[Anandtech] Discrete Q2 GPU Marketshare - AMD rises to 30%

[USER8000]

People should look at the latest Steam survey:
http://store.steampowered.com/hwsurvey/videocard/

There are no Polaris cards in there now - must have been the Aliens from Fallout which abducted them all,since they were there in the last survey.

But,valid shipping figures need to be ignored - just remember that folks!! Companies like JPR and Mercury Research have never been used by AMD and Nvidia,right??

http://www.nvidia.com/object/IO_20020429_6513.html

Oh,wait, isn't that Nvidia using the same figures some people here are telling to ignore? Seems Nvidia is fine using them both so we should all use it!!

 

[Dribble]

People should look at the latest Steam survey:
http://store.steampowered.com/hwsurvey/videocard/

There are no Polaris cards in there now - must have been the Aliens from Fallout which abducted them all,since they were there in the last survey.
!

Perhaps less sarcasm and more reading would help. They are there in the DX10/11/12 systems page - this is the same August survey page that's been banded around this forum since it came out. There are not enough of them to show up on the PC Video card usage.

 

[AtenRa]

There is a RX 480 and 460 in Win 10 + DX-12 GPUs

AMD Radeon RX 480
-
-
0.00%
0.05%
0.11%
+0.06%

AMD Radeon RX 460
-
-
-
-
0.02%

http://store.steampowered.com/hwsurvey/directx/

 

[USER8000]

Perhaps less sarcasm and more reading would help. They are there in the DX10/11/12 systems page - this is the same August survey page that's been banded around this forum since it came out. There are not enough of them to show up on the PC Video card usage.

Perhaps less sarcasm and more reading would help - people like you only want to take JPR and Mercury Research figures when it suits you and your mates agendas. However,since Nvidia is using the same companies figures on their website,its really interesting how all of a sudden the figures don't count and Nvidia uses those figures for a reason since they cover the full market.

However,the same people all of a sudden were quoting the same companies figures when Nvidia gained marketshare and there was no issue.

If you have an issue with Nvidia using JPR and Mercury Research then you should take it up with them.

 

[96Firebird]

I'm not sure what is worse, losing marketshare and turning a loss, or gaining marketshare and still turning a loss...

 

[USER8000]

AMD rolled in both GPUs and CPUs into one segment of the company,ie,Computing and Graphics. Hence,that segment is also accounting for AMD CPU sales too. It probably was a sneaky move by AMD to mask the bigger issues they have on the CPU side,although they seemed to have significantly reduced the losses from the same point last year,so all of this is having an impact.

 

[zinfamous]

I'm not sure what is worse, losing marketshare and turning a loss, or gaining marketshare and still turning a loss...

well, they've lessened the bleeding. gaining marketshare and turning less of a loss are still improvements from AMD's perspective, otherwise it's not an informative claim that you make without looking at where they were before.

It's not all roses, though, obviously, because I agree with the earlier poster that a lot of this supposed gain in unit adoption comes from nVidia shipping less units. But as I've said dozens of times before in these forums, we won't really know anything until Q3 numbers are released. Most everything else so far is speculation regarding Polaris and Pascal sales.

When you look at AMD today compared to this time last year, the year before that, and even Q1 of 2016, they are obviously doing better, but it's not a terribly enviable position.

 

[turtile]

I think the biggest reason AMD planned Polaris was the fact that they didn't have enough R&D to spend on a high end Polaris and build Vega. And the most important part is the wafer agreement. AMD is barely selling CPUs so my assumption is that AMD had many wafers it would have to buy without any use. So they moved GPU production to GF so they wouldn't lose on the agreement. And why not gain market share to get your architecture in use for more developer exposure instead of throwing money out of the window on unused wafers...?

 

[raghu78]

AMD rolled in both GPUs and CPUs into one segment of the company,ie,Computing and Graphics. Hence,that segment is also accounting for AMD CPU sales too. It probably was a sneaky move by AMD to mask the bigger issues they have on the CPU side,although they seemed to have significantly reduced the losses from the same point last year,so all of this is having an impact.

yeah thats the important information. AMD is masking abysmal CPU/APU (Computing) sales by combining with dGPU (Graphics). i am guessing the dGPU is probably bringing in more revenue than CPU/APU. Its likely that the dGPU is probably at breakeven or slight loss (Graphics) but the CPU/APU (computing) is contributing to the majority of the losses.

