Officially 10 Years since the ATI AMD merger. Was it a good idea?

[TheRealDeal]

Can't agree there - in some ways there isn't a clear winner as the propitiatory/closed system nature of G-Sync tends to be a significant negative over some of its potential advantages but at the same time G-Sync has some not insignificant technical merits (proper handling of games and programs that don't use exclusive fullscreen, etc.) that FreeSync lacks especially now if you have FastSync + G-Sync working properly together. Having used both even without any bias one way or the other there is no way I'd choose FreeSync over G-Sync if I had to pick one for every day use.

I believe Freesync works in windowed mode when using dx12. This might be a problem now but one that will most likely disappear. It's a little like Crossfire/Sli in dx11 as crossfire does not work in windowed mode but i believe in Multi Gpu dx12 it does. Dx12 seems to fix a lot for AMD tbh. I don't know why they can't get it working in dx11 but it would have been a plus if both these techs did.

 

[humbug]

That is how it read to me too, maybe its the way you type stuff out and not people reading into things.

This is what i said.

Which doesn't necessarily mean providing competitive products for the sake of keeping Nvidia in check to benefit Nvidia consumers.

AMD need to focus on products that are profitable, if they cannot sell enough products in a tier like the low end or very high end to justify production costs and R&D then they should not bother.

Chasing Marketshare at all costs was the mistake ATI made.

If you read that as "AMD must surrender the top end" then you need to go back to school.

 

[Rroff]

I believe Freesync works in windowed mode when using dx12. This might be a problem now but one that will most likely disappear. It's a little like Crossfire/Sli in dx11 as crossfire does not work in windowed mode but i believe in Multi Gpu dx12 it does. Dx12 seems to fix a lot for AMD tbh. I don't know why they can't get it working in dx11 but it would have been a plus if both these techs did.

The problem is WDDM - unless MS rewrites it there is no proper way to support it AFAIK regardless of DX11 or 12 - even nVidia are using a nasty hack taking advantage of having extra hardware with the G-Sync FPGA to make it function - AMD are left with even more horrific work arounds:

While not exactly spelling it out, this answer seems to indicate that for the time being, AMD doesn't think FreeSync will work with Microsoft Store sold games in the forced borderless windowed mode. NVIDIA has stated that G-Sync works in some scenarios with the new Gears of War (a Universal Windows Platform app), but it seems they too have issues.

 

[humbug]

Why do MS force windowed mode?

I can't see any reason for it other than to try and thwart G-Sync / Free-Sync because their rubbish consoles don't have it.

Don't buy games from MS store...

 

[LoadsaMoney]

Because he's an Nvidia fanboy.

Couldn't care less, I'll buy ATI or Nvidia certainly wouldn't stick to one like a fan boy.

In current lines, it seems 50/50 split- G-sync/freesync- freesync wins. Memory- ATI here. Power usage/temps- Nvidia. DX12/Vulkan- ATI. Not a clear winner in the two.

No im not, i just think Nvidia are too far ahead of them now, they already have much, much faster cards out than them as it is, and theres more coming, Vega will have to go some to beat the 1070/80/XP/Ti, and whatever else they have out by then, when their next card down in the stack, is only a 2yr old 970 performer, the increase in performance it will have to have over it, will have to be monumental, and i wouldn't care about the Dx12 performance yet, as we've only got a few Dx11 patchers, so i doub't Nvidia are even worried, as by the time Dx12 is getting properly used, and theres a fair amount of games out using it, they'll be much better at it, remember tessellation, AMD had that long before Nvidia, and we all know what happened there when it finally started getting used.

Id love to be wrong. and Vega is the monster for AMD that hits back, but i just can't see it sadly, i just think theres going to be mass disappointment again, just like there was with the Furys.

 

[Rroff]

Why do MS force windowed mode?

I can't see any reason for it other than to try and thwart G-Sync / Free-Sync because their rubbish consoles don't have it.

Don't buy games from MS store...

In general non exclusive fullscreen mode tends to have higher compatibility ironically, alt-tabs faster and plays nicer if you are interfacing with other programs i.e. UWP's goal is an eco system where apps can interact with games, etc. some developers seem to favour it not sure if its because they are too lazy to bother with sanity checking/managing assets on alt-tabbing back into a game or something.

