AMD Official Statement - Polaris is NOT delayed

[guskline]

That's not the issue at all. It's the brand and they need to do something to make the brand stand out. When you overhear dumbasses at microcenter or Best Buy give customers a speech as to why Nvidia is better for literally the dumbest of reasons, you have a brand problem. If I was a product manager, I would have a month long trip across the states and have launch parties at every decent sized city and invite these types of people.

Excellent point.

Best Buy in our area really has less and less custom parts. The Micro Center in Philadelphia (actually Wayne) is @ 1 hour 10 minutes from me BUT has a TON of custom parts. They have more experienced salespersons but that also depends.

 

[tential]

That's not the issue at all. It's the brand and they need to do something to make the brand stand out. When you overhear dumbasses at microcenter or Best Buy give customers a speech as to why Nvidia is better for literally the dumbest of reasons, you have a brand problem. If I was a product manager, I would have a month long trip across the states and have launch parties at every decent sized city and invite these types of people.

I've been saying this for the longest time now. The problem isn't how good or bad of a product p10 is but the fact Nvidia controls mindshare and nothing amd has done recently has aimed to change that. A good price/performance part means nothing because people will be told the 1070 is the best gpu no matter what and contrary to popular belief a lot of gamers Dont look at graphs or benchmarks of gpus, and even when presented with contradictory data, still will believe Nvidia is best no matter what.

You can't beat that with good prices/performance you have to market.

 

[RussianSensation]

I too thought AMD might price Polaris aggresively until it came to light they will price R9 470X at $169.Since this price is hardly what one would call aggressive,i think AMD is going for healthy profits on each card sold rather than aggressive pricing.Although its just a rumor about 470X,i feel its right.Polaris will indeed be cheaper,however don't get your hopes up on expecting better price/performance than Nvidia.

Good point. $169 for R9 470X that will be slaughtered in performance by fire-sale R9 390 cards and isn't a good deal for someone who cares more about performance and price/performance than perf/watt. I have a feeling (based on previous history of AMD card clearance) that once R9 390 starts going on clearance that we will see $200-225 R9 390 cards.

What were you expecting when their competition is selling mid range cards for enthusiast prices? AMD ain't no charity, they will try to mild the low to midrange segment as long as possible, it would be stupid not to for any business.

AMD is in a tough spot -- try to gain market share with very aggressive pricing right off the bat (aka HD4000-6000 series), but lose profits. I call that cash flow empty market share. This also moves AMD back towards the budget brand image, which is what they don't want. On the other hand, they almost can't price Polaris 10 too low since they know over time they'll need to adjust price downward when 1050/1060/1060Ti launch. It means the counter balance to very low pricing strategy is that they can sell Polaris 10/11 at higher prices until August-September when the rumoured GP106 and more cut-down GP104 are slated to release. I still think $299 for Polaris 10 with ~ 390X level of performance is too high. $299 with 980Ti/Fury X performance would be really good but the small die size, no rumours of 1.5-1.7Ghz 14nm clock speeds and limited memory bandwidth with GDDR5 suggest we have to be realistic that it's not going to be this fast.

770 2GB was $329 very quickly after the 280X release, not $399.

Nope. I remember R9 280X vs. 770 generation and every time someone was building a mid-range rig, it was impossible to recommend the 770. 770 2GB first dropped from $400 to $380 and it took 8-10 months before it went down to $330-340 range.

May 30, 2013 = $420
February-March 2014 = $340
April-June 2014 = $330

http://camelcamelcamel.com/EVGA-GeF...-P4-2774-KR/product/B00CZIQXBA?context=browse

http://camelcamelcamel.com/GTX770-D...OC-selected/product/B00D3F7CFK?context=browse

Even on the best deals, it took 770 at least 6 months before it dropped to $330
http://camelcamelcamel.com/EVGA-GeF...KR-Graphics/product/B00CZIQZ40?context=browse

The comparison to 770 2GB is also strange considering we all knew that 2GB cards are DOA for anyone cross-shopping 280X vs. 770 around that time. That means the proper comparison in hindsight for those who kept defending 2GB cards was 280X vs. 770 4GB.

For a card that cost more throughout its entire life-cycle, 770 was VRAM gimped or too expensive and aged far worse.
https://www.techpowerup.com/reviews/ASUS/GTX_950/23.html

I paid $330 for my 770-4 mb after returning two Radeon cards that had the dreaded blackscreen disease. Two computer shops couldn't fix the problem nor could I. Had never purchase Nvidia card up before this. I'm well into my 60's and ran IT before I retired so its not like I haven't traveled a few blocks. These were 290X's.