The Q2 market share numbers are decent for AMD. But we have to wait and see if this is a temporary bump for AMD since Nvidia was clearing out older Maxwell inventory. Nvidia's guidance for Q3 2017 is massive at USD 1.7+ billion. I think AMD is going to struggle to keep the 30% market share as Nvidia would have GP107 release in Oct and thus their entire stack would be out in the market GP102 , GP104 , GP106, GP107. AMD has only 2 chips released Polaris 10 and Polaris 11 (3 GPUs total). Vega does not seem to be anywhere close to release. I think the earliest launch prospects for Vega are mid-late Q1 2017. Even then I think the smaller Vega (around 350 sq mm) is probably the one to launch and will face off against GP104. The bigger Vega chip might launch at the end of H1 2017 or early Q3 2017.

 

[DeathReborn]

AMD was specifically talking about mainstream low priced VR ready cards since December 2015 and you are seriously implying AMD released the RX 480 at $200/239 because of NVIDIA GTX 1070/1080 prices ?? Is this a joke or what ??

Thank god we had NVIDIA releasing the GTX 1080 FE end of May at $699 and then GTX 1070 FE 10th of June at $449 and forced AMD to release at the end of June the RX 480 4GB at $199 and 8GB at $239.

It probably stopped the RX 480 8GB from being $299, after all up to $300 was their TAM.

 

[Pariah]

AMD was specifically talking about mainstream low priced VR ready cards since December 2015 and you are seriously implying AMD released the RX 480 at $200/239 because of NVIDIA GTX 1070/1080 prices ?? Is this a joke or what ??

Thank god we had NVIDIA releasing the GTX 1080 FE end of May at $699 and then GTX 1070 FE 10th of June at $449 and forced AMD to release at the end of June the RX 480 4GB at $199 and 8GB at $239.

No, it's not a joke. As someone else said earlier, scale back the sarcasm and concentrate more on reading. The 1070 and 1080 were known quanties long before the 480 was announced. When I said AMD heard what Nvidia was doing I was referring to rumours about the 1060 leaking out. If you think there is no truth to that affecting the 480 launch, then please tell us why the first batch of 4GB 480's were actually 8GB cards.

 

[Piroko]

It probably stopped the RX 480 8GB from being $299, after all up to $300 was their TAM.

The original statement actually was that they would increase the TAM significantly beyond what was available at the time (~330$ HD290?):
http://arstechnica.com/gadgets/2016/04/amd-polaris-will-be-a-mainstream-gpu/

I wouldn't be surprised if their worst case plan was to launch the HD480 at as low as 200$ for the 8GB variant (it is a small enough chip to do so) but they chose to push the clocks and sell it more expensive. The 4GB variant was a backup plan because inventory level wasn't good enough to launch the 470 earlier as the replacement 200$ SKU.

 

[AtenRa]

No, it's not a joke. As someone else said earlier, scale back the sarcasm and concentrate more on reading. The 1070 and 1080 were known quanties long before the 480 was announced. When I said AMD heard what Nvidia was doing I was referring to rumours about the 1060 leaking out.

AMD is talking about mainstream GPUs and increasing the VR TAM with Polaris before NVIDIA released any Pascal cards. So yes it looks like a joke when people suggesting that RX 480 4GB MSRP at $200 was made because of NVIDIA Pascal GPUs.

If you think there is no truth to that affecting the 480 launch, then please tell us why the first batch of 4GB 480's were actually 8GB cards.

I dont know, you may ask AMD if you want. I can only guess it was made because of lower volumes of Polaris 10 Chips ?? Sell some of the 8GB cards as 4GB $200 models and then when we have more Polaris 10 chips we make the real 4GB cards ??

 

[zinfamous]

If you think there is no truth to that affecting the 480 launch, then please tell us why the first batch of 4GB 480's were actually 8GB cards.

That was explained rather clearly at the time of release: Supply of 1gb GDDR5 chips was nonexistent at the time, so all of those cards were populated with the 2gb chips because it was all that was available within the release schedule.

There were some stories/links in the 480 or whatever threads here during the week of that release when word started to get around.

 

[cbn]

Market share is relevant. But its just as important to understand the product mix of it. Specially when the market is moving upwards in SKUs while the overall segment is decreasing.

I wonder how much of this is due to CPUs at the lower end (eg, Core i3) not being able to keep up with dGPU progress? This vs. iGPU size increase.

My guess is that it is the former that has the stronger role. (ie, it is not the iGPU size increase that causes the decrease in dGPU market share, but rather the number of CPUs capable of feeding the increasingly more powerful dGPUs that is decreasing).

