The future of AMD in graphics

[Yotsugi]

Ever since we heard of Navi we know of it as the Polaris replacement

It's actually top to bottom, replaces everything besides the dedicated FP64 part (it has none).

So I take that as having a size limit that would probably prevent it from being competitive Nvidia's offering

The bigger one will give them a fair fight.

 

[Stuka87]

I highly doubt AMD would limit the performance of Navi just to try to save face if they surpass Vega VII. That would seem rather ludicrous to do.

 

[Yotsugi]

I highly doubt AMD would limit the performance of Navi just to try to save face if they surpass Vega VII. That would seem rather ludicrous to do.

Because they won't.
They will slap the EOL sticker on 7 soon enough anyway.

 

[AtenRa]

I highly doubt AMD would limit the performance of Navi just to try to save face if they surpass Vega VII. That would seem rather ludicrous to do.

They may not need to even release a NAVI die that will be able to compete against Vega 20. They could keep Radeon Vii for 12-16 months and then in Q2-Q3 2020 they may release a next gen chip to compete against NVIDIAs 7nm gpus.

 

[Topweasel]

They may not need to even release a NAVI die that will be able to compete against Vega 20. They could keep Radeon Vii for 12-16 months and then in Q2-Q3 2020 they may release a next gen chip to compete against NVIDIAs 7nm gpus.

I mean this is the most likely thing to happen. People react weird that AMD would sit on the Vega 20 just because it's not really faster then the 1080ti. Bit if you look at it, it's 98% of the performance of NVidia's top < $1k video cards. Considering the 2080 just came out we are basically looking at 4 years at that performance mark. There will be small impetus to push performance higher ahead of NVidia's next series.

That said Vega20 was a late decision as a pipe cleaner due to the growing popularity of their MI lineup. The Radeon VII even more of a late decision based on the 2080 pricing leaving the door open to sell the Radeon VII at a profit. So it's very possible that it's performance envelope fits within what AMD was targeting the large Navi will sit at. I don't think so. But it is possible.

 

[Yotsugi]

That said Vega20 was a late decision as a pipe cleaner due to the growing popularity of their MI lineup

Vega20 is ancient as all hell.
It's a long overdue Hawaii replacement.

The Radeon VII even more of a late decision based on the 2080 pricing leaving the door open to sell the Radeon VII at a profit

Now that's semi-valid.

So it's very possible that it's performance envelope fits within what AMD was targeting the large Navi will sit at

Big Navi is higher.

 

[AtenRa]

I didnt know where else to put this,

Intel Discrete GPU Boss Raja Koduri Explains Why He Left AMD
https://hothardware.com/news/intel-...-explains-why-he-left-amd#AX6gCVXcB4tQBuYo.99

 

[maddie]

I mean this is the most likely thing to happen. People react weird that AMD would sit on the Vega 20 just because it's not really faster then the 1080ti. Bit if you look at it, it's 98% of the performance of NVidia's top < $1k video cards. Considering the 2080 just came out we are basically looking at 4 years at that performance mark. There will be small impetus to push performance higher ahead of NVidia's next series.

That said Vega20 was a late decision as a pipe cleaner due to the growing popularity of their MI lineup. The Radeon VII even more of a late decision based on the 2080 pricing leaving the door open to sell the Radeon VII at a profit. So it's very possible that it's performance envelope fits within what AMD was targeting the large Navi will sit at. I don't think so. But it is possible.

This right here is AMD's problem. As you stated "Bit if you look at it, it's 98% of the performance of NVidia's top < $1k video cards", is how I see AMD's approach to marketing. When AMD claims that they're competing in the high end GPU market, this is what their logical thinking entertains. They don't seem to realize the rabid nature of the loudest enthusiasts today.

Using the ubiquitous car analogy, if you don't have the absolute fastest car, you're driving a piece of junk.

What amazes me is how the majority swallow this propaganda. That is a very difficult battle to fight.

 

[ozzy702]

This right here is AMD's problem. As you stated "Bit if you look at it, it's 98% of the performance of NVidia's top < $1k video cards", is how I see AMD's approach to marketing. When AMD claims that they're competing in the high end GPU market, this is what their logical thinking entertains. They don't seem to realize the rabid nature of the loudest enthusiasts today.

Using the ubiquitous car analogy, if you don't have the absolute fastest car, you're driving a piece of junk.

What amazes me is how the majority swallow this propaganda. That is a very difficult battle to fight.