 

[nashathedog]

The 290x launched at the end of October, the 780ti launched at the beginning of November. There was less than 2 weeks between launch dates. So the 780ti was the top card, it doesn't change what I said.

Yes but it doesn't change the fact that AMD managed to beat the Titan taking the lead, Just because Nvidia had the full fat chip sat on a shelf for when it happened doesn't mean it never happened hence the 2900 wasn't the last time AMD made a push for the top slot, There's also the 7970 ghz which was arguably the top card until the Titan released which was almost a year after the 680 released.

The last time AMD/ATI made an effort for the ultimate high end, that wasn't a dual card, was with the 2900xt. Everything since then from AMD has been at slightly lower than highest tier and at a good price,

Nvidia announced the 780ti 4 days after the 290x released in order to take back the top slot.

 

[varkanoid]

The last time AMD/ATI made an effort for the ultimate high end, that wasn't a dual card, was with the 2900xt

Nvidia announced the 780ti 4 days after the 290x released in order to take back the top slot.

ATI Radeon HD2900xt

 

[Ian Evey]

This is what i said.



If you read that as "AMD must surrender the top end" then you need to go back to school.

I think AMD should stop chasing the £400+ market, they don't sell enough GPU's in that segment anyway.

Let Nvidia have it, they already do and there is no way AMD can take it from them.
So put all R&D into making good sub £400 GPU's.

I read this, if you think Amd should stop chasing the £400 market then that means the high end doesn't it?

 

[FuryXGuy]

The last time AMD/ATI made an effort for the ultimate high end, that wasn't a dual card, was with the 2900xt. Everything since then from AMD has been at slightly lower than highest tier and at a good price, like the model they used for the 4xxx cards. Great performers at a great price, but not top end. The 290x and the 7970 were the same. The 290x was released to go against the 780, but the 780ti was the top card. I think the 7970 is the only AMD card that was faster than Nvidia's top card. And that was because Nvidia didn't release a GK100 chip. If Nvidia had released cards based on the GK100, then the 7970 would have found itself where all the top AMD cards find themselves, slightly above the second tier NVidia cards, but behind their top card.



I would argue that the only reason the AMD 7970 and 290/x cards sold in any numbers was because of the mining craze. It was at it's peak back then and because Nvidia stripped out a lot of it's compute stuff for Kepler, the AMD cards were king for miners.

And I don't think having the top card really does make that much impact on sales, certainly not as much as you and others make it out to have and it certainly didn't for ATI. If having the top card really makes all that much difference than you would imagine that during the time of the 9700pro and 9800 pro that they would have increased their market share. I mean they completely dominated everything at the high end for those years.

No they didn't, they actually lost market share. In fact around the time of the 9700pro to the 9800pro Nvidia gained huge market share. In fact they gained market share for 6 straight quarters.

And please stop comparing the graphic card market to the car market. Those comparisons don't work.

Well you are right about the R9 290X getting slightly thumped by the 780 Ti but look at the 780 Ti now. R9 290X is creaming it in benchmarks. You are right about the mining craze that helped the sales of the R9 290/X series but that wasn't the only reason, it was damn good card for the price it offered with 4GB compared to the 3 GB of the 780/Ti. HD 7970 was another top tier card but the mining craze wasn't as big as when the R9 290X came out, so that's wasn't the sole reason it was successful. When it came out in 2011 it was the fastest single GPU you can buy. Only after the release of the GTX 680 did nVidia get the crown back. AMD fought back with the 7970 Ghz edition. So, the 7970 series was a great card.

And secondly I never said the sales of high end cards will make or break AMD; I stated in my previous post where most cards sell in the $200 - $300 price range like 85% of the market. What I did state that it helps AMD keeping up with the technological leadership and increasing their profit margin as the profit margins are greater on the high end.

Secondly, market share doesn't change over time. As other have mentioned, nVidia had great good will among gamers because they had created sucessful products years before the 9700 Pro launched with the TNT 2, GeForce, GeForce 2, GeForce 3, GeForce 4. One successful product launch wasn't going to change people's mindset ovvernight. I mentioned that earlier. It took 2+ years for ATI to overtake market share, following several refreshes of the 9000 series (ie 9800 Pro/9800 XT) then after the release of the X800 did they overtake nVidia in market share.

