Vega/Navi Rumors (Updated)

[Veradun]

I wonder why AMD is not using GDDR5X? Licensing issue maybe? I mean, it seems like a good technology - a cost efficient way to implement higher memory bandwidth without requiring highly specialised HBM2.

We lack a whole lot of the insight to make even barely accurate statements on this matter: how much an HBM stack costs to AMD, how much a GDDR5x chip would cost them in terms of $$ and in terms of differences in die size, power consumption, pcb and components. That's not that clear.

 

[T1beriu]

Both Polaris and Vega were developed in China, which may explain why they are so underwhelming.

He failed, and should be fired. So should the Chinese development team, and R&D brought back to the USA.

Wow. Nationalism and xenophobia are strong in this one.

You know ATI was a canadian company, right?

Or that almost half of IT engineers in the US are from outside the country?

 

[antihelten]

I have to ask (I haven't watched the stream): was the demo live? If not, does the playback Hz necessarily match the recorded FPS of the video? It would be an utterly lazy oversight if the projector was set to 30Hz or couldn't do 60, but it's possible, I suppose. And if it was live, I assume whoever was playing wasn't using the projector as the main display, but as a mirror. The question then is how the signal was split. I'd assume they used some sort of HDMI splitter (rather than Windows display controls - can Windows even render the same game mirrored on two displays?), which again begets the question of whether the splitter might be "downsampling" a 60Hz signal to 30Hz by, for example, skipping every other frame.

The language used by Lisa Su suggested that it was live, but who knows.

Regarding using a splitter or whatever they used for the presentation, I certainly wouldn't rule out that this could be a potential source for issues with the refresh rate.

Also, I know I'm not very sensitive to that kind of stuff, but how does the recording "clearly show" that the refresh rate was 30Hz?

If you pause the video in youtube and use , and . to skip a frame back/forwards, you will see that the projector only updates once for every two frames of the video, and since the video is 60Hz, the projector must be running at 30Hz.

On the AMD side it's 1:16.

Except for Hawaii, which was 1:8 for the gaming cards (and 1:2 for professional cards)

 

[HumblePie]

Wish I could throw in my 2 cents here to the discussion. Was hoping to, because friend's wife went to computex and gave me a link to watch and said be ready... Then we find out that the big spoil there was really threadripper/epyc more than anything else. I think they were planning on saying more than they did about Vega, but I can only make speculative conjecture as to why what little about vega was revealed. Kind of sucks because I was hoping for a bit more myself.

 

[CatMerc]

Except for Hawaii, which was 1:8 for the gaming cards (and 1:2 for professional cards)

Obviously I meant for gaming focused chips, such as Fiji, Tonga, Polaris, Cape Verde, Bonaire, Pitcairn, they're all 1:16.
Hawaii and Tahiti are the exceptions, with 1:2 and 1:4 respectively.

 

[Paratus]

The language used by Lisa Su suggested that it was live, but who knows.

Regarding using a splitter or whatever they used for the presentation, I certainly wouldn't rule out that this could be a potential source for issues with the refresh rate.



If you pause the video in youtube and use , and . to skip a frame back/forwards, you will see that the projector only updates once for every two frames of the video, and since the video is 60Hz, the projector must be running at 30Hz.



Except for Hawaii, which was 1:8 for the gaming cards (and 1:2 for professional cards)

Except for Tahiti which was 1:4 and over 1 TF double precision.

I still don't think they've released a mainstream card with more theoretical DP performance.

 

[Yakk]

I'm really curious as to what went so horribly wrong with Vega to cause this perported massive delay to the point of almost falling 1 generation behind at the high-end. I guess those are the risks when going for these massive gains, it worked out for Ryzen, not for Vega. Even I'm discouraged by Vega at this point, unless in a surprised move AMD just skipped this generation and Vega really is a next gen (wishful thinking I know).

 

[raghu78]

We all know what the excuse will be if the Frontier Edition is slower than some expect in games... "The driver's aren't ready! Wait for RX!"

How are you going to judge a RX Vega gaming card without launch driver ? The Frontier edition is going to launch with Pro driver. You would be one of the first to rejoice if Frontier Edition is slower than 1080 Ti . But the fair assessment of RX Vega will come on launch day with press reviews based on launch driver. If RX Vega fails to impress on launch day then there are no excuses. AMD cannot hide behind the excuse that its a new architecture and drivers will take time to improve. Nvidia delivers fantastic performance on launch day and if AMD wants to win a portion of the enthusiast segment they have to deliver on launch day. This time AMD has very little time left before Nvidia unleashes GV104 and GV102. I am pretty sure GV102 will destroy Vega 10 so if Vega has to succeed then AMD has to deliver on launch day.

 

[T1beriu]

The Frontier edition is going to launch with Pro driver

I think that's not true. I saw Raja's comment denying that. Sadly I can't remember where. Or I'm just imagining it.