 

[swilli89]

I think the biggest reason AMD planned Polaris was the fact that they didn't have enough R&D to spend on a high end Polaris and build Vega. And the most important part is the wafer agreement. AMD is barely selling CPUs so my assumption is that AMD had many wafers it would have to buy without any use. So they moved GPU production to GF so they wouldn't lose on the agreement. And why not gain market share to get your architecture in use for more developer exposure instead of throwing money out of the window on unused wafers...?

Very clearly obvious. AMD didn't have the cash to build both high end and low end cards, only had resources to pay a team to do one so they chose low end as that probably afforded them more OEM design wins. Only question is if they had more resources would Vega have been brought to market by now or would there have been a scaled up Polaris large enough to fight GP104? That's really the question.. Vega is rumored to be a large revamp of GCN so its not just larger but a truly modified uArch, but just how different is it?

 

[tajoh111]

AMD was specifically talking about mainstream low priced VR ready cards since December 2015 and you are seriously implying AMD released the RX 480 at $200/239 because of NVIDIA GTX 1070/1080 prices ?? Is this a joke or what ??

Thank god we had NVIDIA releasing the GTX 1080 FE end of May at $699 and then GTX 1070 FE 10th of June at $449 and forced AMD to release at the end of June the RX 480 4GB at $199 and 8GB at $239.

The gtx 1070 launching with dual price points was incredibly business savvy but greedy. It succeeds on a marketing point but at the same time capitalizes on AMD ability to not compete.

If AMD had been closer to the gtx 1070, Nvidia could release a designated gtx 1070 the same day, that actually launched at 380. It was completely in their capacity because we saw barebone consumer cards at tradeshows, which were supposedly the 380 kind.

However 400 dollar 1070 do exist. Also to get back to the matter, the 380 1070 is an important matter because AMD would have to use this benchmark cost as a comparison point for it's rx 480. Reviews would use this figure along with the founders edition pricing and the rx480 would get hammered if it launched at $300. Finfet is an expensive node, and it would have been justifiable to launch 300 dollars.

However with the 1070 being dramatically faster than it, 300 dollars doesn't make sense.

AMD was quoting the 350 and under market original when talking about Polaris originally.

And look at the original price of the 7870. 350 dollars.

If AMD original intention was to release the 240/200. Some in their marketing department needs to be fired.

THis is because look at all the pricing. It's priced too closely together.

rx 480 4/8gb, 200/240.

rx 470, 4/8gb 180,/210

rx 460, 2/4gb 120/140.

The Problem is AMD pricing encourages cannibalism. A rx 470 4 gb sale will rightfully take an rx 460 4gb sale and a rx480 4gb sale will rightfully take the sale away from a rx480 8gb, rx 470 4/8gb.

Add in that the rx 480 has the lowest margins and among the lowest volume productions(since it requires a full die) and AMD priced their products terribly. The only thing that makes the rx 470 sellable is the rx 480 4gb being out of stock. And the rx 460 4gb/2gb are just bad buys in general. The rx 470 is 70-80% faster and only costs 40 dollars more. Add in it most ties the gtx 950 which launched a long time ago at a price of 160 a years ago(street price is now 100-120 dollars), which means the market is very saturated and provides little upgrade incentive to these previous card owners and it's no wonder it was panned in reviews.

Also, if anyone expects 480 to perform anywhere near 1070 while having 1/3rd of rop performance. is delusional. the card was designed to replace 380/x which costed 200-240$ at their launch while having half vram. nothing to do with 1070's pricing.

The 1070 is incredibly cut down and it has less shaders than even the rx470. Yet it beats the rx480 on average by 40-50%. They should be much closers.

The problem for AMD stands with tonga. Tonga and the gtx 980 only differ in size by 12 percent yet the later is 50% faster. Add in AMD higher transistor density which reduces yields and make the cost to produce very similar and you have simply a poorly designed GPU. These problems carried over to polaris and what is ultimately leading to the poor profit margins for AMD.

These differences is what lead to the huge gap between the gtx 1080. It shows the stagnation of GCN as an architecture.

People forget, Nvidia gm204 lineup used to be their midrange lineup. And Nvidia used to have to use a bigger die than AMD to compete with them, but that changed dramatically in Nvidias favor during maxwell.

If you think the gtx 1070 pricing has nothing to do with the rx 480, your a fool.

Look at the r9 285, pricing. AMD was charging 250 for a cut down card prior to any competition. Gtx 970 launches with 50 percent better performance for 330 and kills r9 285 demand and forces AMD to rebadge and reprice tonga into the 380 and 380x and to be price at 240 and 200. Nvidia well priced card have forced AMD to reconsider their pricing as AMD has done the same with the gtx 280 and 4870 pricing.

