| HollyGamer said: The TDP i mentioned is 330 a typo it's the same number that mentioned on Anand tech (310w) |
Hence why I thought I would provide clarification.
| HollyGamer said: Ok, but we don't have an exact architecture lay out of Navi, AMD might have something on their sleeve like more Rop's on single CU. It's the same design GCN but it's still "Navi" a next gen mainstream GPU to replacing Polaris. |
We do have an exact architectural layout of Navi. It's called Graphics Core Next. - That is the GPU architecture.
It's not next-gen, it's not a dramatic departure to what we have witnessed over the past 7 years it's an iterative update.
Next-gen graphics technology doesn't happen until AMD deviates from the 7+ year old GCN architecture.
| HollyGamer said: Even PS4 to PS4 Pro can have a graphic boost by 2,5 times more. |
You would hope that the Playstation 4 Pro can boost GPU capability by that much.
Considering that it is:
CPU Clock: 33.12%.
Theoretical GPU GFLOP performance: 127.7%
Pixel Fillrate: 127.7%
Texture Fillrate: 127.7%
Raw Bandwidth: 23.6%
The Pro also implements various features like Delta Colour compression as it inherits a few new tricks introduced into Tonga.
Which means Bandwidth could potentially be up by 73%.
And it's no wonder why most games are sitting around the 1440P mark with a good uptick in fidelity and framerates.
| HollyGamer said: what's the point of 7nm if there is no performance gain. You seem pessimist, while historical data pointed different way. |
I never once claimed there was no performance gain.
| HollyGamer said: If you " theory " is true, PS5 performance will not have a generational leap, it will be even smaller leap compared to PS3 to PS4 gap, while PS3 to PS4 already considered small. |
The Playstation 5's performance will be a significant one over the base Playstation 4.
It will be a much smaller step over the Playstation 4 Pro... Which is the downside of releasing a mid-generation console.
Because Sony isn't investing into the development of chips directly for it's consoles anymore (And nor should they, it's prohibitively expensive!), Sony is thus at the mercy of PC's microprocessor development cadence, that means... If AMD stagnates, then so do consoles... And AMD certainly has stagnated... Part of that is because AMD made Radeon into it's own separate entity again after they combined them with their CPU teams... Plus they had to cut a ton of costs due to profitability so they lost a ton of engineers over time.
They are slowly reversing all that, but we won't see the real fruits of that positive change until after the next-gen consoles launch.
| HollyGamer said: If you said using Zen2 is big improvement , you said on others thread, it's not. CPU is not affecting rendered pipeline directly, it might rid of bottle neck but it will not affecting the shaders count etc. |
I didn't say anything about whether Zen 2 is a big improvement or not in this thread?
| HollyGamer said: You also need to remember what happen with Polaris , it's able to gain more performance compared to previous Pitcairn mainstream GPU and comparable to Hawaii high end GPU or even to Nvidia Maxwell (GTX 970- GTX 980). |
Every iterative update Graphics Core Next generally gets an uptick in performance when compared to the previous generation.
As for Polaris vs Maxwell... Polaris still looses to Maxwell in terms of performance... Where a Geforce GTX 970 will more often than not, beat the RX 480.
But in doing so, the Geforce 970 consumes only marginally more power... This is despite the fact that Polaris is built at 14nm verses the Geforce 970 being built at 28nm. Ouch.
nVidia is simply far ahead of AMD in the GPU space.
Things even start to look worst when you throw the Radeon RX 580 or RX 590 into the equation against the GTX 980... As AMD pretty much threw away the idea of releasing Polaris on it's optimal efficiency curve and just drove up clockrates instead.
| HollyGamer said: I believe we are on the times where 2016 high end GPU will be a mainstream GPU like what happen with polaris. So 14 teraflop is achievable .Hell even 2016 AMD able to bring 12,5 teraflop with vega 64. |
Is 14 Teraflops achievable? Sure. Is it likely? Probably not for Navi as Navi is a Polaris replacement targeted for the mainstream and not high-end.
I don't think people fully comprehend how terrible AMD's GPU's are right now for gaming... For compute they are the ducks nuts.
| HollyGamer said: I wouldn't hold my breath, everything is possible until we have a concrete data is just pure speculation , but based on the current data i have my own theory, i am not close to any possible result either. |
Well, no. Everything is not possible, we are still bound by the laws of physics and common sense.