Happens. I have tried 3 different Radeon cards in the same older Asus motherboard and only 1 of 3 worked properly. Sometimes it's a major mobo compatibility issue with older boards. With my Z170 board I've tested 7970, R9 295X2, R9 390 and all of them run amazing, no crashes, no freezes, no black screens. I believe you that some Radeons had the black screen of death as I bought a refurbished 7970 off Newegg during the bitcon mining days and the card failed within 24 hours just giving me a black screen (monitor not detected). At the same time, I had many friends with dead GeForce 4 Ti 4200 and GeForce 8 series. Doesn't mean I won't recommend them NV cards. Bad cards happen to almost everyone eventually if you buy a lot over your lifetime.

I will be following the tech info from AMD and then read the analysis to see how close the Polaris releases are to the GTX 1070. I'm a realist. I seriously doubt either Polaris will challenge the GTX 1080 but why should they? Especially for the prices bantered about (up to $400.00)

Agreed. NV was vastly ahead of AMD in perf/watt with Maxwell vs. Tahiti/Hawaii GDDR5 parts. Even after they incorporated HBM1, they were still behind. That means it's realistic to expect AMD to keep trailing Pascal in perf/watt. Couple that with a smaller die size of Polaris 10 and no way it should be even close to the 1070. If AMD's 232mm2 Polaris 10 is as fast as NV's 314mm2 1070, that means for next gen vs. next gen cards AMD would have pulled off a 35% increase in performance per mm2 over Pascal. No chance I'd believe that.

I owned a GTX 970 and sent it back after the 3.5 Vram scandal was exposed. I "moved down" to a R9 290 on principal so for me it made sense BUT it seems quite evident that the GTX 970 was a marketing phenomenon, despite this Vram exposure and it made Nvidia TONS of $$ despite this. AMD had a tough time answering with HOT Hawaii.

NV's viral marketing on forums/online, professional invested YouTubers and Twitch Gamers who are sponsored by NV/product reviewers did a horrible job (or a good job if you are NV) of portraying the true reality of Hawaii cards in after-market form.

They need a price AND performance competitive gpu to the GTX 1070.

Unfortunately, I only see two paths for AMD. Price cards higher to make $ and relay on objective/AMD buyers or deliver more performance at the same price as NV. Delivering less performance at a lower price (i.e., price/performance) has failed AMD since HD4000-R9 290 series. It's why Lisa Su realized this and raised prices. The market voted with their wallets that price/performance is a metric they don't prioritize unless within its own preferred brand choice.

That's not the issue at all. It's the brand and they need to do something to make the brand stand out. When you overhear dumbasses at microcenter or Best Buy give customers a speech as to why Nvidia is better for literally the dumbest of reasons, you have a brand problem. If I was a product manager, I would have a month long trip across the states and have launch parties at every decent sized city and invite these types of people.

I sold a 750 1GB non-Ti for only $50 less than my Sapphire 1Ghz HD7970 edition. The guy who bought it off me builds PCs in Toronto. After he took the card from me, I asked him how is he paying this much $ for a 750? He said he'll sell it for $150 CAD no problem inside a pre-built system. :sneaky:

Look at this:

EVGA GTX780 SC = $250 CAD used
Zotac AMP 780 = $250 CAD
XFX R9 290X = $260 CAD used

Who would buy a 780 over a $260 XFX R9 290X?

980 is going used for $400 CAD but is what 15-20% faster than an after-market 290X at 1440p?

This is clearly a branding issue. The only way I see AMD overcoming this is to offering the fastest card on the market and pricing every single SKU that has more performance than NV for the same price as NV. Then hold on tight for 10 years and keep the flagship performance crown. Over time a new generation of gamers will emerge and will be more likely to buy AMD. A lot of the older generation that grew up on OpenGL or FPS won't buy AMD and they never did even during GeForce 5 or 7 days when NV was producing awful GPUs compared to ATI.

Excellent point.

Best Buy in our area really has less and less custom parts. The Micro Center in Philadelphia (actually Wayne) is @ 1 hour 10 minutes from me BUT has a TON of custom parts. They have more experienced salespersons but that also depends.

In Russia, Central Asia or UAE, I could hardly find AMD Radeon cards and when I could, they were overpriced.