Nah, it's more that products that run super loud/hot because they consume 50-150% more power for the same performance as the equivalent NVIDIA GPU can't command the same price and interest.

AMD's architecture is so bad that even with a huge process advantage the Radeon VII gets beat consistently by the 2080 while consuming significantly more juice and running much hotter/louder.

Yes yes, we all know you can undervolt and tinker with things to do the job that AMD should have in the first place, but time and time again AMD illustrates why the market has swung so far in favor of NVIDIA.

I hope that NAVI and it's successors rectify the poor job that AMD has done since the 7970 (290 was a flop that wasn't fixed until 390 and everything after 390 has been sad in contrast to NVIDIA's offerings), and that excitement and competition can re-enter the market. The only AMD GPU I can recommend my friends and family are super cheap used 570/580 cards that I test/refurbish for them. In contrast I've built several Ryzen boxes for friends.

 

[maddie]

Nah, it's more that products that run super loud/hot because they consume 50-150% more power for the same performance as the equivalent NVIDIA GPU can't command the same price and interest.

AMD's architecture is so bad that even with a huge process advantage the Radeon VII gets beat consistently by the 2080 while consuming significantly more juice and running much hotter/louder.

Yes yes, we all know you can undervolt and tinker with things to do the job that AMD should have in the first place, but time and time again AMD illustrates why the market has swing so far in favor of NVIDIA.

I hope that NAVI and it's successors rectify the poor job that AMD has done since the 7970 (290 was a flop that wasn't fixed until 390 and everything after 390 has been sad in contrast to NVIDIA's offerings), and that excitement and competition can re-enter the market. The only AMD GPU I can recommend my friends and family are super cheap used 570/580 cards that I test/refurbish for them. In contrast I've built several Ryzen boxes for friends.

You're only thinking recently. This situation has not always existed. In fact your argument is exactly this strategy. Wherever you're (not you personally) dominant, becomes the rallying cry.

What was the argument when all of your points were reversed with AMD/ATI dominant? I do agree however, that now is probably their worse period in years.

 

[AtenRa]

290 was a flop

Hawaii was one of the best architectures of the last 5-6 years, easily beaten the more expensive and much bigger die sized kepler GK110 GTX 780Ti that was released at the same time.

 

[ozzy702]

You're only thinking recently. This situation has not always existed. In fact your argument is exactly this strategy. Wherever you're (not you personally) dominant, becomes the rallying cry.

What was the argument when all of your points were reversed with AMD/ATI dominant? I do agree however, that now is probably their worse period in years.

I was buying and recommending AMD GPUs for the 4000, 5000 and 7000 series. Outside of mining/compute, nothing from AMD has remotely been attractive since the 7970.

 

[ozzy702]

Hawaii was one of the best architectures of the last 5-6 years, easily beaten the more expensive and much bigger die sized kepler GK110 GTX 780Ti that was released at the same time.

I agree, Hawaii was awesome once it's potential was realized in the 390. The rollout of the 290 was dismal and a flop.

 

[Yotsugi]

Nah, it's more that products that run super loud/hot because they consume 50-150% more power for the same performance as the equivalent NVIDIA GPU can't command the same price and interest.

No one cared about that when nV was running super loud and hot.

 

[Ajay]

No one cared about that when nV was running super loud and hot.

Yes we did, there were threads on it in every major hardware forum. This has been a problem with numerous cards. Here is one notable card (5800): https://www.anandtech.com/show/1062/3

Edit: typo

 

[Yotsugi]

Yes we did, there were threads on it in every major hardware forum

The market didn't, outside of like 5800.

 

[tajoh111]

No one cared about that when nV was running super loud and hot.

This was mostly mitigated because Nvidia had the performance crown at the time. But when AMD was cheap and efficient vs Nvidia, even the crown wasn't enough to lose a bunch of marketshare. AMD did not outsell Nvidia but they got pretty close which is impressive considering Nvidia's strong Desktop partner presence.

At it's worst in terms of efficiency and lackluster performance per dollar, Nvidia had marketshare in the mid 50's even with the performance crown. Because of the size of the company and the big dies they were using to compete against AMD, Nvidia were also not really making money from the gaming market and it was the professional market margins that was allowing the company to be profitable.

