Radeon VII

[Ian Evey]

Ahhhhh AMD, always promise so much and deliver so little.

Ahhhhh Nvidia, promise so much and charge the earth for it.

 

[4K8KW10]

Lol. You do know that not until recently AMD have been in the red for MANY years? This means not enough money for R&D. Not only that, but it takes ages to make the changes your are talking about. Can’t just do it in a couple of months. Things are playing out from decisions made years ago.

I don't think that doubling the 2304 shaders of Ellesmere and making it a viable gaming SKU would have taken so much money...
I think RTG strategy and execution is extremely poor and wrong.

 

[nashathedog]

AMD Radeon VII will become available in reference design only

https://www.guru3d.com/news-story/amd-radeon-vii-willb-ecome-available-in-reference-design-only.html

Navi coming sooner rather than later perhaps?

From the picture it looks like the fans are actually in the heatsink, If you look along the top edge of the card in picture 3 you'll see what I mean, It looks like instead of using a solid block heatsink in a 2.5 slot package they've kept it from going beyond 2 by hollowing out the heatsink for the fan's to drop into it, I'm not sure that was a good way to go, Based on Vega it isn't going to be good news for temps & noise. They should have gone AIO or copied what was done with the 3 fan Fury's and Vega's.

 

[melmac]

Ahhhhh AMD, always promise so much and deliver so little.

Ahhhhh Nvidia, promise so much and charge the earth for it.

Not really accurate statements this time around though. AMD didn't promise anything with Vega 2.

And if you combine both statements it works better for Nvidia's latest cards, Promised so much, delivered so little and cost the earth.

 

[melmac]

I don't think that doubling the 2304 shaders of Ellesmere and making it a viable gaming SKU would have taken so much money...
I think RTG strategy and execution is extremely poor and wrong.

I think you should stop thinking. It would have cost a fortune to make the changes you are suggesting. Please tell me how you were going to just double the shaders without doing a massive redesign.

 

[melmac]

My questions are two and very simple:
1. Why doesn't AMD want to make the Vega 20 specifically for gaming?

Because they don't want to spend money. The new Vega cards are just a chance for them to sell of chips that didn't make the cut for the Instinct cards without investing a lot of time or effort. They only have finite resources and Navi is their goal and that's what they are concentrating on.

 

[LePhuronn]

Actually AMD won't kill FP64 performance on RVII, which is why it will disappear from the shelves.

Well they have I'm afraid. AMD’s Director of Product Marketing Sasa Marinkovic told TechGage "Radeon VII does not have double precision enabled."

https://techgage.com/news/radeon-vii-caps-fp64-performance/

 

[LePhuronn]

My questions are two and very simple:
1. Why doesn't AMD want to make the Vega 20 specifically for gaming?
2. Why was the last new gaming SKU coming from them back in 2016, the Ellesmere from the Arctic Islands, which was renamed to Polaris?
(Vega = Greenland).

1. Because there is nothing new added to Vega 20 which would have any real benefit to gaming. Also AMD have been running dangerously close to bankruptcy for many years so they don't have the money. Also AMD focussed development on a pure gaming architecture with Navi. Also are investing R&D budget into moving away from GCN with Arcturus.
2. I have no idea what your question was, but if you're asking "why was the last gaming SKU released in 2016" then the answer is the same as above: AMD have been close to bankruptcy for years, and simply did not have the resources to tear up their long-standing plans. You do know microprocessors are conceived, planned and designed YEARS in advance, right?

Also Vega doesn't equal Greenland. Vega is a star, specifically the 5th brightest star in the Northern Hemisphere sky. The brightest star in the Northern Hemisphere is the North Star, otherwise known as Polaris. Did you know the 2nd brightest start in the Northern Hemisphere is called Arcturus?

Speaking of stars, did you know that Gamma Cassiopeiae was used during NASA space missions as an easily identifiable navigation aid? It was even nicknamed "Navi" by astronaut Virgil Grissom.

Polaris, Vega, Navi and Arcturus...now why do they sound familiar?

