Hence why 'compute' is irrelevant for gaming and why the 1080Ti, RTX 2080 and 2080TI spank the Vega VII in gaming benchmarks even though AMD like to flaunt the Flops numbers about just like Sony do with their consoles. The reason a lot of articles dont even mention it is because it's meaningless when other card specifications affect that throughput. The numbers you mentioned in that last post would indicate twice the performance........in theoretical maximums involving memory bandwidth then yes, but in real world performance, and certainly gaming (where the card is aimed at, its behind the same priced competing cards) then no.

There's no change again.....it's getting very tiring watching AMD cherry pick certain sets of numbers that make their cards appear as if they are competing at the top end even though they are yet to manufacture a card of recent that does anything other than offer bang for buck. If it's not HBM, GFlops or memory bits it's something else. They aren't competing at all so they aren't forcing Nvidia to drop their outrageous pricing of high performance GPUs.

The problem with Vega VII is that most people are aware it's just a rehashed, gimped 7nm Instinct MI50 that's the same price as a 2 year old Nvidia GTX 1080ti but still cant reach the same gaming performance (I'm not sure why AMD claim it's competing with the RTX 2080 as it doesn't have the same feature set). We'll see what Navi has to offer in time but patience is wearing thin after so many promises over recent years, and the inability to break Nvidias performance dominance. By the time Navi does appear, Nvidia will have increased the performance watermark yet again and moved to smaller and more efficient fabrication. Nvidia appear to have performance to spare but AMD are always running at their max ceiling.

I try to be un-emotional when evaluating.

To state the obvious:

In every competition someone comes in first and someone comes in second.

The fact that AMD comes in second a lot is not a bad thing -- because the difference (between similarly positioned items) is usually not visible to the unaided eye (its only measurable). For example, in a blind gaming test of Vega 64 vs GTX1080 -- half the players picked Vega 64 and half picked GTX1080 because they really could not see the difference. For example, I can play any game on ultra settings with my setup. Usually, when possible, I mod the games to be even more graphics intensive -- with rare exception they stay on ultra. I go by what I see -- not what I measure.

If a gamer gets enjoyment knowing their GPU gives more FPS even though it looks no different -- fine -- there is no wrong way to enjoy gaming.

To compete, AMD charges lower prices. Without that competition what would "unopposed" Intel and "unopposed" Nvidia be providing. How much slower would progress have been. And, how high would "unopposed" prices be?

Competition has winners and losers -- but, its better than the alternative. So far, AMD is holding its own.

Competition is good

But that's the real problem with AMD, they've promised so much with Vega over the last few years but failed to compete at the top end, and then had to backtrack and claim they are offering bang for buck every time.

As you mentioned games don't look any different but the key factor is always the minimum framerate. With VR and 4K resolutions becoming more and more popular it's only Nvidia that can support the 1440p @ 90fps and 4K @ 60fps in the demanding titles. Another reason why the Vega VII isn't going to cut it at the $699 price point......the 1080Ti scrapes in for most games whilst the GTX 1080 and RTX 2080 fall short, which is exactly where the VII will drop in. AMD has no GPU to offer that level of performance.

Competition will be good when AMD actually get there. They are great at throwing irrelevant performance numbers around but not so good at delivering cards that actually compete in-game hence why Nvidia are under no pressure to lower the prices of their top two tiers of cards.

The best bang for buck at the moment is a used 1080Ti at ~$350-400 as it's only the RTX 2080Ti that is beating it at a ridiculous retail price.

In keeping with the title of the thread: The Future And Current Status of AMD

Looking at Steam GPU usage figures for December 2018 (most current):

Overall:
Nvidia 74.18 percent of all GPUs used.
AMD 15.08 percent of all GPUs used.

Selected Models:
GTX 1060+GTX 1050Ti+GTX 1050 = 29.86 percent of all GPUs used (most popular price/performance grouping by far)
RX580+RX480+RX470 = 1.52 percent of all GPUs used (same price/performance grouping)

GTX 1080Ti = 1.68 percent of all GPUs used (more than AMD's above three mainstream combined)

RX Vega 64 and RX570 = not enough to make list -- lumped under "other". Definitely under 0.15 percent each.