And the car comparison is extremely valid. Don't you think Tesla learned a lot by creating the high end Roadster; technologies that were developed was later implemented on the cheaper Model S. Even Elon Musk mentioned that putting out the high end Roadster helped pave the way in making cheaper Model S.

 

[FuryXGuy]

At the moment they haven't got a choice, They can't just release a high end card because they ought to, The cold hard truth is they haven't got one to release, It's as simple as that they're not focusing on the Mid range no matter what they say, It's all PR, They simply had nothing available or else it would have been out already. The sales team and PR team are focusing on the Polaris cards because that's what they have in there hands, Meanwhile the gpu development team are likely to be more focused on getting Vega ready than anything else.

AMD have had a few lean years due to low sales, the recession, debts and what funds they had available being spread out between Zen, HBM and everything else that's on the table or on the cards, That's limited how much could be done at the same time, Hopefully that's going top change over the next couple of years, But to do that both Zen and Vega need to deliver solid products.

You can see how they've struggled by how many refreshes we've had, Look at Pitcairn for example that went 7850, 265 and then the 370, How many time have they revamped the 8 core cpu's? Giving us cpu's like the 9590 and the 'e' etc.

Hopefully they've now been through the worst of times and survived, learning a lot as a business so let's hope things start picking up. They deserve it and we need it.

I agree. They simply don't have Vega ready, which is unfortunate because that is letting nVidia reign free on the high end with the GTX 1070/1080 abated.

I do agree that 2017 will be the year for them with Zen and Vega. 2016 was a good sign with Polaris being fairly successful and them clawing back market share. But they really need something for the high end and with Zen hopefully they will become competitive with Core i7/i5 for the first time in 10 years.

 

[FuryXGuy]

If AMD could build a monster GPU that was competitive against Nvidia’s offerings and could be manufactured cheaply enough they would have done it by now. One thing that must be stopping them is power draw, the power usage on cards like the 390Pro and 390X were ridiculous, watt for watt those cards should have been keeping pace with the 980Ti’s and Titans instead the 390 was competing against the 970 and the 390X against the 980 (non ti) in the majority of DX11 titles.

AMD’s GPU’s have fallen into the same trap of AMD CPU’s where management have made decisions based on future technologies that don’t work out and suffered from the same judgement issues that have plagued it’s CPU’s (such as when Bulldozer would herald an era when Software/games would be largely compatible with multi-threaded CPU’s). One of the biggest failings is CPU overhead, AMD can’t/won’t write drivers which allow their hardware to take advantage of multithreaded CPU’s, the command engine in the 290/390/Fury is capable but game programmers need to write for it specifically (given most games are written for consoles it’s wonder why AMD cards don’t perform better on PC). Nvidia cards can call upon multiple CPU threads via it’s drivers given them a huge advantage. AMD choose to incorporate a ACE engine long before it would become relevant in DX12 games, sure it’s meant the 390’s are still relevant today but so what? It doesn’t make AMD any more money.

I suppose when you look at it from a far AMD are designing products for 5 years in the future rather then what’s needed for the current day and it’s cost them dear, there gambling the market will go a certain way and if it doesn’t it costs them dear. With Zen it looks like they have their house in order again (let hope so) but the jury is still out on it’s GPU’s, Polaris is disappointing, it failed to address the problems of prior cards. Polaris smacks of a lack of vision, lack of leadership or a lack of funds, probably a combination of all three but I do agree they need to some cards into the enthusiast end of the market as it’s the only part of the PC market that’s growing (or didn’t AMD read the memo?).

I do agree that they produced products that were too far ahead of it's time that were not beneficial ie 8 core bulldozer when many windows and other applications didn't use much multicore.

Having said that implementing ACE engine did benefit them and it was a smart move. Many people chose the RX 480 over the GTX 1060 simply because it was more forwarding looking architecture and it is paying dividends with DX 12 and Vulkan games where it's beating the GTX 1060. I expect it to get better. Sure the RX 480 could have done better. I was somewhat disappointed with the performance, it should done better than R9 390/290X level performance, but they did improve on the power front and they did sell very well gaining them close to 34% market share. While you are right about the 390 not making them more money, this is simply because they are phasing out the 390 but using ACE in RX 480 did pay dividends for the reasons I mentioned above.

