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

[CAT-THE-FIFTH]

This is true. Good analogy with Maxwell vs the Hawaii. This is why I stated you have to smart about die size. Interestingly Carrel Killbrew, the guy I quoted above who was AMD/ATI GPU architect was actually one of the proponents of the "small die strategy". Having a small die doesn't necessarily mean giving up technology leadership. I call it the "smart die" strategy. R9 290/290X was going up against GTX 780/780 Ti, it was not intended to go against Maxwell. Fury was the one which was supposed to counter Maxwell, unfortunately, I just think that the Maxwell architecture and design was just better than Fijii. And it was late to market. it should have been release in 2014 or late 2014 early 2015. That's one of the reason which hurt them.
As a matter of fact, 2014 was the only years over the past 16+ years where AMD/ATI didn't release a top to bottom series of GPU's or even refreshed parts.

Anandtech had a very good article about how Carrell Killbrew came up with the "small die" strategy with the RV770 and it effectively up ended nVidia in that generation cycle. As a matter of fact nVidia actually had losses in the years covering 2008/2009 thanks to HD 4800 series and the 5800 series:

Link

^^ This is probably my all time favorite article on Anandtech.

They also have one about the HD 5800 series.

Hawaii is impressive considering is DP performance and small die,but if you look at Bonaire and Tonga they really didn't change much and performance/watt stagnated.

The problem is AMD at that point were more worried adding various features and not about engineering faster,and more efficient cards.

I don't understand why didn't fathom Nvidia would be doing that??

They should have been thinking more of improving performance/watt for the midrange GPUs, etc to get these into more laptops and desktops.

When the GTX750 series was released,the AMD cards were made to look like relative power hogs.

During the Kepler times,the AMD midrange cards held their own - the HD7850 and HD7870 were efficient cards.

The problem is that they kept rebranding them and Nvidia appeared look ahead in the technology stakes for cheaper cards,which have been traditional ATI/AMD strong points.

Fiji also seemed more like a technology demonstrator,ie,a very expensive HD4770.

 

[FuryXGuy]

The argument that Nvidia can just use the 1080 to kill the RX 480 is not a valid one, that's Nvidia setting their own coffers alight to hurt AMD, they would burn money on everyone one they sold.

No one is saying nVidia will do that. But they can always use their technological advantage and use it in their mid range and in their next generation. There is a lot of benefit coming out of being on the cutting edge and having technological leadership.

Ignoring the high end would not only hurt AMD but it will continue to hurt them and have people view them as the "cheaper" option.

And with respect to die size. Having a larger die size doesn't mean it's better, it's about having good architecture with reasonably die size that can compete with the high end. This was implemented with the HD 48xx series and masterfully executed in the HD 58XX series.

Fury X was a smaller die size than the 980 Ti but it didn't perform well against the 980 Ti when it came out and it should have come out earlier. If it had come out earlier and performed better than the 980 Ti, I don't think it would have done worse as it did in terms of sales. Also, the 4GB limit hurt it's image.

It's not a bad card. I just think it's pricing was way to close with respect to the performance it had when compared to the 980 Ti.

 

[humbug]

No one is saying nVidia will do that. But they can always use their technological advantage and use it in their mid range and in their next generation. There is a lot of benefit coming out of being on the cutting edge and having technological leadership.

Ignoring the high end would not only hurt AMD but it will continue to hurt them and have people view them as the "cheaper" option.

And with respect to die size. Having a larger die size doesn't mean it's better, it's about having good architecture with reasonably die size that can compete with the high end. This was implemented with the HD 48xx series and masterfully executed in the HD 58XX series.

They already are, Pascal is Pascal no mater what range of GPU's the technology resides in the GTX 1060 is Pascal and there fore the best Nvidia can do right now in that range, the RX 480 is competitive, its gaining AMD marketshare despite not having any cards in the high end.

Fancy that... for the first time in ever AMD don't have a high end competitor and AMD stop the marketshare bleed, infact reverse it.

 

[FuryXGuy]

Hawaii is impressive considering is DP performance and small die,but if you look at Bonaire and Tonga they really didn't change much and performance/watt stagnated.

The problem is AMD at that point were more worried adding various features and not about engineering faster,and more efficient cards.

I don't understand why didn't fathom Nvidia would be doing that??

They should have been thinking more of improving performance/watt for the midrange GPUs, etc to get these into more laptops and desktops.

When the GTX750 series was released,the AMD cards were made to look like relative power hogs.

During the Kepler times,the AMD midrange cards held their own - the HD7850 and HD7870 were efficient cards.