That was explained rather clearly at the time of release: Supply of 1gb GDDR5 chips was nonexistent at the time, so all of those cards were populated with the 2gb chips because it was all that was available within the release schedule.

There were some stories/links in the 480 or whatever threads here during the week of that release when word started to get around.

No way was this the case.

The rx 480 4gb is only specified to use 1750mhz 1gb ddr5 which is very plentiful considering it is found on most maxwell products and is still being used on some products today including AMD's lower lineup.

The fact that the first batch of rx480 4gb not only had 8gb of memory, they used 2000mhz vs 1750mhz shows it was a late move knee jerk release. There was no way the 1750mhz 1gb drr5 was low in supply.

 

[AtenRa]

For the last time, AMD was talking increasing the VR TAM since January 2016 and Mainstream GPUs as first priority in the Capsaicin event March 14th, months before NVIDIA announced the Pascal GTX 1080 in May 7th.

So people need to get their facts straight, no NVIDA Pascal card pricing had anything to do with the price of Polaris 10 in RX-480 and 470. The $199 RX 480 4GB price was planed way before any NVIDIA Pascal card announcements.
Actually it is the other way around, NVIDIA was originally going to release GTX 1080 at $699, GTX 1070 at $449 and GTX 1060 at $300. But because of Palaris 10 RX 480 4GB $199 price, they invented the Founders Edition BS and released the same products with dual pricing. GTX 1060 FE at $300 is way overpriced if we compare it to the $199 RX 480 4GB, but GTX 1060 6GB (non Founders BS Edition) at $249 is extremely competitive vs RX 480 8GB $239. We have those competitive mainstream cards because of AMDs low Polaris prices not because NVIDIA wanted a GTX 1060 6GB at $249.

 

[Dribble]

Personally I think Polaris has not hit it's clock speed targets due to using GloFo node not TSMC. If it had clocked a bit higher it would have slotted in under the 1070 nicely and AMD could have charged $300, perhaps even more. That I would have thought was the original plan, and probably GloFo was promising better silicon then it delivered.
It's been discussed in the past how the Ati part of the company is upset at being hamstrung to the CPU division and would really like to get bought out. I get the feeling part of that is being forced to use inferior silicon due to the AMD wafer agreement, which means even if they equal Nvidia in design (impressive given their smaller design team) they still loose.

 

[SolMiester]

Except the Fury does handle 4k very well, it even performs nearly the same as the 1070.

https://www.techpowerup.com/reviews/ASUS/GTX_1060_STRIX_OC/26.html

Its performing closer to the 8GB 1070 than the 6GB 980 TI. Guess HBM does work better than GDDR5 after all.

I think you should re-read the post, you seem to have got lost there somewhere

 

[tajoh111]

For the last time, AMD was talking increasing the VR TAM since January 2016 and Mainstream GPUs as first priority in the Capsaicin event March 14th, months before NVIDIA announced the Pascal GTX 1080 in May 7th.

So people need to get their facts straight, no NVIDA Pascal card pricing had anything to do with the price of Polaris 10 in RX-480 and 470. The $199 RX 480 4GB price was planed way before any NVIDIA Pascal card announcements.
Actually it is the other way around, NVIDIA was originally going to release GTX 1080 at $699, GTX 1070 at $449 and GTX 1060 at $300. But because of Palaris 10 RX 480 4GB $199 price, they invented the Founders Edition BS and released the same products with dual pricing. GTX 1060 FE at $300 is way overpriced if we compare it to the $199 RX 480 4GB, but GTX 1060 6GB (non Founders BS Edition) at $249 is extremely competitive vs RX 480 8GB $239. We have those competitive mainstream cards because of AMDs low Polaris prices not because NVIDIA wanted a GTX 1060 6GB at $249.

That's backwards.

The dual pricing of gtx 1080/1070 was launched way before the polaris pricing was known. So its idiotic to think that the 199 pricing had anything to do with 599/380 pricing. At that point, AMD pricing was an unknown quantity. Hence it's impossible to say Nvidias dual pricing was in reaction to AMD. It only makes sense because Nvidia's pricing and performance were a known quantity, that it was AMD that reacted. Not the other way around. And because of this, gtx 1060 pricing was likely not effected by polaris because their current 300/250 strategy is priced relative to other founders edition/base pricing. If their original attention was to only price it at 300, it would be poorly price and sell very badly since gtx 1070 can be found for 400 and are much faster. Because it is a mainstream offering, the gtx 1060 has to have strong price to performance.

Nvidia beside their top end price their products pretty fairly and aggressively. The fact that the gtx 1070 has similar price to performance as the rx480 shows this.