There comes a point where a chip simply becomes to large to financially be viable at 7nm, especially as the process is still fairly young.
| HollyGamer said: 4k on 60 FPS is possible , hell even 1070 can do that |
That's because nVidia is far more efficient. It's chips are smaller, less power hungry and perform far better than the AMD equivalent.
| HollyGamer said: your reason that we have 3 TF performance GPU on 2010 is funny , 3TF on that year is a high end GPU, 14 teraflop in 2019 to 2020 will be a mainstream GPU so it's possible we have that number on navi. |
Flops alone is irrelevant.
Back in 2010 we were playing around with Terascale 2 and Terascale 3 hardware for the AMD side... Which leveraged the VLIW approach... And that generally meant it had multiple caveats, some of which were performance-impacting in newer titles, but generally were extremely solid performers in titles that leveraged older techniques that were common in games during that era.
Ironically... I have such a GPU sitting on the shelf looking at me right now, that was back in the days before AMD threw out it's small-core philosophy and was on top of the game when it came to performance.
| Bofferbrauer2 said: A Vega 64 has 64 ROPs that's true, along with 4096 Unified Shaders, 256 TMU and organized into 64 CU. I think remembering the actual limit comes from the Shader logic limiting those to 4096, but due to how it's organized that results into 64 CU, while the ROPs could be increased. Hence why there are rumors about an increased number of ROPs in Navi but no increase in CU. |
Increasing the ROPS isn't an easy thing either though, you need to do a full rework of the rest of the chip (memory included!) to take full advantage of it, that is costly.
Instead, with Vega 7... AMD increased ROP performance by dialing up the clockrates.
In saying that, I wouldn't be surprised if Vega 7 was still ROP limited, I know it is certainly geometry limited, despite there being a substantial uptick in geometry performance... But their decision to stay with 4x Geometry Units isn't doing them any favours.
I feel like Graphics Core Next is simply unbalanced for today's games and AMD needs to do a full rework rather than just iterative updates with a couple new features tacked on, that will hopefully happen after Navi is done and dusted.
| Bofferbrauer2 said: PS4 Pro chip is about the same size as the original PS4 chip, which is about 350mm2. A Vega VII, which is produced in 7nm, is already is 331mm2. How do you want to reach that power in a ~350mm2 package if 95% is already occupied by the GPU - and that GPU isn't even strong enough for the leak. While Navi could be more powerful, it will need more transistors, and thus die space, for that. Getting that much power out of such a small chip is pretty much impossible with the 7nm process. |
400mm2 is probably the upper limit of what we can expect from a chip size-wise for consoles, irrespective of node... Otherwise chip costs will increase at a non-linear rate.
| Bofferbrauer2 said: Actually March 2011 (March 8 to be exact), and it's a dual GPU, the Radeon HD 6990, so those 5 TFlops are highly theoretical Terascale was actually not nearly as bad as you seem it to remember, especially the 4xxx and 5xxx series did compete pretty well with NVidia. In fact, at the time AMD was making fun of NVidia being so unefficient and running hot (Termi, anyone? Here in a Hitler rant ^^) that they even made videos about it (can't find them now). GCN was of course much better, but Terascale wasn't that bad either. You could also have come with the 7990, also a dual GPU, but with 8TFlops in May 2013. Or that a 12TFlops GPU existed in 2014 already (R9 295X2, 500W TDP). Or a 16TFlops GPU in April 2016 (Radeon Pro Duo, 350W TDP). All of them are dual GPU, btw, which makes the TFlops highly theoretical. |
Flops is always theoretical.
The Radeon 5000 series certainly stood the test of time, that's for sure.
It wasn't until games like Battlefield started pushing more and more geometry that the older Terascale parts started to show their limitations in games as they were extremely geometry limited.
| Werix357 said: Yeah I think including a SSD would drive up the price to high I just hope they the hardware is made with option to upgrade to m.2 SSD's 3000mb/s would help remove most loading and I imagine open world games would less of a technical challenge. |
Well... We are also on the cusp of next-generation mechanical hard drives right now... Drives with heat-assisted magnetic recording, helium... And/or Multiple actuators and can do 500MB/s or more in sustained transfers.
Whether next-gen will leverage that remains to be seen... But even improving over the base Xbox One/Playstation 4's archaic, slow and terrible 5400rpm 500GB drives (That generally do 60-80MB/s) is probably not a hard ask even with budget drives.
--::{PC Gaming Master Race}::--