The problem is that they kept rebranding them and Nvidia appeared look ahead in the technology stakes for cheaper cards,which have been traditional ATI/AMD strong points.

Fiji also seemed more like a technology demonstrator,ie,a very expensive HD4770.

Yes, I do agree they lacked a good mid range product post 7850/7870 series. That's why I stated you would have to have strong mid-range product in order to offset cost's of the developing high end product. Yes, they did lose site of the mid/low end with respect to power consumption and getting into the laptops. I can barely find any laptops with descent AMD graphics chips for gaming, unlike the 2008 - 2010 years.

With respect to Fijii, it should have come out earlier and should have cost's less. nVidia blindsided AMD when they released the 980 Ti right before Fury X came out. If they had came out in 2014, let's say couple of months after Maxwell launched, then it would have been a better launch product. Having to go up against the 980 Ti at the price it launched, relative to the performance it had, was going to be an uphill battle.

Similar to what happened with the R600, it was late to market and was a larger chip and the performance wasn't great compared to the price it launched and the performance it had relative to the 8800 series GPUs.

 

[Rroff]

They already are, Pascal is Pascal no mater what range of GPU's the technology resides in the GTX 1060 is Pascal and there fore the best Nvidia can do right now in that range, the RX 480 is competitive, its gaining AMD marketshare despite not having any cards in the high end.

Fancy that... for the first time in ever AMD don't have a high end competitor and AMD stop the marketshare bleed, infact reverse it.

You aren't understanding what people are saying - sure it is doing them well now but they can't ride that horse forever - if they keep pushing cards at the highest in the RX480 bracket in a few years max their market share will utterly bottom out.

As things are it has been a good move for them - though I still contend it would have played a little better with another card above the 480 that was more competitive with the 1070 but its not a strategy they can always play.

 

[CAT-THE-FIFTH]

Yes, I do agree they lacked a good mid range product post 7850/7870 series. That's why I stated you would have to have strong mid-range product in order to offset cost's of the developing high end product. Yes, they did lose site of the mid/low end with respect to power consumption and getting into the laptops. I can barely find any laptops with descent AMD graphics chips for gaming, unlike the 2008 - 2010 years.

With respect to Fijii, it should have come out earlier and should have cost's less. nVidia blindsided AMD when they released the 980 Ti right before Fury X came out. If they had came out in 2014, let's say couple of months after Maxwell launched, then it would have been a better launch product. Having to go up against the 980 Ti at the price it launched, relative to the performance it had, was going to be an uphill battle.

Similar to what happened with the R600, it was late to market and too larger and the performance wasn't great compared to the price it launched and the performance it had relative to the 8800 series GPUs.

Its a shame since they increased die size by 50% over the HD7800 series with the R9 285 and R9 380 series.

Nvidia went from using a salvage 300MM2 die to a one barely larger than 200MM2 with half the RAM chips.

The R9 285/R9 380 are better cards than a GTX960 but Nvidia must have been making much more profit per card.

At least with Polaris 10 and 11 things are much closer IMHO OFC.

 

[humbug]

You aren't understanding what people are saying - sure it is doing them well now but they can't ride that horse forever - if they keep pushing cards at the highest in the RX480 bracket in a few years max their market share will utterly bottom out.

As things are it has been a good move for them - though I still contend it would have played a little better with another card above the 480 that was more competitive with the 1070 but its not a strategy they can always play.

They don't need to and they ain't, Polaris is just a tweaked and die shrunk Tonga gap plug while they get on with developing their own brand new architecture.

It may or may not be better in some ways, or every way than Pascal, or Volta..... we know nothing about it.

What i said and still am is that if its not, if AMD cannot make it cheap enough for enough performance to fight Nvidia at the enthusiast end they should not bother.

These things do not gain them a reputation for anything other than failure and at a financial cost they cannot afford.

This suits Nvidia consumers in the sort term, which is why the argument keeps getting made that AMD should make cheap top end cards come what may, it does nothing but harm to AMD in the short and long term.

 

[FuryXGuy]

They already are, Pascal is Pascal no mater what range of GPU's the technology resides in the GTX 1060 is Pascal and there fore the best Nvidia can do right now in that range, the RX 480 is competitive, its gaining AMD marketshare despite not having any cards in the high end.

Fancy that... for the first time in ever AMD don't have a high end competitor and AMD stop the marketshare bleed, infact reverse it.

I am not arguing against the RX 480. But they still don't have a card to compete with the GTX 1070 and GTX 1080. If they did their profit margins would be higher and they would be in lock step with nVidia in terms of technology.