 

[shankly1985]

AMD Responds to Radeon VII Short Supply Rumours

"While we don't report on production numbers externally, we will have products available via AIB partners and AMD.com at launch of Feb. 7, and we expect Radeon VII supply to meet demand from gamers. "

https://www.overclock3d.net/news/gpu_displays/amd_responds_to_radeon_vii_short_supply_rumours/1

 

[subbytna]

Isn't the entire and central point here the Radeon 7 is not a new product? It's Vega 20 + HBM packages that didn't make the cut for an Instinct card and would otherwise have just been thrown into the bin. But as it turns out, AMD could smash some more power through it, take the clocks up and lo and behold it can supposedly fight with the RTX 2080. easy win.

So all this talk about "but why couldn't they do this" or "couldn't they do that" is totally moot. No, they couldn't do this because AMD have not made anything new.

However, given that these packages were destined for the trash anyway, there's no real excuse to pitch the card at $699. Could've done $599 easily because that's still $599 they wouldn't otherwise have had.

The price WILL drop to that price I'd say fairly soon. When the free games offer dies off.

 

[Panos]

My questions are two and very simple:
1. Why doesn't AMD want to make the Vega 20 specifically for gaming?
2. Why was the last new gaming SKU coming from them back in 2016, the Ellesmere from the Arctic Islands, which was renamed to Polaris?
(Vega = Greenland).

1. Vega 20 IS NOT a gaming GPU (neither is Vega 10). Vega 20 is a salvaged, and chopped, product that was going to the bin
No point to design a gaming Vega (10/20), because it needs whole new R&D and that costs to make it work for GDDR ram alone.

2. Because it takes 5 years to make a new GPU. 5 years ago AMD was almost bankrupt. So they focused on money making products (Zen, Fury, Vega) in the lucrative business market, and of course on SOC for consoles.
Also Polaris is a great GPU. Is very well priced and beats the Nvidia competitors at those price brackets.

And I would like to point you out the obvious. When AMD had the better product at the top range, like R9 290X/390X, 7970, 5870 etc was selling less than the worse performing and more expensive Nvidia counterparts.
Look at the GTX1050Ti & GTX1060 against the RX570 and RX580. Nvidia has the worse performing and more expensive products compared to AMD on those brackets.
Yet there are people in this very forum, promoting as "better card" the 1050Ti & 1060 instead of the 570 & 580.

So spare us the anti AMD rhetoric. AMD is there to make money, not to lower Nvidia prices.
As long as you continue buying overpriced Nvidia products, the prices will go up regardless the competition. Expect the next **80Ti to sell for £2000 and then next **60 at £500 bracket.
And that is down to you and everyone who blindly supports Nvidia.

 

[Gerard]

AMD Responds to Radeon VII Short Supply Rumours

"While we don't report on production numbers externally, we will have products available via AIB partners and AMD.com at launch of Feb. 7, and we expect Radeon VII supply to meet demand from gamers. "

https://www.overclock3d.net/news/gpu_displays/amd_responds_to_radeon_vii_short_supply_rumours/1

For a change they're shooting down the bs.

 

[D.P.]

Course you will, just flash it with an mi50 bios and be done with it. I recon these will be unlocked and flashed within a few days.

I doubt it, we will see. Also, once the MI50 is on there then there will be no gaming drivers. So it would really be for people looking at high DP throughput on the cheap.

 

[4K8KW10]

1. Because there is nothing new added to Vega 20 which would have any real benefit to gaming. Also AMD have been running dangerously close to bankruptcy for many years so they don't have the money. Also AMD focussed development on a pure gaming architecture with Navi. Also are investing R&D budget into moving away from GCN with Arcturus.
2. I have no idea what your question was, but if you're asking "why was the last gaming SKU released in 2016" then the answer is the same as above: AMD have been close to bankruptcy for years, and simply did not have the resources to tear up their long-standing plans. You do know microprocessors are conceived, planned and designed YEARS in advance, right?

Also Vega doesn't equal Greenland. Vega is a star, specifically the 5th brightest star in the Northern Hemisphere sky. The brightest star in the Northern Hemisphere is the North Star, otherwise known as Polaris. Did you know the 2nd brightest start in the Northern Hemisphere is called Arcturus?