GP104 performance probably caught everyone off guard. Essentially, noone expected the gtx 1080/1070 to be clocked between 1700-1800mhz. From AMD's own experience, and the lackluster polaris clocks, they probably expected 1450-1500mhz from Nvidia.

Prior to the launch of the gtx 1080, who expected a 314mm2 gtx 1080 to beat a gtx titan x by 30%? Based on AMD own experience, they likely expect gp104 to be alot slower. Their performance per transistor decreased from 390x to Polaris 10, so they probably didn't expect much from Nvidia.

Why Nvidia released dual pricing has little to do with AMD and everything to do internally with their partners.

The founders edition was invented to increase margins for partners. Basically what the founders editions has allowed is for partners to increase their profit margins by 30 to 70 dollars which doubles their margins which were in the 10 to 15% area before.

This is because prior to this founders editions crap, Nvidia's attractive reference design, forced board partners to sell after markets at pricing similar and often below the reference solution. Looks matter. What adding a premium to the reference design does it allows partner to sell stripped down products at the lower end(where this wasn't even possible as they wouldn't sell) and sell them at the minimum price, while attach significant premiums to aib cards, which are far more expensive a premium in the past. Basically everything without a stripped down cooler has a lightning or classified like premium over the lowest base price. Where on the gtx msi gaming 1080 x, which used to have a premium of just 30 dollars over the base lower price as was the case with the 980/980 ti, it now has 100 dollar premium!!

This is the same upcharge as EVGA hybrid and MSI seahawk editions!!! Msi lightning have nearly the same premium and these are limited edition cards. Now Partners like MSI are charging 150 dollars for hybrid cards.

Because 380/599 cards are so barebones and Nvidia's reference card costs 450/699, partner cards sell at these ridiculous upcharges. Overpriced founders editions add to increase the appeal of partner cards rather than decrease the appeal as the original reference cards did. Partners are very happy Nvidia did this and it adds a secondary effect.

What this means for Nvidia is partner that sell both Nvidia and AMD gpu's are way more likely to advertise for Nvidia than AMD. You hardly see any adds for AMD card ads from MSI, Asus or gigabyte. This founders edition has allowed these partners to double or triple their margins, and thus partners prefer advertising Nvidia cards.

For all your theories, how do you know AMD priced their cards at 240/200 when they were as vague as possible when it came to pricing and performance?

How do you know AMD planned to price their card 199/240 prior to the 1070/1080 price and performance were revealed? Everything indicates otherwise.

From the first batch of rx480 4gb's having 8gb of physical memory that were at a spec'ed at speeds that the rest of the batches were not going to get(shows a last ditch/last minute change) to their bad market segmentation, nothing today indicates 240/200 was the original intention.

If $199, rx480 were originally in the cards, they would of had rx 480 with 4gb of physical memory that were only speced at 1750mhz from day 1. It BS that there was low quantities of 1gb 1750mhz modules since it has been used in most of maxwells lineup for years and continues to be used on most of AMD's polaris cards today. It scream last minute change and poor planning. In other words the 199 pricing was a knee jerk reaction from AMD, that backfired because it reduce the appeal of the rest of AMD lineup. And this leads to the second point, AMD pricing for their product stack is so bad that if AMD planned the pricing on their products like this from the beginning, it would mean morons are running the company.

AMD pricing, particularly when you add partner cards to the mix, cannibalize their own sales. AIB rx470 4gb should not be the same price as a ref rx 480 4gb. AIB rx 470 8gb should not be the same price as ref rx 480 8gb. The rx 460, 4gb, which should be AMD highest volume card, should not have the worst price to performance and AMD rx 480 4gb, should not be the best value when it is the lowest volume and lowest margin product. If not for mining, it's a pricing disaster.

If AMD did not decrease their price before launch and the rx 480 8/4gb still launched at 300/250, it would still achieve AMDs goals on increasing their VR capable install base(nowhere in their early presentations do they mention actual pricing of polaris). Lowering the price to 240 and 200, has no effect on this install base because the main bottleneck is still the price of a VR headset. People that can spend 500+ on VR gear, are not more likely to get into the vr ecosystem because of a 50 dollar decrease on a videocard.

Also if you do a videocard pricing BOM estimate based on the cost of previous BOM of other videocards and you add the increased cost of finfet chips, you will see how grim the margins are on Polaris cards. The the 6870 cost AMD 116 dollars to produce, and used manufacturing tech on a very mature 40nm node GPU which is half the cost of todays 14/16 finfet designs(which should add 30-40 dollars to cost). Add in partners cut, retailers cut, distributers cost, packaging, and you will see AMD priced their cards as low as they can go, while making some profit.