That is my argument. You can't just focus on the mid end. You should try to capture the mid end AND the high end. I for example read about people who wanted the best and had previously brought AMD's high end (ie the HD 6970) but would like to buy something similar to the GTX 1070/1080 from AMD side but since AMD doesn't have anything to compete with that they went with the GTX 1080/1070. That's one lost sale of AMD.

They did stop the market share bleed. But what else could they have done to get more market share? Do you think they would have gained more market share just with the RX 480 or with the RX 480 + GPUs that compete with GTX 1080/1070?

I think the answer is clear. That's what me and others have been saying, they can't just ignore the high end if they want to maximize their market share and their profits. Not to mention risk of getting trampled by nVidia buy them consistently putting out better and better parts.

This is what cost them in the CPU market, where intel just kept on pushing out better and better parts and then eventually suffocating AMD and essentially pushing them out of the server market. If they had been in lock step with intel in terms of technology they not only would have been able to compete with intel in the high end and in the mid end also (without taking losses).

 

[CAT-THE-FIFTH]

Yeah,AMD if fighting the £300+ market need to launch decent cards. The sub £300 market is a bit more tolerant of less than perfect launches IMHO.

Its another issue - in the last few years AMD card launches always had some niggle or problem. Nvidia by comparison have almost perfect card launches.

 

[Rroff]

That is my argument. You can't just focus on the mid end. You should try to capture the mid end AND the high end. I for example read about people who wanted the best and had previously brought AMD's high end (ie the HD 6970) but would like to buy something similar to the GTX 1070/1080 from AMD side but since AMD doesn't have anything to compete with that they went with the GTX 1080/1070. That's one lost sale of AMD.

I see the other way quite a bit as well - people eyeing up the RX480 but are swayed by the 1070/1080 image wise and even with everything going for the 480 end up buying the 1060 due to association with the higher Pascal cards and/or take a lot of deciding to go with the 480 over the 1060 even when its the more obvious choice for them.

Its easy to forget that the average consumer is nothing like the average poster here.

 

[humbug]

I am not arguing against the RX 480. But they still don't have a card to compete with the GTX 1070 and GTX 1080. If they did their profit margins would be higher and they would be in lock step with nVidia in terms of technology.

That is my argument. You can't just focus on the mid end. You should try to capture the mid end AND the high end. I for example read about people who wanted the best and had previously brought AMD's high end (ie the HD 6970) but would like to buy something similar to the GTX 1070/1080 from AMD side but since AMD doesn't have anything to compete with that they went with the GTX 1080/1070. That's one lost sale of AMD.

They did stop the market share bleed. But what else could they have done to get more market share? Do you think they would have gained more market share just with the RX 480 or with the RX 480 + GPUs that compete with GTX 1080/1070?

I think the answer is clear. That's what me and others have been saying, they can't just ignore the high end if they want to maximize their market share and their profits. Not to mention risk of getting trampled by nVidia buy them consistently putting out better and better parts.

This is what cost them in the CPU market, where intel just kept on pushing out better and better parts and then eventually suffocating AMD and essentially pushing them out of the server market. If they had been in lock step with intel in terms to technologically they not only would have been able to compete with intel in the high end and in the mid end also (without taking losses).

I'm not saying AMD should step out for no good reasons.

Read my response to Roff above you.

 

[FuryXGuy]

I'm not saying AMD should step out for no good reasons.

Read my response to Roff above you.

Well I do agree that they should come up with a high end that doesn't cost too much too produce. That's not what I am arguing about. I am arguing about completely surrendering the high end which you were advocating by telling only to focus on the mid-end. Which is silly.

They have competed on the high end as well as mid end before. The 9700 Pro/9800 Pro line ups, X800 line ups, and with the 7970 line ups. Even with the HD 200 series line up I would say they had a descent line up in the mid-end.

 

[humbug]

Well I do agree that they should come up with a high end that doesn't cost too much too produce. That's not what I am arguing about. I am arguing about completely surrendering the high end which you were advocating by telling only to focus on the mid-end. Which is silly.

That is silly.

And speaking of silly, the Fury-X is the silly sort of thing i'm talking about, a card like that just because it has HBM?

Fiji was way to big, way too expensive to make for the performance it had, all it did was set in stone AMD's reputation of getting it wrong at the enthusiast range.

So unless they have a GPU that like Pascal is nice and small, efficient, cheap to make and fast at the high end AMD should not bother.

A 300mm^2 sub 200 watt chip with cheap memory IC's is not going to be an horrendous anvil around their neck if Nvidia decide to hack £100 + off the 1080 to deny it sales.

 

[LoadsaMoney]

Well I do agree that they should come up with a high end

They can't.