A costly mistake for AMD is they did not want to leverage their performance per watt advantage into a giant monolithic chip during the 4870 series and 5870 series. They would have obtained the performance crown for sure. I guess the reason why is they didn't have the professional market presence to justify the R and D expense of such a big chip. Also Nvidia's chips while inefficient were outstanding in compute and professional applications.

https://hothardware.com/reviews/nvi...nd-5000-series-workstation-gpus-review?page=5

Why the performance per watt matters so much now is Nvidia is leveraging that directly into performance across all power ranges from the low end and high end. Something AMD should have done during the 4870/5870 generation. An enthusiast might not care about power consumption but if that performance per watt allows them to be 30-40 percent faster flagship to flagship in a 300 watt profile, it matters even to enthusiasts.

 

[Shmee]

The after market 290s were far from flops. They actually did quite well, such as the 290 Tri X from sapphire. I would agree hawaii was a good architecture. Fiji as well. But now it is time for new stuff to come from AMD, and I am looking forward to Arcturus.

 

[Staples]

If all else fails, AMD still have the Xbox and Playstation to make graphic chips for. I myself have bought AMD cards because of their Freesync however now that NVIDIA supports it, I as well as others may jump ship on their next purchase.

 

[Topweasel]

If all else fails, AMD still have the Xbox and Playstation to make graphic chips for. I myself have bought AMD cards because of their Freesync however now that NVIDIA supports it, I as well as others may jump ship on their next purchase.

Now there is the google streaming service on top of it. I see AMD getting a lot of money to develop new video card tech that will keep them in the discreet game even if they aren't going to punch as high as people want.

 

[maddie]

Now there is the google streaming service on top of it. I see AMD getting a lot of money to develop new video card tech that will keep them in the discreet game even if they aren't going to punch as high as people want.

Yep. Steady, or should I say more dependable income streams allow much more flexibility for management in longer term planning.

 

[railven]

The after market 290s were far from flops. They actually did quite well, such as the 290 Tri X from sapphire. I would agree hawaii was a good architecture. Fiji as well. But now it is time for new stuff to come from AMD, and I am looking forward to Arcturus.

I would agree with the original poster ONLY because of mining. R9 290 didn't get to enjoy its well earned fame. It released around the cusp of bit mining bubble and I skipped R9 290 for GTX 780 because a Lightning 780 cost LESS than a Ref 290. It wasn't until the 390 rolled around, but by then the mining bubble had mostly popped, and even the 390 didn't get it's fame due to bargain bin 290s being sold left and right. Why buy a shiny 390 when the 290 was going for at some points half the price.

I personally feel the 290 has the most warm reception, at least around here, because lots of people scooped them up for <$200. Hell I think I paid like $150 or something for my Powercolor 290 that's still chugging away in the basement PC.

 

[Topweasel]

Vega20 is ancient as all hell.
It's a long overdue Hawaii replacement.

Now that's semi-valid.

Big Navi is higher.

That's not what I saw. I saw NCU or whatever it was (New compute architecture) which was Navi long before Vega 20. Vega 20 didn't start showing up till about the time VegaFE launched. You could be right, just like you could be right on Navi being bigger or faster. I am not sure that is the market Navi is targeted at. That has always been rumored to be more of a Polaris replacement then a replacement for both it and Vega. Doesn't mean it's not or you are wrong. Just consider me sceptical.

 

[ozzy702]

I would agree with the original poster ONLY because of mining. R9 290 didn't get to enjoy its well earned fame. It released around the cusp of bit mining bubble and I skipped R9 290 for GTX 780 because a Lightning 780 cost LESS than a Ref 290. It wasn't until the 390 rolled around, but by then the mining bubble had mostly popped, and even the 390 didn't get it's fame due to bargain bin 290s being sold left and right. Why buy a shiny 390 when the 290 was going for at some points half the price.

I personally feel the 290 has the most warm reception, at least around here, because lots of people scooped them up for <$200. Hell I think I paid like $150 or something for my Powercolor 290 that's still chugging away in the basement PC.

Yup. I which is why I skipped the 290 and finally went with the 390 a year later. Better cooler, more VRAM, and half the price.

 

[Yotsugi]

That's not what I saw

I dunno what you saw.

Vega 20 didn't start showing up till about the time VegaFE launched

It showed up waaaaaaay before, I think someone actually leaked their roadmaps into public.

You could be right, just like you could be right on Navi being bigger or faster

Both Navi dies that will launch in ~August will be small (and fast).

That has always been rumored to be more of a Polaris replacement then a replacement for both it and Vega

It replaces everything, top to bottom.
Bigger gaming part is probably late'19-early'20 though.