 

[RussianSensation]

^ The entire analysis above would have made a lot more sense except for the fact that Polaris 10 is an R9 380/X market segment replacement. For the 100th time, we have 4 designating a new generation, and the 2nd number is the class of card it is replacing:

R9 380/X -> RX 480

R9 380 was a $199 videocard and R9 380X was a $229 videocard. These were not $299-349 videocards:

http://www.anandtech.com/show/9387/amd-radeon-300-series/2

http://www.anandtech.com/show/9784/the-amd-radeon-r9-380x-review

The other major flaw in the argument is that AMD already had a mid-range $329 R9 390 8GB a long time ago before Polaris 10 launched. How would AMD aim to expand TAM if they released a $299-$349 RX 480 and $249-279 RX 470 when they already were selling $250 R9 290 and $330 R9 390? The logic makes no sense. The only way to try and expand TAM was to bring R9 290/970 VR-spec to lower price levels. It's why AMD emphasized how the $349 VR-minimum was too high of a price for the average/mainstream consumer.

Secondly, next generation similar tier cards tend to replace the previous level tier at similar TDP levels. What does that mean? It means in much the same way how 140-150W 970 and 170-180W 980 were replaced by 140-150W 1070 and 180-190W 1080, AMD would need to replace the R9 390/X with a card having ~ 200-225W TDP, leaving 250-300W for the Fury/X replacement [of course we don't know yet if AMD will abandon the 250-300W tier this generation because this segment for AMD barely sells].

Does it look like Polaris 10 was meant to be an R9 390/970/980 replacement meant to compete with GTX1070/1080?! Absolutely not.

Thirdly, previous rumours indicated that AMD's flagship succeeding Fury X would be 15-18B transistors. How can you have such a massive chip and such a small Polaris 10 and nothing in the middle? That's because there is a mid-range tier AMD has not released yet.

This is the same flawed argument HardOCP used when Polaris 10 launched, completely misunderstanding AMD's naming convention, market strategy, product planning and SKU positioning.

It doesn't get more obvious:

RX 470 replaced R9 270X
RX 480 replaced R9 380/380X

You might as well call this the mainstream GPU segment. AMD is yet to release the performance and high-end segments.

We don't know if AMD will have 2 more tiers because companies change their strategies. At the very least we can somewhat reasonably estimate that AMD should have at least 2 more distinct tiers above Polaris 10 -- the $329-449 segment (hint: R9 390/X replacement tier), and $549-699 segment (hint: Fury/X replacement tier).

The idea that AMD planned to have RX 480 at $300-350 before 1070 launched simply makes no sense.

All of this comes down to the fact that people to this date cannot fathom, believe and accept that AMD and NV are bifurcating GPU generations now. 7970/680 = mid-range and the high-end tier got pushed back by both firms as 780/780Ti/R9 290/X. Next generation we got the exact same thing where 970/980 were mid-range and because AMD had no $ to design a new product, they refreshed 290/X as 390/X. Then we got the high-end which was 980Ti/Fury X.

This is the 3rd generation in a row that the cycle is repeated:

RX 470/480/1060 = these are low-end / mainstream videocards. For crying out loud, the 1060 is GP106, the exact (!) lineage successor to the $149 GF106 GTS450:
http://www.anandtech.com/show/3909/nvidias-geforce-gts-450-pushing-fermi-in-to-the-mainstream

The next GPU tier is mid-range / upper-mid-range. That's GTX1070/1080 and AMD's response to date is nothing yet.

The next GPU tier is high-end/flagship. That's 1080Ti/Titan XP and AMD's response to date is nothing yet.

Finally, for nearly 2 decades of GPU generations, next gen replacement tiers are faster than the previous generation tier they replaced. RX 470/480 are not faster than R9 390/X tier, which automatically means they can't be their replacements. It's why RX 480 is 60-64% faster than R9 380, similar to how 1070/1080 are ~ 65% faster than 970/980.

It seems the whole premise that RX 480 was meant to be a $300-350 card meant to compete with GP104 simply comes down to PC gamers refusing to accept that GP104 is mid-range/upper-mid-range $199-249 GTX560/Ti tier. It's a way for them to feel better about paying $400-700 for 1070/1080 since they think: "Well, Polaris 10 was meant to be the competitor but AMD failed miserably. No competition means I have no choice but to pay these prices." That's why the next explanation they have is: "AMD failed so hard with RX 480 against 1070, they had to reduce the price from $349 to $239."

Too much hanging out on HardOCP forums, huh?