Speaking of stars, did you know that Gamma Cassiopeiae was used during NASA space missions as an easily identifiable navigation aid? It was even nicknamed "Navi" by astronaut Virgil Grissom.

Polaris, Vega, Navi and Arcturus...now why do they sound familiar?

Ok In Navi and Arcturus we trust.
But Greenland is Vega and Vega is the new name of the GPU previously known as Greenland. Why Mr. Raja decided to rename it, push silly marketing and release unreal videos (remember Poor Volta) in his defense, I do not know.

1. Vega 20 IS NOT a gaming GPU (neither is Vega 10). Vega 20 is a salvaged, and chopped, product that was going to the bin
No point to design a gaming Vega (10/20), because it needs whole new R&D and that costs to make it work for GDDR ram alone.

2. Because it takes 5 years to make a new GPU. 5 years ago AMD was almost bankrupt. So they focused on money making products (Zen, Fury, Vega) in the lucrative business market, and of course on SOC for consoles.
Also Polaris is a great GPU. Is very well priced and beats the Nvidia competitors at those price brackets.

And I would like to point you out the obvious. When AMD had the better product at the top range, like R9 290X/390X, 7970, 5870 etc was selling less than the worse performing and more expensive Nvidia counterparts.
Look at the GTX1050Ti & GTX1060 against the RX570 and RX580. Nvidia has the worse performing and more expensive products compared to AMD on those brackets.
Yet there are people in this very forum, promoting as "better card" the 1050Ti & 1060 instead of the 570 & 580.

So spare us the anti AMD rhetoric. AMD is there to make money, not to lower Nvidia prices.
As long as you continue buying overpriced Nvidia products, the prices will go up regardless the competition. Expect the next **80Ti to sell for £2000 and then next **60 at £500 bracket.
And that is down to you and everyone who blindly supports Nvidia.

You are right. This is the most scary part of the business. To produce something better and only few customers to appreciate it....

 

[LePhuronn]

Ok In Navi and Arcturus we trust.

That was always my stance.

Zen started the road back for CPUs and returned the company to profitability. The 3rd iteration is now poised to return AMD to the top of the CPU food chain across all segments. It will be the same in the GPU space: Vega did its job in bringing money into RTG, but it will be Navi that returns RTG to profitability. The future is with Arcturus as AMD wipe the slate clean and bring Lisa Su's guidance to a unified AMD.

Brand new architecture and cutting edge manufacturing process (TSMC 7nm+) from a company now unified under the guidance and vision of a proven leader, and a company which now generates a healthy profit. All eyes will be on Arcturus, and I hope it's another Zen. That doesn't mean we get an Nvidia killer straight off the bat, but it does mean the ball is rolling and build up momentum to become competitive across all metrics and hopefully reclaim the crown by Arcturus's 3rd iteration.

But Greenland is Vega and Vega is the new name of the GPU previously known as Greenland. Why Mr. Raja decided to rename it, push silly marketing and release unreal videos (remember Poor Volta) in his defense, I do not know.

Fair enough then. But I guess after running out of Pacific islands names and moving to Artics islands, naming your processor after a largely unpopulated frozen wasteland hardly instills a fantastic image. So in my mind it makes sense to pre-empt that rubbish, use stars instead and rename the first product of the family to Polaris. Greenland was likely renamed because Ellesmere was renamed.

But then Koduri always came across as being full of **** anyway.

 

[Panos]

I doubt it, we will see. Also, once the MI50 is on there then there will be no gaming drivers. So it would really be for people looking at high DP throughput on the cheap.

Actually
https://www.4gamer.net/games/446/G044684/20190115124/

If something doesn't make sense in english, blame google translate from japanese to english.

....Before this section they talk about Ryzen 3000 CPUs.....

Q : Radeon VII has 16 GB of HBM 2 installed. Is this the necessary capacity to play today's game?

Adam Kozak: I think it is so in a high-end gaming environment. Although the current high class GPU is becoming a standard of graphics memory capacity of 8 GB, there are increasing cases where it is necessary to play for example in 4 K and HDR environment, for example 10 GB or 11 GB .

An example of recent games is "Battlefield V", "Call of Duty: Black Ops 4", "Rise of the Tomb Raider", and "The Division 2" coming out from now on apply.

