Exactly
Soldato
- Joined
- 8 Jun 2018
- Posts
- 2,827
*Rumor
We really don't have much to go on this time around with Radeon rdna 2 gpus. From the rumors it is alleged that the performance per watt has gone up from 50% to 60%. Also all remnants of GNC have been removed with RDNA 2 which will impact performance.
There is supposedly four skus. 2 in the mid-range, 2 in the higher end. I find it hard to believe that 60 cu's is the max. Top teir rdna2 is alleged to use hbm2/hbm2e.
Rumor does have it AMD expected the performance of ampere and was not impressed. Whatever that means. And contrary to the conspiracy that Nvidia is shortening supply I've heard that they just don't have enough units to sell.
Suggesting that is why AMD is so calm cool, collective and quite. AMD is expecting their units to be low up until they are ready to release.
We really don't have much to go on this time around with Radeon rdna 2 gpus. From the rumors it is alleged that the performance per watt has gone up from 50% to 60%. Also all remnants of GNC have been removed with RDNA 2 which will impact performance.
There is supposedly four skus. 2 in the mid-range, 2 in the higher end. I find it hard to believe that 60 cu's is the max. Top teir rdna2 is alleged to use hbm2/hbm2e.
Rumor does have it AMD expected the performance of ampere and was not impressed. Whatever that means. And contrary to the conspiracy that Nvidia is shortening supply I've heard that they just don't have enough units to sell.
Suggesting that is why AMD is so calm cool, collective and quite. AMD is expecting their units to be low up until they are ready to release.
Soldato
- Joined
- 21 Apr 2007
- Posts
- 2,645
60 CU for Navi 22 =
60 CU for biggest Navi =
60 CU for biggest Navi =
No, because AMD will have moved onto RDNA 3.
Soldato
- Joined
- 26 Oct 2013
- Posts
- 4,236
- Location
- Scotland
60 CU for Navi 22 =
60 CU for biggest Navi =
Why does it matter how many CUs there are? If 60 CUs beats a 3080 does it matter at all that it's Navi 21?
Number e-peen and envy. makes me laugh that some keep going on about transistor count and density like it's a super important metric, and X company needs a die size of A in order to compete with Y company. It's one of many reference points to consider, not the be-all and and-all metric.
Transistor density and die size mean nothing if one arch is vastly superior to another, if 1 transistor is faster than 3 of another, who cares how big the die is?
For AMD die size matters because if they want to produce x cards that compete with the 3080 then the smaller they can make that competitor the fewer wafers used on GPUs and the more wafers in their supply agreement can be used for Zen3 and other higher margin products.
And we have a rough estimate of the xbox performance which has 52CUs and 1800mhz using the same architecture. Roughly 2080/2080s?
We could add 10-15% for a 60CU part and another 20% for 2200mhz clock speeds. Bit of IPC improvements on top. At that point it's gotta be in range of a 3080.
I entirely agree, but like I said die size is just one consideration, and the example you give is a business and resource one, not necessarily a performance one. And certainly not the only metrix that will determine if they can top the 3080.
- Status