-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
Next-gen PS5 and next Xbox speculation launch thread |OT5| - It's in RDNA
- Thread starter Mecha Meister
- Start date
- Status
If TFlops is less than 10 then marketing will be using multipliers. "2x better than..."
Skipping trying to explain why 8.5tf (rdna) > 12tf (gcn).
Skipping trying to explain why 8.5tf (rdna) > 12tf (gcn).
so basically, AMD has inedeed sacrificed compute for better performance, if its true that 5700 chip at 8.5TF is faster than a vega 64, then even that chip will fit in thurrot's leak of over 2x performance multiplier. if this is true, then everyone here should be dropping their expectation from the clouds they are back to reality of 8.5~9.5TF range.
also remember that we already know the power consumption of the 5700 is 150W and 5700XT is 180W so we likely wont surpass the 9.5TF figure.
They will just use performance .
Like this perform 3 or 4 times better etc etc .
Which takes us back to the original rumor from a few days ago that the Andaconda will be 2x the performance of Xbox One X...
I don't think there's been any confirmation for power draw.
Also 1750-1900mhz seems way too wide of a range for clocks, unless 1 model is watercooled.
Edit: seems like it's a normal range for their boost clocks but a 580 pretty much runs at the boost clock and not 100mhz below it.
Last edited:
i am pretty sure that got leaked 2 weeks ago.
is it really a "lower end" if its performing faster than every vega gpu on the market other than vega 7?
Well the lower end in the pure flops game. People wanted a high number to latch onto
I remember seeing TBP of 180W for the low end and 225W for the high end model but it hasn't been confirmed.
I think most people were on 10 to 12 train .
So it be a little less but they did change from GCN so that was always possible after that news.
This thread has been like 5 TF to 20+. It depends on what day and what time is it 🕵🏻♂️
Its not "lower end", though. Just that it seems Navi 5700 XT with 8.5TF at typical gaming clocks will perform ~equal with a Vega 64. It might explain why the dev kit spec leaks are coming in at 12-14TF. They currently have a Vega 64 to mimic performance?
It's seems very unlikely though because 12.5/8.5 is almost a 50% gain while AMD themselves have said 25%.
Yeah nothing at all makes sense. I really hope we have some answers by the end of E3. Maybe some more developers can comment on next gen.
Not great at the maths myself but didn't Lisa Su say at Computex that the combined architecture+7nm would result in ~50% better than Vega 14nm?
I might have that wrong, though....
I know that the 1.25x Navi-Vega multiplier has been claimed by the AMD CEO itself but I'd rather wait some independent benchmark before applying it indiscriminately . For example it could be up to 1.25x on selected titles
They gave both a 50% & 25% number: 50% increased performance/watt & 25% better performance per clock. Compared to what? No idea because AMD never said that. I'm assuming the better performance per clock is with the same amount of CUs otherwise it doesn't make a lot of sense to mention it.
That was 1.25x per clock - but aren't clocks higher again with Navi?
She claimed a 50% improvement per watt vs vega. Edit - I assume that comparison was vs Vega.
I'm not sure what flop performance a Navi with the same watt consumption vs a Vega 64 will be though - if the disparity will be so high or why that would be.
In the one game comparison they showed at computex, vs the rtx2070, you could perhaps extrapolate the performance of the Navi chip they were referring to is similar to a vega64. Maybe that's what the b3d poster is referring to? Maybe it is indeed a much smaller chip in terms of power consumption - and also paper flops - but we will see.
Yeah, there is still quite some confusion over this that will hopefully be mostly answered in the next 10-36 hours. That Brad Sams has said Microsoft will talk in 2X, 4X multipliers (as well as the 'performance leader' language in the Scarlett documents) fits with Navi having lower paper TF.
So,it seems paper TF number will be the last thing we will hear from MS and Sony cause it's tricky from PR/"power" perspective...
Yes, but that means 1.25x per flop unless you change the CUs and in that case it's a completely worthless number to give.
I think it is starting to look like that, yes. Show us the games I say.....
Sony already stated they are using Navi.
So that would make 9.2 tflops with 40 CUs at 1.8 Ghz performing like 11.5 tflops Vega
Good news being Navi is a real (next gen) new architecture from AMD finally able to compete with Nvidia, well maybe, we'll see.
EDIT: sorry bad formatting, Sony only confirmed Navi, not the Tflops number. I edited my post
Last edited:
i know but is the dev kit right now navi based? i doubt it . all these leaks are coming from dev kits that are vega based .
so if a leak sees a 12.9 TF vega based means navi PS5 and PS5 NAvi dev kits will have 10.3 TF of RDNA rather than GCN when NAvi is ready. same performance between a 12.9 Tf vega and 10.3Tf of navi,,,no?
Indeed, if confirmed, this is where we will go. An 8tf machine (at least for the PS5).
EDIT: It would give more credence to the 4tf Lockhart too.
If the actual perf for Navi flops turns out to be as different as is being suggested, then the voting from the last thread (Polaris/Vega flops) goes out the window anyway. It would be irrelevant because people who were predicting for a low flop number would be "right", and people who predicted for high perf would be "right' as well.
Last edited:
Stated by Cerny in the Wired interview and then by AMD in the Computex Keynote.
Last edited:
Compete on what? They are barely there with the RTX 2070. At a higher power draw, Navi pulls 225W versus the 2070's 180W.
This would be very awesome but I doubt it'll happen.. October/Nov 2020 more likely
So my final 'just for fun' prediction pre-E3 Microsoft/AMD.