Also, in graphics content production and video editing, there is no need to worry about having too much graphics memory capacity. I dare not even explain this. We believe that achieving a graphics memory capacity of over 8GB in high-end class GPUs will become commonplace after 2019.

The Vega 7 nm has a die size that is 68% smaller than Vega 10, but as more space on the GPU package is available, two sets of HBM 2 can now be mounted. This doubles not only the capacity but also the memory bus bandwidth (point to note).

Q: Radeon VII has 60 arithmetic units. In the Radeon Instinct series announced earlier, it corresponds to "MI 50", but there are 64 "MI 60" in the Radeon Instinct ( related article ). Does Radeon VII with 64 arithmetic units appear?

Adam Kozak: In the product name Radeon VII, we only release models with 60 arithmetic units. I can not say anything beyond that.

Q: Radeon Instinct MI 50 and MI 60 are compatible with PCIe Gen.4. May I think that Radeon VII is so?

Adam Kozak: Vega at the chip level of 7 nm, PCIe Gen.4 is supported, but Radeon VII is released as a PCIe Gen.3 compatible card. In the meantime, although we will also introduce PCIe Gen.4 compatible products, for the time being we will be Gen.3 support products. Please recognize that it corresponds to PCIe Gen.4 from the new AM4 platform that appears after the release of the 3rd generation Ryzen.

Even with PCIe Gen.3 connection, there will be no problem in game use in the first place. Radeon VII boasts overwhelming computing performance over 13.8 TFLOPS, and ultra-fast memory performance reaching 1 TB/s is unrelated to PCI Express.

Q: Does Radeon VII support Infinity Fabric based multi-GPU?

Adam Kozak: We do not correspond. It does not have the connection terminal for Infinity Fabric Link. Under the DirectX environment, it is only "mGPU" compatible with the Radeon RX Vega series.

Q: Is there new information related to Radeon VII's video interface? Especially I'm worried about compatibility with HDMI 2.1.

Adam Kozak: Support around the video output interface is the same as Vega 10 generation.

Q: What is the TDP of Radeon VII?

Adam Kozak: It's 300W. " Radeon RX Vega 64 " was 295 watts, so it got a little up.

Q: How does AMD think about compliance with DirectX Raytracing (DXR)? Both the user and the game development side are seeking official remarks on AMD's response posture. Also, UL adds DXR compatible benchmark to "3D Mark".

Adam Kozak: I understand the importance of DXR, but I think it is too early to use it in games. Ray tracing is useful for professional video production sites, but there are Radeon Pro Render for such applications.
The game has only the corresponding title Battlefield V , the benchmark also " Port Royal Only. It will be unrelated to 99% of gamers. It is an understanding that there is little benefit for gamers so far.
However, there is no doubt that it will support either. I can not say anything about that time.

Q: NVIDIA is trying to bring AI-based image processing called DLSS (Deep Learning Super Sampling) to game graphics processing. Please let me know what AMD thinks about this trend.

Adam Kozak: Last year's GDC 2018, Microsoft announced a framework "Windows ML" for developing machine learning based applications on the Windows 10 platform, and "DirectML" that makes it available from DirectX (Windows It became effective (October 2018 Update of 10) ( related article ).
We are currently experimenting with obtaining the evaluation version SDK of DirectML, but Radeon VII shows excellent results in that experiment. By the way, Radeon VII scored about 1.62 times the "GeForce RTX 2080" in "Luxmark" which utilizes OpenCL-based GPGPU-like ray tracing renderer . Based on these facts, I think NVIDIA's DLSS-like thing can be done with GPGPU-like approach for our GPU.

Q: NVIDIA has concluded a partnership with Open Source game live broadcasting software "OBS" (Open Broadcaster Software), and said it worked on improving the performance of OBS. Is there any such move in AMD?

Adam Kozak: In Vega generation Radeon, OBS has already demonstrated satisfactory performance. Even if you play "Fortnite" in full HD with one Radeon VII and simultaneously deliver it from OBS to Twitch at full HD 60 fps and bit rate 6 Mbps, the frame rate of the game can be maintained at 123 fps. Without delivering, if you just played the game, the frame rate is 127 fps, so even if you deliver it at the same time the frame rate drop rate is only 3%. Either way it's over 120 fps, so I think that this is enough performance.