R9 270/X -> RX 470
R9 380/X-> RX 480 4/8GB
R9 390/X -> RX 490 (?)
Fury/Fury X -> Who knows what this will be called.

Since AMD is lagging behind NV in perf/mm2, and perf/watt, they need 14nm to mature for higher yields to clock Vega higher to be competitive. They need prices of larger 14nm die to drop to be able to make $$ with a larger Vega chip than GP104. They also need HBM2 to mature, hit volume production and prices to drop to be able to have $329-499 HBM2 cards that actually make $. If you want to argue that AMD should have designed a Vega hard with GDDR5X instead of HBM2, that's a fair criticism to make against AMD.

Considering for every single GPU generation, except for GeForce 4 where ATI skipped a generation entirely & HD2900/3870, NV never beat the equivalent ATI/AMD by more than 20%, and taking into account that AMD actually had faster cards last 2/3 times (7970Ghz > 680, R9 290X > 780Ti, 980Ti > Fury X), it's laughable to suggest that RX 480 was meant to cost $300-350 and meant compete with GP104, and ended up 50-80% slower:

https://www.techpowerup.com/reviews/MSI/GTX_1070_Gaming_Z/26.html

 

[zinfamous]

^all of that,but just to point out they haven't yet replaced 380X, as that will be the RX 485 refresh (with RX 490 followed by the 495 refresh--assuming those are Vega chips).

But that's only according to "AMD's official" naming structure that was leaked a month or so ago.

Granted, this was released by wcctech and/or maybe some other speculation-liberal sites, so I don't know if this is really confirmed

 

[tajoh111]

^ The entire analysis above would have made a lot more sense except for the fact that Polaris 10 is an R9 380/X market segment replacement. For the 100th time, we have 4 designating a new generation, and the 2nd number is the class of card it is replacing:

R9 380/X -> RX 480

R9 380 was a $199 videocard and R9 380X was a $229 videocard. These were not $299-349 videocards:

http://www.anandtech.com/show/9387/amd-radeon-300-series/2

http://www.anandtech.com/show/9784/the-amd-radeon-r9-380x-review

The other major flaw in the argument is that AMD already had a mid-range $329 R9 390 8GB a long time ago before Polaris 10 launched. How would AMD aim to expand TAM if they released a $299-$349 RX 480 and $249-279 RX 470 when they already were selling $250 R9 290 and $330 R9 390? The logic makes no sense. The only way to try and expand TAM was to bring R9 290/970 VR-spec to lower price levels. It's why AMD emphasized how the $349 VR-minimum was too high of a price for the average/mainstream consumer.

Secondly, next generation similar tier cards tend to replace the previous level tier at similar TDP levels. What does that mean? It means in much the same way how 140-150W 970 and 170-180W 980 were replaced by 140-150W 1070 and 180-190W 1080, AMD would need to replace the R9 390/X with a card having ~ 200-225W TDP, leaving 250-300W for the Fury/X replacement [of course we don't know yet if AMD will abandon the 250-300W tier this generation because this segment for AMD barely sells].

Does it look like Polaris 10 was meant to be an R9 390/970/980 replacement meant to compete with GTX1070/1080?! Absolutely not.

Thirdly, previous rumours indicated that AMD's flagship succeeding Fury X would be 15-18B transistors. How can you have such a massive chip and such a small Polaris 10 and nothing in the middle? That's because there is a mid-range tier AMD has not released yet.

This is the same flawed argument HardOCP used when Polaris 10 launched, completely misunderstanding AMD's naming convention, market strategy, product planning and SKU positioning.

It doesn't get more obvious:

RX 470 replaced R9 270X
RX 480 replaced R9 380/380X

You might as well call this the mainstream GPU segment. AMD is yet to release the performance and high-end segments.

We don't know if AMD will have 2 more tiers because companies change their strategies. At the very least we can somewhat reasonably estimate that AMD should have at least 2 more distinct tiers above Polaris 10 -- the $329-449 segment (hint: R9 390/X replacement tier), and $549-699 segment (hint: Fury/X replacement tier).

The idea that AMD planned to have RX 480 at $300-350 before 1070 launched simply makes no sense.

All of this comes down to the fact that people to this date cannot fathom, believe and accept that AMD and NV are bifurcating GPU generations now. 7970/680 = mid-range and the high-end tier got pushed back by both firms as 780/780Ti/R9 290/X. Next generation we got the exact same thing where 970/980 were mid-range and because AMD had no $ to design a new product, they refreshed 290/X as 390/X. Then we got the high-end which was 980Ti/Fury X.

