Fudzilla: AMD Navi is no high end GPU

[TheoneandonlyMrK]

Sounds great, but Titan V is out now and not being poked around with in a lab... actually i take that back, GV100 is being used by reseachers all over the world right now.

Let's hope AMD deliver.

Amd dont tend to charge consumers 3 grand and also have vega 7nm, shrunk instincts for them with full speed double precision and packed math ?? that should make for quite the ai engine.
but this could also cater to both angles ,who knows.
Titan V is out now so is 1080Ti etc etc , vegas and polaris sells still,some vote with their wallet, others make their own choice life goes on.

 

[Fluffmeister]

Amd dont tend to charge consumers 3 grand and also have vega 7nm, shrunk instincts for them with full speed double precision and packed math ?? that should make for quite the ai engine. but this could also cater to both angles ,who knows.

AMD don't tend to offer anything competitive when it really matters, at least not recently. They have tried their luck with rad pro duos and other such nonsense, but hey ho.

Compared to aging overpriced Hawaii based FirePros like the W9100 Titan V is a bargain.

Titan V is out now so is 1080Ti etc etc , vegas and polaris sells still,some vote with their wallet, others make their own choice life goes on.

Life does indeed go on, despite all the tears and tantrums.

https://www.anandtech.com/show/12741/nvidia-announces-record-q1-fy-2019-results

 

[TheoneandonlyMrK]

AMD don't tend to offer anything competitive when it really matters, at least not recently. They have tried their luck with rad pro duos and other such nonsense, but hey ho.

Compared to aging overpriced Hawaii based FirePros like the W9100 Titan V is a bargain.



Life does indeed go on, despite all the tears and tantrums.

https://www.anandtech.com/show/12741/nvidia-announces-record-q1-fy-2019-results

Yeah well that told me eh, moan about last gen Amd cards then show Nvidias financial results, And.......

Amd did well too but it is irrelevant to a debate about the next gen from Amd as is hawal??.

 

[Melvis]

In the real world high end GPU's just dont sell that often, in all the yrs I have been building PC's the biggest baddest GPU's that I have sold for clients or put into gaming PC's is GTX 970/GTX 1060's and 1 GTX 295 Many yrs ago but that was for a mate of mine. The normal crowd of people where I live anyway just cant afford the super high end.

 

[Fluffmeister]

Absolutely, I just hate to see people continuing to wait for Vega... now at 7nm, and a product not even meant for consumers, quite the AI engine though!

 

[TheoneandonlyMrK]

Absolutely, I just hate to see people continuing to wait for Vega... now at 7nm, and a product not even meant for consumers, quite the AI engine though!

I didn't wait and im happy with my one at 14nm, but this threads still about Navi, not Vega or Nvidia or Ai or what any company earned last year.

Any chance you have some thoughts on Navi possibly being ,polaris's replacement??

 

[Divide Overflow]

Any chance you have some thoughts on Navi possibly being ,polaris's replacement??

If Navi is supposed to be a multi-chip module design philosophy, it's possible it could be be both the mid and higher end solution. Just scale the number of GPU modules to meet the performance target.

 

[xkm1948]

MCM in theory sounds nice, just like many of RTG's promises that were supposed to "revolutionize gaming" while in end never picks up any steam.

IMO RTG will probably just stay in the mid~low end performance range in the next 4~5 years. They bet a lot on the ditection of GPU accelerated computing hence the entire arc of GCN. Too bad most of those fields prefer Nvidia's CUDA. The only compute GCN is good for apparently is crypto-mining.

RTG need to take a good look at itself and decide on the direction. Do they wanna focus on AI and scientific computing where Nvidia is absolutely dmoniating, or do they want to focus on gaming where they may still have a slim chance of catching up?

Also open standards everything ain't gonna work. At the end of the day most companies and research institutions prefer solid tech support and Nvidia offers some of the best tech support for researchers and developers.

 

[Vya Domus]

Open standards work just fine , stuff like OpenCL is used more than most even realize.

It's not that they have a chance at anything, they are already invested in pretty much everything. They are doing fine , they can't increase their marketshare by a few dozen % overnight.

 

[TheoneandonlyMrK]

MCM in theory sounds nice, just like many of RTG's promises that were supposed to "revolutionize gaming" while in end never picks up any steam.

IMO RTG will probably just stay in the mid~low end performance range in the next 4~5 years. They bet a lot on the ditection of GPU accelerated computing hence the entire arc of GCN. Too bad most of those fields prefer Nvidia's CUDA. The only compute GCN is good for apparently is crypto-mining.

RTG need to take a good look at itself and decide on the direction. Do they wanna focus on AI and scientific computing where Nvidia is absolutely dmoniating, or do they want to focus on gaming where they may still have a slim chance of catching up?

Also open standards everything ain't gonna work. At the end of the day most companies and research institutions prefer solid tech support and Nvidia offers some of the best tech support for researchers and developers.

Always rather negative but i get you, your Opinion is not the same as mine though, consider this, their GPU tech Has still been all sold off ie what Amd made, they sold , pretty much.
If that got better well that'd be nice, but in reality Vega 64 is 10-20 percent slower then a 1080ti not half or less the performance , so a gap that is not anywhere near the gap intel had on them in x86, and they caught that up enough to make intel stock owners worry, and to sell a boat load of chips.

And how did they do that, Innovation , Risk, and doing it first, as ever.

their modular and 2.5D designs are presently unmatched from the point of view of being out there ,in use and fully tested , others are just now getting onto this band waggon.

Apply Zen type modularity to their GPU's and they could be great, they might not be but I would hope for it personally, because the innovation on team greens side has all but gone in consumer gaming space, they used it all up on Ai and data center grade interfaces etc.

and finally Nvidia cards are equal to Amd cards at mining?? and more Gtxs are mining then radeons I would also wager, Amd do some Algos better as do nvidia but Nvidia have caught up some what with miner software optimisations such that their Roi is now better , though still , mining is presently hard to justify initial investment on imho.

@Divide Overflow Agreed , thats what I have been saying.

 

[JRMBelgium]

I live by a rule when it comes to GPU's. I wont spend more then 500$, I only spend it on AMD hardware and I want at least 50% performance increase.
This ment I had to wait a few years to upgrade from my R9 390x to Vega 56. This also means that I'll probably wont upgrade my GPU next year and wait one more year.
But since I'm playing on 1080p and I'm not planning on changing that, I expect I wont run in to problems with my overclocked Vega 56.
I only play Battlefield. And eaven if the next game is 30% more demaning, I'll still be able to run it at 60fps+.

Nvidia can release whatever they want. I wont contribute to their monopoly.
And with my x470 motherboard, I expect to upgrade to Ryzen 2 and perhaps Ryzen 3 after that.
There was no excuse to buy AMD hardware a year ago. But these days, there is no excuse not to buy AMD.

 

[Vayra86]

I live by a rule when it comes to GPU's. I wont spend more then 500$, I only spend it on AMD hardware and I want at least 50% performance increase.
This ment I had to wait a few years to upgrade from my R9 390x to Vega 56. This also means that I'll probably wont upgrade my GPU next year and wait one more year.
But since I'm playing on 1080p and I'm not planning on changing that, I expect I wont run in to problems with my overclocked Vega 56.
I only play Battlefield. And eaven if the next game is 30% more demaning, I'll still be able to run it at 60fps+.

Nvidia can release whatever they want. I wont contribute to their monopoly.
And with my x470 motherboard, I expect to upgrade to Ryzen 2 and perhaps Ryzen 3 after that.
There was no excuse to buy AMD hardware a year ago. But these days, there is no excuse not to buy AMD.

Its your choice... The relevance of this eludes me though

 

[kruk]

MCM in theory sounds nice, just like many of RTG's promises that were supposed to "revolutionize gaming" while in end never picks up any steam.

MCM is not something "nice" or a gimmick ... It's something nVidia and AMD will have to implement properly, if they want to keep increasing performance of their GPUs as we are nearing process node limits. The same story holds for stacked dram (HBM). Its future iterations are going to be massively important for MCM designs, therefore looking at the current HBM implementations only as the way to reduce power consumption is extremely shortsighted.

For the doubters, here is what we could read on TPU 1 year ago: https://www.techpowerup.com/234908/nvidia-laying-groundwork-for-multi-chip-module-gpus
It's from nVidia and their sketch of MCM GPUs confirms everything that was written above.

AMD knew exactly what they were doing when they released Fury and Vega. Those chips were a test platform for things to come. Fury as a test drive for HBM and Vega as a test drive for a module interconnect:

In “Vega” 10, Infinity Fabric links the graphics core and the other main logic blocks on the chip, including the memory controller, the PCI Express controller, the display engine, and the video acceleration blocks

The logical continuation is ofcourse Navi being a first MCM design, but now that Raja is gone, we will see what the actual outcome will be ...

 

[xkm1948]

The departure of Raja is probably one of the best things happened to RTG in a long time. Now they at least will be lead by a competent leader Dr. Su.

RTG ain’t gonna die. They may just focus on consoles and industry solutions.