Q: CEO Dr. Lisa Su did not mention anything about the next-generation GPU "Navi" (Navi, development code name) this time at the keynote lecture.
At the timing of CES 2018 she had predicted to release Navi in 2019, but what does this mean? Is Navi going well?

Adam Kozak: I can not say anything about Navi, but the development situation is going well. Lisa did not mention Navi in this keynote lecture because I did not want to dilute the presentation impact of Radeon VII. Although this is a personal opinion to the last.

...After this section they talk about APU.....

So to sum up.
a) AMD doesn't answer question about DXR saying "we cannot talk about this this time". However in Luxmark is 62% faster than the RTX2080 (so is faster than the 2080Ti also by around 30%).


b) About DirectML (DirectX based DLSS), we have excellent results but getting certification


Food for thought.

 

[Vince]

I doubt it, we will see. Also, once the MI50 is on there then there will be no gaming drivers. So it would really be for people looking at high DP throughput on the cheap.

If the features are not cut in hardware then you can guarantee some clever chap will sort it out. Every AMD card I have ever owned could in some way be modified or changed if it was based on enterprise based hardware or if it was a lower grade version of another part. Back in the day the 9500 could be unlocked to 9700.. some baffin cards can unlock cores and shaders. Vega can be flashed to Instinct Mi25, vega56 to vega64 the list is just endless, 7970 can be flashed to 280/280x, with no hardware locks it's all very simple.

Literally every card I can remember can easily be flashed and if you mess it up massively there is always the proven unbrick method of bridging pins 1 and 8 during boot. Which I am ashamed to say I have had to do recently with an rx550.

 

[Gerard]

From the picture it looks like the fans are actually in the heatsink, If you look along the top edge of the card in picture 3 you'll see what I mean, It looks like instead of using a solid block heatsink in a 2.5 slot package they've kept it from going beyond 2 by hollowing out the heatsink for the fan's to drop into it, I'm not sure that was a good way to go, Based on Vega it isn't going to be good news for temps & noise. They should have gone AIO or copied what was done with the 3 fan Fury's and Vega's.

At CES a few YouTubers had the case opened and put their ear to the card and stated it was really quiet.

I think personally the fans are a bit small, the hub on the fan is quite big. Had they made the hub smaller and made the blades longer that makes for better airflow. Pretty nice looking design though overall.

The third pic seems to look like the heatsink is actually extended out to the shroud, could be the fans aren't cut into the heatsink but more the heatsink is wrapping around the fans? Third pic really looks like an extension.

 

[james.miller]

If the features are not cut in hardware then you can guarantee some clever chap will sort it out. Every AMD card I have ever owned could in some way be modified or changed if it was based on enterprise based hardware or if it was a lower grade version of another part. Back in the day the 9500 could be unlocked to 9700.. some baffin cards can unlock cores and shaders. Vega can be flashed to Instinct Mi25, vega56 to vega64 the list is just endless, 7970 can be flashed to 280/280x, with no hardware locks it's all very simple.

Literally every card I can remember can easily be flashed and if you mess it up massively there is always the proven unbrick method of bridging pins 1 and 8 during boot. Which I am ashamed to say I have had to do recently with an rx550.

X800gto2

 

[Panos]

If the features are not cut in hardware then you can guarantee some clever chap will sort it out. Every AMD card I have ever owned could in some way be modified or changed if it was based on enterprise based hardware or if it was a lower grade version of another part. Back in the day the 9500 could be unlocked to 9700.. some baffin cards can unlock cores and shaders. Vega can be flashed to Instinct Mi25, vega56 to vega64 the list is just endless, 7970 can be flashed to 280/280x, with no hardware locks it's all very simple.

Literally every card I can remember can easily be flashed and if you mess it up massively there is always the proven unbrick method of bridging pins 1 and 8 during boot. Which I am ashamed to say I have had to do recently with an rx550.

290 flashing to 290X & 390X
Vega 56 flashing to 64
Fury Pro to FuryX

ATI 9500np to 9700Pro via custom drivers or just a pencil