Anaconda have:
- slightly stronger Zen 2 CPU 3.2GHz
- slightly stronger non-NAVI GPU @ ~12TF's*
- 20GB RAM (16GB GDDR6 & 4GB DDR4)
- Custom SSD
- non-hardware RT
- more likely £449 to £499
Which leaves PlayStation 5 to have:
- slightly weaker Zen 2 CPU @ ~3GHz
- slightly weaker custom NAVI GPU @ ~11TF's*
- 20GB RAM (16GB GDDR6 & 4GB DDR4)
- Custom NVMe
- Hardware RT
- possibly £399 to £449
*Comparative power
Lockhart is:
- whatever
Ha...so although Scarlet has a little more power PlayStation 5 has the ability to display more data (a bit like the throughput of PlayStation 2 v XBOX). Regardless, the perceptible differential between to two will be negligible overall.
Apart from my fun musings it's curious timing for AMD to reveal all about NAVI tomorrow. Considering it's after Microsoft's conference one would expect Scarlet to naturally have NAVI technology. Sony have already announced NAVI technologies so AMD can now spill all the beans once Microsoft confirm details, even if they are not using NAVI.
Also curious AMD want to start to talk about Ray-tracing...but maybe they can't until Microsoft tells all.
At this stage I think Microsoft have to go through some details to ensure AMD can stay on-point with it's own product timetable.
QUESTION: Did we find out why Sony's lawyers will be hard at work this weekend?
Anaconda have:
- slightly stronger Zen 2 CPU 3.2GHz
- slightly stronger non-NAVI GPU @ ~12TF's*
- 20GB RAM (16GB GDDR6 & 4GB DDR4)
- Custom SSD
- non-hardware RT
- more likely £449 to £499
Which leaves PlayStation 5 to have:
- slightly weaker Zen 2 CPU @ ~3GHz
- slightly weaker custom NAVI GPU @ ~11TF's*
- 20GB RAM (16GB GDDR6 & 4GB DDR4)
- Custom NVMe
- Hardware RT
- possibly £399 to £449
*Comparative power
Lockhart is:
- whatever
Ha...so although Scarlet has a little more power PlayStation 5 has the ability to display more data (a bit like the throughput of PlayStation 2 v XBOX). Regardless, the perceptible differential between to two will be negligible overall.
Apart from my fun musings it's curious timing for AMD to reveal all about NAVI tomorrow. Considering it's after Microsoft's conference one would expect Scarlet to naturally have NAVI technology. Sony have already announced NAVI technologies so AMD can now spill all the beans once Microsoft confirm details, even if they are not using NAVI.
Also curious AMD want to start to talk about Ray-tracing...but maybe they can't until Microsoft tells all.
At this stage I think Microsoft have to go through some details to ensure AMD can stay on-point with it's own product timetable.
QUESTION: Did we find out why Sony's lawyers will be hard at work this weekend?
Last edited:
i m sure he is talking about 40 CU and 9.2 Tf number you threw out there as part of official detailsStated by Cerny in the Wired interview and then by AMD in the Computex Keynote.
Ok, I want to be ridiculous here. I want MS to go crazy. Lockhart is supposed to match the PS5 and Anaconda is a 700-800 dollar beast. That would make me buy the anaconda immediately.
It's curious that AMD still can't talk specifics about their collaboration with Microsoft, they openly boast about their Stadia and PS5 involvement, but can't seem to do the same with Microsoft, who here thinks MS is using a GPU from NVIDIA?
I agree, Nvidia architectures are still in their own league. I remember when Vega was supposed to close the gap with pascal, the magic of double rate FP16, the HBM etc.
Nvidia has still both the performance and efficiency crowns firmly on its head no matter what AMD pulls off
Sorry. The second part was only me speculating, I edited my post.
No one. Zilch.
The only reason they haven't is because next gen Xbox hasn't been revealed yet.
Not to mention they have a contract with AMD.
MS still wants to sell some Ana systems.
Think there was document with MS thinking the ratio could be 1\3 for Ana sales wise and that won't happen at that price.
Yea, when i saw the 70/30 targeted split, it became obvious the price target for anaconda is 499. Higher than that and you wont see this split.
I'm not sure I totally follow.
What I took from Su's comment is that Navi, for a given die size, at - say - 100Mhz will be 1.25x the performance of GCN at 100Mhz.
But if Navi also lets AMD clock higher than GCN - at a given wattage - then the performance delta might be larger still. For a given TDP you might have a Navi at 120Mhz and the GCN part at 100Mhz. Leading to an overall per-watt gain of 1.5x that she talked about.
The missing piece of the puzzle is what Navi at a given die size, and a given power usage, at that clock rate actually translates to in terms of flops. And it seems to be suggested than on paper it may be less than the same size/wattage GCN parts. Perhaps that's due to changes to the cu/sps. If that suggestion turns out to be true, I guess we will get a better idea of that and why that is during/after AMD's E3 brief, and why the real performance gain is larger despite that.
I'm hoping for a surprise early release for one of the xbox skus. (at least 6 months before PS5)
Otherwise I'll be getting a PS5 and an xbox further down the road, further down the road turned out to be never this gen for xbox.
PS5 is launch day regardless.
I'd say a lot of other non Americans would be in a similar mind set.
Otherwise I'll be getting a PS5 and an xbox further down the road, further down the road turned out to be never this gen for xbox.
PS5 is launch day regardless.
I'd say a lot of other non Americans would be in a similar mind set.
- Status