 

[FuryXGuy]

I see the other way quite a bit as well - people eyeing up the RX480 but are swayed by the 1070/1080 image wise and even with everything going for the 480 end up buying the 1060 due to association with the higher Pascal cards and/or take a lot of deciding to go with the 480 over the 1060 even when its the more obvious choice for them.

Its easy to forget that the average consumer is nothing like the average poster here.

This is true. There is a lot of association with respect to "image wise". You see that in the car market. Audi wouldn't be Audi if they didn't compete with BMW as a sports Luxury brand in the high end. If Audi just produced mid end cars and if BMW had produced a mid-end car at the same price as Audi's some people would choose the BMW mid-end car because of BMW's association with producing high-end luxury car despite the fact that mid-end Audi car may have the same features or even have better features than the mid-end car, it may even have all-wheel-drive which is more expensive to implement and the BMW may not even have a all-wheel-drive but people may still choose the BMW without all-wheel-drive and less features because people wouldn't consider BMW as high-end luxury brand.

AMD can't compete with nVidia just on the mid end forever because people will associate with AMD as the cheaper brand, then it already is. This is what Lisa Su alluded to as AMD can't be considered just the cheaper brand. That's why they need to continue to produce 7970's, R9 290X's because that has an effect on their mid-tier, lower tier image wise. ATI wouldn't have been able to get over 50% market share in the mid-2000's if they had just focused on the mid-end. They continued to pump out high end cards like the 9700 Pro/9800 Pro/9800XT the X800 XT PE/X850 XT which finally enabled them to take more market share from nVidia for the first time because people started to associate ATI with making high end cards.

If AMD continues to do that they will end up being like Hyundai. In the late 90's/early 2000's Hyundai was able to capture significant market share by focusing on the low to mid end market. But when they started to make high end luxury cars they didn't make much headway because the brand has been tainted with being makers of cheap cars. That's why they started the Genesis brand representing the mid to high end car. As a matter of fact they don't even put the Hyundai logo on the cars but a completely different logo, the Genesis logo on the cars because they are trying to distinguish those range of cars from the cheaper Huyndai cars.

I would argue that AMD doesn't nearly have a big challenge as Hyundai has with respect establishing Genesis as a premium brand as Genesis would have to go up against BMW, Audi, Mercedes Benz, which are already well established luxury brand for decades.

All AMD needs to do is put out high performing GPUs that outguns nVidia for 2 - 3 generations, kind of like what ATI did with subsequent generations with the 9700 Pro/9800 Pro/9800 XT/X800 XT PE to establish itself as a maker of high performance enthusiasts cards and finally beat nVidia and cross that 50% market threshold or at least get close to that.

But we all know that is easier said then done.

 

[FuryXGuy]

That is silly.

And speaking of silly, the Fury-X is the silly sort of thing i'm talking about, a card like that just because it has HBM?

Fiji was way to big, way too expensive to make for the performance it had, all it did was set in stone AMD's reputation of getting it wrong at the enthusiast range.

So unless they have a GPU that like Pascal is nice and small, efficient, cheap to make and fast at the high end AMD should not bother.

A 300mm^2 sub 200 watt chip with cheap memory IC's is not going to be an horrendous anvil around their neck if Nvidia decide to hack £100 + off the 1080 to deny it sales.

You read nothing about what I stated earlier. They need to make a high end card that is not expensive to make.

Why in the world are you comparing the Fury X to the GTX 1070/1080 they are completely different generation on a totally new node. That is silly.

Fury X should be compared to the 980 Ti and it was poorly executed and yes it was expensive to make but there were other factors that didn't make it as competitive like it came late to market and had closed loop cooler (adding to the costs) and it didn't perform as well. 980 Ti had a larger die than the Fury X by the way.

So, I would pin the Fury X not selling as much to the reasons mentioned above mainly the performance.

 

[FuryXGuy]

They can't.

Not yet at least.

I think Raja Kudori may have something to say about that.

https://www.youtube.com/watch?v=24rFeHo1S-o

 

[queamin]

In dx12 fury not that far behind 1070 and people are saying Vega will have trouble beating a 1070 come off it, Nvidia are on the back foot here with dx12 it is only because Amd has not got any high end cards out yet.

 

[hernaldo]

OP is comparing apples to oranges and doesn't take into account any of the financial situations that AMD and ATI was facing in early 2000s.

To put it kindly, it's a terrible OP and is akin to measuring time with a tape measure in my humble opinion.

 

[LoadsaMoney]

In dx12 fury not that far behind 1070 and people are saying Vega will have trouble beating a 1070 come off it, Nvidia are on the back foot here with dx12 it is only because Amd has not got any high end cards out yet.

LOL