This is the 3rd generation in a row that the cycle is repeated:

RX 470/480/1060 = these are low-end / mainstream videocards. For crying out loud, the 1060 is GP106, the exact (!) lineage successor to the $149 GF106 GTS450:
http://www.anandtech.com/show/3909/nvidias-geforce-gts-450-pushing-fermi-in-to-the-mainstream

The next GPU tier is mid-range / upper-mid-range. That's GTX1070/1080 and AMD's response to date is nothing yet.

The next GPU tier is high-end/flagship. That's 1080Ti/Titan XP and AMD's response to date is nothing yet.

Finally, for nearly 2 decades of GPU generations, next gen replacement tiers are faster than the previous generation tier they replaced. RX 470/480 are not faster than R9 390/X tier, which automatically means they can't be their replacements. It's why RX 480 is 60-64% faster than R9 380, similar to how 1070/1080 are ~ 65% faster than 970/980.

It seems the whole premise that RX 480 was meant to be a $300-350 card meant to compete with GP104 simply comes down to PC gamers refusing to accept that GP104 is mid-range/upper-mid-range $199-249 GTX560/Ti tier. It's a way for them to feel better about paying $400-700 for 1070/1080 since they think: "Well, Polaris 10 was meant to be the competitor but AMD failed miserably. No competition means I have no choice but to pay these prices." That's why the next explanation they have is: "AMD failed so hard with RX 480 against 1070, they had to reduce the price from $349 to $239."

Too much hanging out on HardOCP forums, huh?

R9 270/X -> RX 470
R9 380/X-> RX 480 4/8GB
R9 390/X -> RX 490 (?)
Fury/Fury X -> Who knows what this will be called.

Since AMD is lagging behind NV in perf/mm2, and perf/watt, they need 14nm to mature for higher yields to clock Vega higher to be competitive. They need prices of larger 14nm die to drop to be able to make $$ with a larger Vega chip than GP104. They also need HBM2 to mature, hit volume production and prices to drop to be able to have $329-499 HBM2 cards that actually make $. If you want to argue that AMD should have designed a Vega hard with GDDR5X instead of HBM2, that's a fair criticism to make against AMD.

Considering for every single GPU generation, except for GeForce 4 where ATI skipped a generation entirely & HD2900/3870, NV never beat the equivalent ATI/AMD by more than 20%, and taking into account that AMD actually had faster cards last 2/3 times (7970Ghz > 680, R9 290X > 780Ti, 980Ti > Fury X), it's laughable to suggest that RX 480 was meant to cost $300-350 and meant compete with GP104, and ended up 50-80% slower:

https://www.techpowerup.com/reviews/MSI/GTX_1070_Gaming_Z/26.html

That doesn't mean anything, if you want to go off code names, the x80 and xx8x series has actually been more expensive than 240 and 200 dollars in the past and 380/380x pricing has more to do with conceding competitive performance vs the gtx 970.

The original chip which the 380 and 380x were based on were was tonga and that started off as a 250 dollar cut down chip. I never said 350 was the price, 300 and 250 were the price i suspected.

Do you known the price 7870 when it started out? $350

You know the price of the 280x/280, $299 and $250.

Basically the 380x price had to change because of competition. Add in that it was a rebadge and it should not be a surprise that it launched at 250 and 200.

A 250 dollar and 300 dollars(for the larger memory config) was still in the mainstream pricing if the past x8xx pricing is anything to go on. Add in finfet costs more than 28nm and AMD poor pricing below and your theory doesn't hold much weight. An extra 50 dollars in pricing all around for AMD would have done wonders for AMD pricing.

 

[USER8000]

Wait,what even in the UK websites like Overclockers UK were selling the RX470 and RX480 as successors to the R9 270/R9 270X and the R9 380/R9 380X. If the pound had not been devalued due to BREXIT and BREXIT fears,they would have slotted in around £200 and under. The same goes with the two versions of the GTX1060.

Plus,look at the die sizes for the chips - Polaris 10 is around 200MM2 to 250MM2 ,which is close to the same size as previous GPUs in cards like the HD6870 and HD7870. The HD6870 launched at the same price as the RX480.

Also,using the HD7850 and HD7870 as comparisons is also forgetting history on purpose. The GTX660 was released six months after the HD7870 and that was the true competitor to the HD7870.

Nvidia made the cut down GTX660TI using a much larger chip which forced the HD7870 price down,and then AMD repositioned the HD7950 as the main competitor for the GTX660TI.

However,the GTX660TI was released in August 2012,and that was still months after the March 2012 launched of the HD7870.

That is why AMD priced it so highly,since it was competing with GTX570 cards at launch.