 

[Veradun]

I think that's not true. I saw Raja's comment denying that. Sadly I can't remember where. Or I'm just imagining it.

Actually it is (true):

The Frontier Edition was designed for a variety of use-cases like Machine Learning, real-time visualization, and game design. Can you play games on Frontier Edition? Yes, absolutely. It supports the RX driver and will deliver smooth 4K gaming. But because it is optimized for professional use cases (and priced accordingly), if gaming is your primary reason for buying a GPU, I’d suggest waiting just a little while longer for the lower-priced, gaming-optimized Radeon RX Vega graphics card

https://www.reddit.com/r/Amd/commen...n_technologies_group_at_amd_and_were/dhqnsgo/

 

[Malogeek]

Well technically he says it supports the RX driver, but not that it will ship with one.

 

[Veradun]

Well technically he says it supports the RX driver, but not that it will ship with one.

Yep, and I'm sure it won't, so you won't be able to infer RX gaming performance using a Frontier unless you wait for the RX to launch (and have its own driver)

 

[raghu78]

I think that's not true. I saw Raja's comment denying that. Sadly I can't remember where. Or I'm just imagining it.

I am quoting Raja Koduri from the reddit discussion

https://www.reddit.com/r/Amd/comments/6bklro/we_are_radeon_technologies_group_at_amd_and_were/

Will the consumer RX version be as fast at the Frontier version?

• Consumer RX will be much better optimized for all the top gaming titles and flavors of RX Vega will actually be faster than Frontier version!

What things does the RX Vega have over the Radeon Vega FE that would make it worth the extra wait

• RX will be fully optimized gaming drivers, as well as a few other goodies that I can't tell you about just yet....But you will like FE too, if you can't wait

Is the Frontier Edition a card like a Titan X (a professional/gaming card?), will we have the choice between RX driver and Pro Driver?

The Frontier Edition was designed for a variety of use-cases like Machine Learning, real-time visualization, and game design. Can you play games on Frontier Edition? Yes, absolutely. It supports the RX driver and will deliver smooth 4K gaming. But because it is optimized for professional use cases (and priced accordingly), if gaming is your primary reason for buying a GPU, I’d suggest waiting just a little while longer for the lower-priced, gaming-optimized Radeon RX Vega graphics card

The way I see it AMD is still optimizing drivers for RX Vega and thus anything before RX Vega launch day might not reflect its best performance.

 

[leoneazzurro]

In a tweet also Raja declared that Vega drivers are not ye toptimized for Prey, so who knows if CF is actually working at all, or only partially.

 

[tg2708]

In a tweet also Raja declared that Vega drivers are not ye toptimized for Prey, so who knows if CF is actually working at all, or only partially.

So why did they say two cards were used? One would think, or I would think that if CF was supposedly not working why did they say two were used. But then again I briefly read somewhere that its a demo of the threadripper performance.

 

[Despoiler]

As crazy as trying to compete with Intel on the cpu manner given how many generations behind they were and yet they surpassed our expectations (the reasonable ones not the people who are expecting amd to compete directly with Intel with just 1 cpu drop).


It's not a crazy idea, it's whether it works.

The difference between Intel and Nvidia is that Intel rested on its laurels. Intel's product roadmap since bulldozer is a confused mess of cancellations, reworks, renames, and date changes. Zen seemingly caught them completely off guard. Nvidia on the other hand has not let up on their execution at all. There is no denying Nvidia's cadence is impressively unrelenting. That is the crux of the issue for AMD. If they need time to work things out to release a good product, fine. The problem is it gives Nvidia that much more of an advantage. It's a double edge sword that AMD cannot afford to wield. The good news is that because one AMD competitors stumbled AMD's coffers will balloon. That means more money for RTG to fight Nvidia.

 

[leoneazzurro]

So why did they say two cards were used? One would think, or I would think that if CF was supposedly not working why did they say two were used. But then again I briefly read somewhere that its a demo of the threadripper performance.

Maybe they wanted to show that the platform was there working (Threadripper + Vega in CF) but drivers are not ready yet.

 

[nathanddrews]

Maybe they wanted to show that the platform was there working (Threadripper + Vega in CF) but drivers are not ready yet.

I suspect that this is the only reason to do so. AMD is on the offensive and despite any delays (perceived or actual), they need to keep their momentum. I would say that a demo of Threadripper + 2x RX Vega CF playing Prey and Threadripper + 4x Vega FE in Blender is pretty epic (but not EPYC).

I have plenty of things I wish for (like RX Vega being faster than 1080Ti for $399), but I think realistically we're going to get Fury X 2.0 with Vega. RX Vega will launch at near-1080 (vanilla) performance and be priced around $500. NVIDIA will drop prices on Pascal and launch Volta, leaving Vega in the dust.