Those numbers indicate why AMD is rumored to be planning a sub $300, new design 7nm Navi for later this year as their first "serious" 7nm GPU -- that price/performance group is what sells best. As noted above in this thread, Radeon VII is a 7nm "place holder" just to keep their name in the news (competitive -- but not special performance vs Nvidia's similarly priced RTX2080).

Hopefully, the pundits correctly analyzed the "leak". Nothing official yet. Repeating: the AMD 7nm Navi GPU is "supposed" to be a new design, under $300 GPU (some say $250) with performance of RX Vega 64 at low power use (no proof, just educated rumor). Navi may also be used in 7nm APUs for thin laptops (AMD hints laptops will be an area of specialization and competition). We'll see this Summer.

We were just talking about AMD's exceptionally fast "compute" being used to boost games of the future. That future is previewed March 15th 2019.



They say compute is a "strength of AMD GPUs across the board" because AMD GPUs have had DX12 and compute built in for years now. Even older AMD GPUs should perform. We'll see how well when the reviews are written.

More AMD compute and using 16GB memory on the GPU -- an RX570 for cryptocurrency mining. Sounds like a year ago -- shudder



So, buy a 7nm Radeon VII 16GB HBM2 GPU. Use it for GRIN Coin mining in its "spare time". It pays for itself. I'm kidding -- though its a thought

Take the attached with more than a grain of salt -- maybe a few hundred pounds. Mostly guesswork. But, it is in keeping with the rumors/leaks.

In one simple figure it provides a snapshot of what might happen this Summer.

I had those two on my main machine the last couple (?) months. My home-made fixes did not always work. Wondered if it was "me" or "them". It was "them" for a change. If a reader has had the same issues, the driver update might fix it.

These will be custom made console (not PC) APUs -- with the core, thread, and shader counts specified by XBox Next and PlayStation 5. The counts will be different in the two gaming consoles -- so, performance will not be identical.

Since Ray Tracing is part of the discussion now, I've conducted some of my own observations.

3DMark has a ray-tracing benchmark out (only a few dollars to add it). I watched the video of it (my card does not have ray tracing). Many fully reflective surfaces. Looks very realistic. Runs at 1080p on RTX2080Ti (they say). Anyone can go to Steam page and watch the video.

Played Assassin's Creed Odyssey main campaign and now back on A.C. Origins (never finished it). No ray-tracing -- but, has reflections (mostly on water surfaces). The reflections look "real" when playing. I have to stop and stare to see the non-real aspects. However, if all the graphics were fully ray traced, I imagine I would easily see (and like) the difference.

Full ray tracing is currently not possible because it costs far too much GPU power -- rather developers must pick and choose what aspects to ray trace (I have read). And, there are other ways to get reflections/shadows/lighting-effects that do not cost much GPU power (e.g. those used by A.C. Odyssey/Origins). For ray tracing, beauty will be in the eye of the beholder -- no wrong way to enjoy gaming.

The article indicates that Radeon VII was logical to keep AMD "in the news" (something we've noted). But, the profit is very small. In that limited sense (they are not making a lot of money off the Customer), its not a bad deal. It releases in less than two weeks.

I'll be interested in the independent test results. However, if its only roughly 25 percent faster than what I've got, I wouldn't "see" or "feel" the difference.

Radeon VII out next week may be being sold at close to estimated design+production cost. Again, the concept is that Radeon VII is just to keep AMD "in the news" -- not to make real money. Think of it as an advertising expense. So, based on the guesswork in the article, customers are getting a bargain -- particularly if they use the compute functionality.



Article

My understanding is that the new Ryzen 3000 series CPUs probably (not for sure) will work with older X300 series and X400 series motherboards. Nominally plug and play. But, I won't be at all surprised if a BIOS update is required.

Depending on performance and pricing, my tentative plan is to buy a 7nm Ryzen 7 3800X (??) to replace the Ryzen 7 1800X in my 470 motherboard. But, I'm open to 570 if it actually makes things work better.

There's a chance some Nvidia GPU prices may rise due to supply shortage. Its not clear what model number GPUs were affected.



AMD has chips made there also. So, AMD may have been affected -- but not reported, yet. The article "speculates" that AMD was not affected.

The first Intel GPUs are planned for 2020 release. As often happens with a totally new product, Intel GPU release to market may be delayed significantly. But, I too expect a good product.

For comparison: