Discussion in 'Videocards - AMD Radeon' started by OnnA, Jul 9, 2016.
You guys think AMD will show us the rest of the Navi lineup beside rx5700?
Maby 5750 & 5770 but we will see (1 day)
Im more into -when- they give us Async Path-Tracing (1440p BFV at 70FPS will be nice)
*when not if*
So we have ATI 9000 name sheme (ATI 9700/9800):
5700 Pro, XT & XTX
5800 Pro, etc.
Which links to this tweet.
Though it sounds like the official unveil and announcement is pretty close also.
So possibly the 5700 XT
40 compute units.
9.75 Terra Flopps
8 GB GDDR6
1905 Mhz boost ("Up to")
1755 Mhz game clock speeds.
1605 Mhz stock clock speeds.
Guessing it tries to boost a bit ("game clocks") but it can go higher if there's room potentially another 150 - 200 mhz and whatever that translates to in terms of performance.
AMD Radeon RX 5700 XT Specs Leak Out: 2560 Cores, 8 GB GDDR6 VRAM, 9.75 TFLOPs Compute Power
The AMD Radeon RX 5700 XT is based on the 7nm Navi GPU which makes use of the latest RDNA core architecture.
According to AMD themselves, the RDNA architecture brings a revamped CU or Compute Unit design which should deliver much better gaming performance than the existing line of GCN based graphics cards.
Some of the key upgrades on Navi GPUs will include:
New Compute Unit Design
Multi-Level Cache Hierarchy
Streamlined Graphics Pipeline
In terms of specifications, the Radeon RX 5700 XT comes with 40 compute units in total and since AMD has already confirmed that the compute unit design still features 64 stream processors, we will be getting 2560 stream processors in total.
The chip itself is clocked at 1605 MHz base clock but includes two additional clock speeds, a boost clock, and a game clock. The boost clock is rated at 1905 MHz while the game clock is rated at 1755 MHz.
What I can make of the differences from these two is that the boost is the maximum speed the reference card would be able to hit but the average clock speeds while gaming is the ‘Game clock’ itself.
With the said boost clock, AMD expects a maximum of 9.75 TFLOPs of single-precision compute from the Radeon RX 5700 XT. The card is also confirmed to feature 8 GB of GDDR6 memory that should run across a 256-bit wide bus interface.
It will be interesting to see if AMD chooses the same 14 Gbps GDDR6 dies as NVIDIA does or go for the more cheaper alternatives that operate at 10-12 Gbps.
In addition to that, it should be pointed out that since there is an XT variant, there should also be a cut-down Pro variant which has always been the case for AMD graphics cards.
So it won’t be a surprise if AMD also has a Radeon RX 5700 Pro ready for announcement during their E3 ‘Next Horizon’ event.
100% Confirmed NAVI is with 'Ray-trace' capabilities in H/W.
Next Gen Xbox - 24 Tera Flops, dedicated Ray Tracing hardware - all our PCs will be obsolete
Microsoft Conference Sunday, June 9th 1:00 PM 4:00 PM 9:00 PM 10:00 PM
41.21m Cyberpunk 2077 & Keanu Reeves
Gears of War 5 (Gameplay):
Bethesda Conference Sunday, June 9th 5:30 PM 8:30 PM 1:30 AM (Monday) 2:30 AM (Monday)
Next-gen Xbox Scarlett Announced at E3 2019; Features Custom AMD Zen 2 and Navi Processor With 8K Visuals
Microsoft has finally announced its next-generation Xbox console during its 2019 E3 press conference.
Codenamed Project Scarlett, Microsoft intends to launch the new console in holiday 2020 and is said to offer around 4 times the power of the Xbox One X.
“The console should be optimized for one thing and one thing only,” Microsoft’s head of Xbox, Phil Spencer said. “Gaming.”
Project Scarlett with sport a custom AMD “Zen 2 and Navi processor and will be able to support framerates up to 120 frames per second and 8K visuals.
In addition, Spencer mentioned that the console will support next-gen Ray Tracing technology and it will pack a new generation SSD that is said to greatly increase loading times by 40x times over the current generation consoles.
Microsoft’s new console will also use the SSD as virtual memory. We’ve included the Xbox Scarlett reveal trailer down below:Introducing Project Scarlett.
Unmatched power and speed ushers in a new level of game play performance and the future of gaming.
With four generations of content that looks and feels unlike anything before. Launching with Halo Infinite Holiday 2020
Some new Leaks
Prices will be lower IMO
499€->649€ price bracket will be reserved for Arcturus HBM3.
I wonder if crossfire will be supported.
Those "Navi 12" leaks make little sense on lowest part. Would be 1st time AMD cut off 40% of actual GPU. That's such a waste unless both defect rate and ability to recover GPU was rather high.
But against that stands "Navi 10" which seems to have very good yields. As it cuts only 14% in worst case.
Then there is "Navi 20" which has double ROPs of Radeon VII, but same number of TMU/SP/... +50% wider IMC than "Navi 10/12". And they think that it is going to be up to 30% bigger than Radeon VII.
What is conclusion? Three variants of "Navi 10" are believable.
Lowest Variant of "Navi 12" is not believable. (RX-580 is cheaper to make, and defect rate on 7nm is unlikely so high.)
"Navi 20" variant is not believable at all since memory itself will cost 50% more than ones for "Navi 10", And GPU itself is considered to cost 20~50% due to lower yields for bigger GPU.
Then there is their speculation on performance. If top variant of "Navi 10" was as they wrote. It would outmatch Radeon VII in gaming due to 112 vs. 64ROPs and improved frontend without need to have RDNA CUs.
- - - -
Only thing which may go for this speculation is that They state "Navi 10/12" is not having DX-R acceleration HW. (Which I would read as CU configured in GCN compatibility more instead of improving FP16 at cost of other compute. But then again "Navi 20" is not really bigger enough than "Navi 10", therefore "Navi 10" could have already been reconfigured.)
- - - -
What interests me most is Number of SPs per TMU. Patents show clearly that it no longer needs be 16SPs per TMU as with GCN. And unless that "Navi 20" changes this ratio, it is again in GCN compatibility more, therefore DX-R optimization would seem rather weird.
- - - -
My expectation was that "Navi 12" would be in GCN compatibility mode as there's only that much silicon space one can use to trade off for improved DX-R. And that "Navi 10" would already have reconfigured CUs.
(I personally want GPU which is not in GCN compatible mode to reap full benefit of new architecture and for GPU to last longer. If neither of "Navi10/12" is, then I may just get "Navi 12" as long as performance improvement over my RX-580 is reasonable as mine runs at 145W top.)
I wonder for how long I have to wait. Well, tomorrow, I'll know what I need/want to know.
- - - -
It is nice to know that I do not know actual configurations. And it is nice to know that those speculators are in same boat
PC Gaming Show Monday, June 10th 10:00 AM 1:00 PM 6:00 PM 7:00 PM
Ubisoft Conference Monday, June 10th 1:00 PM 4:00 PM 9:00 PM 10:00 PM
AMD Next Horizon Gaming Monday, June 10th 3:00 PM 6:00 PM 11:00 PM 0:00 Midnight
Square Enix Conference Monday, June 10th 6:00 PM 9:00 PM 2:00 AM (Tuesday) 3:00 (Tuesday)
Found here: https://twitter.com/SpartanGeekTV/status/1138091993459056640
Hmm AMD caught up to the performance level of NVIDIA in Assassin's Creed Odyssey, guess they have a pretty hefty CPU in the test rig but even so the driver situation for D3D11 showed that game as one of the worst performing titles between AMD and NVIDIA so that's curious to see.
Wildlands too even if the situation wasn't as problematic in that game.
Those caught my eye too. As well as Civ6 which they are claiming parity (as there is rarely any difference in game play between 1%).
I have to wonder if there is something in the drivers that is helping the 5700?
And, 115% in Exodus at 1440p?
If all this is true it's looking pretty salty to me. And Nv is answering with their "super" series of cards which seems to confirm these improvements IMO.
And don't forget OC headroom, if massive Radeon VII could be OC to ~2100~2200 then this card can get around +20% with good cooling.
Let's hope for good pricing, don't care if only Blower comes first- Got new Accelero Extreme III in box ready for action .
In BF5 and Metro Exodus that is around 2080 level performance. Interesting if this is true.
Not that long until the stream now...
Not in the market for a new card personally but it sucks to have to wait for the 3rd party models, I wouldn't recommend a stock blower from AMD to my worst enemy.
Yeah I'm no expert but from a few weeks of reading up on the driver situation and features like driver command lists and multi-threading, NVAPI and deferred contexts and render threads and much more even if this isn't a entirely CPU limited 1920x1080 or lower run for 2560x1440 or even 3840x2160 the D3D11 titles especially Anvil Next should be fairly rough on AMD compared to NVIDIA.
Tests from earlier this year although this was for 1920x1080 put Assassin's Creed Odyssey at a 40% performance deficiency with a AMD Vega 64 against a Pascal 1080 I think it was with most other tests hitting a 5 to 15 % performance lead or penalty and then Anvil Next and the use of eight additional threads resulting in a very different CPU situation with NVIDIA and NVAPI against AMD and a different overall performance figure as long as the NVIDIA GPU system has a CPU that can carry this workload. (Hexa works but Octa ensures it's smooth.)
NVIDIA spent around two years I think possibly more adding in and updating the D3D11 driver code for better multi-threading and support for D3D11.1 feature driver command lists in addition to what NVAPI allows for which AMD in 2016 discouraged DCL for 11.1 (GPU Caps Viewer should still show it as unsupported as a optional 11.1 component or feature.) and driver wise I've heard AMD is at most dual core plus how threading and the command cues and lists are all executed and dealt with although performance isn't a total loss against NVIDIA even so but with focus on Vulkan and D3D12 even if adaption hasn't been that quick or smooth likely hasn't seen any major driver changes for D3D11 and it's code though who knows.
VLK and D3D12 has some of this as mandatory features whereas 11 has that and other as optional plus solutions like NVAPI doing it's thing in various ways (Not too sure about all of it or well not sure on most of it really how it works.) but we'll see next month as actual reviews do come in and some of these games here should still end up in the benchmarks we're going to get from Ryzen 3000 on Zen2 and NAVI and it's changes on the GPU side too.
Or that's what I'm thinking at least, will be interesting to see how it does and well maybe another four or six months and how later drivers could make a difference too as well as usual really I guess though maturity and level of difference this makes can vary as can whatever AMD is changing either listed or unlisted or changes to the GPU functionality via the driver and changes therein.