 

[96Firebird]

How are you going to judge a RX Vega gaming card without launch driver ? The Frontier edition is going to launch with Pro driver. You would be one of the first to rejoice if Frontier Edition is slower than 1080 Ti . But the fair assessment of RX Vega will come on launch day with press reviews based on launch driver. If RX Vega fails to impress on launch day then there are no excuses. AMD cannot hide behind the excuse that its a new architecture and drivers will take time to improve. Nvidia delivers fantastic performance on launch day and if AMD wants to win a portion of the enthusiast segment they have to deliver on launch day. This time AMD has very little time left before Nvidia unleashes GV104 and GV102. I am pretty sure GV102 will destroy Vega 10 so if Vega has to succeed then AMD has to deliver on launch day.

Please refer to the context of my post; the post above mine helps.

 

[antihelten]

Obviously I meant for gaming focused chips, such as Fiji, Tonga, Polaris, Cape Verde, Bonaire, Pitcairn, they're all 1:16.
Hawaii and Tahiti are the exceptions, with 1:2 and 1:4 respectively.

Tahiti is another one yes, I was just thinking from Maxwell and on (since you mentioned Maxwell). Technically Hawaii was of course released before Maxwell, but it was re-released with the 300 series (renamed to Grenada).

Also I would say that both Tahiti and Hawaii are gaming focused chips first and foremost. Sure they are stronger in compute than other chips, but that doesn't necessarily imply that they are primarily focused on the professional market (that would be true for chips like GP100/GV100).

Except for Tahiti which was 1:4 and over 1 TF double precision.

I still don't think they've released a mainstream card with more theoretical DP performance.

An interesting thing about Tahiti is that it's the only high DP performance card from AMD that wasn't gimped in the gaming version. Hawaii was capable of a 1:2 ratio, but was gimped to 1:8 in the gaming version. If Hawaii had been given it's full DP performance in the gaming version like Tahiti was, then it would have been at almost 3 TFLOPS.

 

[CatMerc]

Tahiti is another one yes, I was just thinking from Maxwell and on (since you mentioned Maxwell). Technically Hawaii was of course released before Maxwell, but it was re-released with the 300 series (renamed to Grenada).

Also I would say that both Tahiti and Hawaii are gaming focused chips first and foremost. Sure they are stronger in compute than other chips, but that doesn't necessarily imply that they are primarily focused on the professional market (that would be true for chips like GP100/GV100).

Should have phrased it better. Obviously Hawaii and Tahiti are gaming focused, but they also packed double precision to act as AMD's HPC GPU. This means they're "less" gaming focused than the others relatively, but they're still gaming focused.

 

[antihelten]

Should have phrased it better. Obviously Hawaii and Tahiti are gaming focused, but they also packed double precision to act as AMD's HPC GPU. This means they're "less" gaming focused than the others relatively, but they're still gaming focused.

I would certainly agree with you that they are less gaming focused than the other GCN chips, and by extension thus also more compute/HPC/professional focused. My point is simply that even with the lower focus on gaming, that is still their primary focus, at least in my opinion (I don't have access to the inner thoughts of AMD obviously).

Regardless of DP ratio, I think one could possibly make the argument that Vega is actually the most professional oriented chip AMD has done to date, given that they are leading with professional cards for it.

 

[Mopetar]

I'm really curious as to what went so horribly wrong with Vega to cause this perported massive delay to the point of almost falling 1 generation behind at the high-end.

Drivers seem to be the biggest culprit based on the evidence we have available.

Given V100 has had HBM2 shipping for a long while and the Vega FE card has 8 GB chips which some still thought impossible to make in quantity prior to being shown. Consumer Vega will probably only have 4 GB chips which aren't as different to manufacture.

There were also reports that 80% of Ryzen dies were good so yields can't be responsible either. Sure Vega is bigger so less likely to yield full die in the same amounts as Ryzen, but AMD can sell cut Vega as well so even if only 40% are full chips they should still have enough chips.

Capacity at GF could be a limitation, but AMD could use Samsung since it's the same tech. With the amended WSA they could do this for a high margin part like Vega.

Respins don't seem too likely as they've had hardware demos for months now so they can't be changing too much unless they're trying to get faster clock speeds.

 

[Arachnotronic]

They can do pre-emptive price drops instead of having to wait. They could sell 1080 for $300 if they wanted to completely price AMD out. There is a reason AMD hasn't dropped speed and pricing for RX

Margins on a $300 1080 would be very close to zero.

 

[tential]

Margins on a $300 1080 would be very close to zero.

I think the point is, if you give Nvidia an inch, they'll leave with your wife and kids.
The more info Nvidia has on what AMD plans to do, the better they can preemptive wreck AMD.
Especially when AMD continually adds in more pit stops to their releases.