I do hope that Zen and Vega turn out to be great. They seem to have better leadership with Raja at the helm of RTG and their current CEO seems to be more competent. I have my fingers crossed.

 

[LoadsaMoney]

Look at the amount of power the 480 needs, its performance, and how hot it is compared to the 1080, then you'll see why they can't do a card to compete with it yet.

 

[Rroff]

Look at the amount of power the 480 needs, and how hot it is compared to the 1080, then you'll see why they can't do a card to compete with it yet.

Supposedly they now have a revised variant of the process that fixes the power/thermal efficiency. Even then though I can't see them doing anything but chasing nVidia at this point unless they undo everything they've done and throw a massive amount of money at big cores.

 

[LoadsaMoney]

Which they can't do, as they're broke.

 

[Freddie1980]

Supposedly they now have a revised variant of the process that fixes the power/thermal efficiency. Even then though I can't see them doing anything but chasing nVidia at this point unless they undo everything they've done and throw a massive amount of money at big cores.

If your saying that due to the review of the XFX GTR 480 showing much better power and thermal efficiency that's probably more due to binning then process improvement. It would be good if it were due to an improved process but I'm not optatmisic.

 

[Freddie1980]

Which they can't do, as they're broke.

There not broke at all AMD has plenty of cash reserves and a bunch of new consoles from Sony, MS and the Apple deal as well to keep them going. The issue with the lack of a competitive top end product may be due to a lack of talnet, by all account GPU designers are being snapped up by the likes to qualcomm to help them build better graphic chips for their SoC's.

 

[eddyr]

I agree. They simply don't have Vega ready, which is unfortunate because that is letting nVidia reign free on the high end with the GTX 1070/1080 abated.

I do agree that 2017 will be the year for them with Zen and Vega. 2016 was a good sign with Polaris being fairly successful and them clawing back market share. But they really need something for the high end and with Zen hopefully they will become competitive with Core i7/i5 for the first time in 10 years.

I would argue that it is a positive considering their circumstances. Nvidia having free reign at the top can't be helped and is the better of the 2 bad options. Which is either forego the high end for a generation and invest that money into a future product or release a card that generates no ROI and at best reinforces a stereotype of being slower or more power hungry than their competitor.

If they had brought out bigger dies out in this last quarter (assuming GF were capable and it was economically feasible), they would have been polaris based and hypothetically you would be looking at ~350mm2 225W chip (perhaps dropping sub 200W over time with process refinements) to compete with the 1080. R&D resources directed to Vega design team delayed or reduced, potentially delayed succeeding tech (or tighter time constraints), higher R&D expenditure overall, etc, which would not help AMD whatsoever.

 

[FuryXGuy]

Thats not what i said, its what you read into but its not what i said.

This is what you said:

I think AMD should stop chasing the £400+ market, they don't sell enough GPU's in that segment anyway.

Let Nvidia have it, they already do and there is no way AMD can take it from them.
So put all R&D into making good sub £400 GPU's.

Explain to me how " stop chasing the £400+ market" is not considered going after the high end? Pretty much all high end graphics have been in the £400+ range at the time they were released.

 

[FuryXGuy]

I would argue that it is a positive considering their circumstances. Nvidia having free reign at the top can't be helped and is the better of the 2 bad options. Which is either forego the high end for a generation and invest that money into a future product or release a card that generates no ROI and at best reinforces a stereotype of being slower or more power hungry than their competitor.

If they had brought out bigger dies out in this last quarter (assuming GF were capable and it was economically feasible), they would have been polaris based and hypothetically you would be looking at ~350mm2 225W chip (perhaps dropping sub 200W over time with process refinements) to compete with the 1080. R&D resources directed to Vega design team delayed or reduced, potentially delayed succeeding tech (or tighter time constraints), higher R&D expenditure overall, etc, which would not help AMD whatsoever.

Not having anything to compete with the high end can never be considered a positive. They could have had Polaris compete with in the mid to low end and Vega compete with the high end.

nVidia just released record revenue with $1.2 billion in revenue just in their Gaming Segment (which includes the GTX 1070 and GTX 1080). Don't tell me that they would have made that much money if Vega had been released couple of months earlier.

GPU companies have had multiple design teams working con currently on products. The sad fact is that AMD is still struggling with execution. Vega should have at the least come out in October of this year NOT 1Q/1H of 2017.