I watched a recent Russian PC hardware YouTuber and he said that in Russia, NV extends credit to Russian retailers. For example, let's say the store wants to buy 1000 GTX1070s and NV sends them for free on credit for 30 days. AMD's AIBs don't offer this option as they ask the retailer to pay the entire cost of GPUs upfront. He then explained that this forces privately owned stores to be more likely to carry NV due to the cost of capital and the upfront capital required to buy so many AMD cards with cash. Now if you have such a retail situation, over a decade of this the consumer who only bought NV for the last 5-10 years is going to be reluctant to switch to AMD. It's like why switch from Colgate to Crest? I was shocked to see that Colgate commanded a 50-100% pricing premium over premium Crest (Blendamed) toothpaste in some countries I visited.

Here is Russian mid-range Oral-B Blend-a-Med branded toothpaste that undercut Colgate by 50%+.

Over here in North America, we know it as the Oral-B's premium Crest Pro-Health/Adanced Pro Health toothpaste. Same product.

In many countries, brand perception plays a far greater role than the actual quality of the product. Latecomers to the market or products that have lower mind-share cannot command the same premium pricing they can enjoy otherwise.

AMD in Russia is like Blend-a-Med, 2nd tier/budget brand and something you buy if you cannot afford NV. Some companies used immense marketing power to turn their image around. For example, Audi in Russia in easily considered on par if not more premium than BMW. In the US, Audi is not considered on the same level as BMW/Mercedes and hence sells about half of what those brands sell. What's the difference between Russians loving Audi and putting it on par or higher than BMW in luxury vs. Americans who consider Audi a re-badged VW (but yet buy Lexus in droves...haha) --- Marketing and branding for Audi was far more successful in Russia than ever in the US.

 

[coercitiv]

nothing amd has done recently has aimed to change that

They are aiming for OEMs and yet you see them doing nothing to change that...

 

[gamervivek]

Even after they incorporated HBM1, they were still behind. That means it's realistic to expect AMD to keep trailing Pascal in perf/watt.

But Pascal's performance comes at high clocks which lead to worse power efficiency improvement compared to AMD's claims. AMD can still be trailing but not as much to make a sizeable difference like Maxwell did.

 

[krumme]

They are aiming for OEMs and yet you see them doing nothing to change that...

And its the only viable strategy. Competing with nv for brand is a pipe dream. They neither have cash competence and culture to compete here. And even if they had its a long term strategy. Building brand is also integrated into product development. They are never ever going there. Period.

Strategy have to reflect eg business, competences, culture and also top management style and strenght. Selling to oem and doing semicustom is perfect for amd. I am sure that the next 5 years will show better results than prior 5.

 

[piesquared]

Well, that's the thing about viral marketing. It infects. Investor biased posts are included in viral marketing as well. I've been saying it for years, the enthusiast boards of yesteryear will disappear or are gone. Moderators refused to do anything about it as forums turned into investor biased shilling. On this forum for example, moderators ignored the shilling as financial related post after financial related post filled the CPU and GPU sections. Completely off topic to the forums, as other non related posts were locked or moved. Maybe 50 out of 100 posts are related to the technology, while the other 50 are driven by viral marketing or financially motivated.
Some suggest that NV's marketing is great, and that AMD's is poor. That is an example of viral marketing in itself, since about the only marketing NV does is viral marketing. It's not hard to sell a product if you swarm message boards to convince the easily convincible of your message. I put the blame squarely on the websites and their forums for allowing it to happen. The question is why? I'd be surprised if any objective reader couldn't see it plain as day.


Mod callouts are not allowed
If you don't like something, use the MD forum. Also, There is no magic flag that says someone is a shill, and to accuse them of that without proof is not allowed.
Markfw900

 

[tential]

People aren't objective based and rarely make rational decisions. We're emotional far more than we realize and make a lot of emotional decision

 

[maddie]

Agreed. NV was vastly ahead of AMD in perf/watt with Maxwell vs. Tahiti/Hawaii GDDR5 parts. Even after they incorporated HBM1, they were still behind. That means it's realistic to expect AMD to keep trailing Pascal in perf/watt. Couple that with a smaller die size of Polaris 10 and no way it should be even close to the 1070. If AMD's 232mm2 Polaris 10 is as fast as NV's 314mm2 1070, that means for next gen vs. next gen cards AMD would have pulled off a 35% increase in performance per mm2 over Pascal. No chance I'd believe that.

The 1080 is 9.0 TFlops and the 1070 is 6.5 TFlops, suggesting that the 1070 is equal to 72% of the full die. Roughly a 227mm^2 GP104 full equivalent.

This would suggest that a full Polaris 10 should be equivalent in perf/mm^2 to GP104 if it is equal in performance to the 1070.

Your calculation would be right if comparing to the 1080 variant, not the 1070.

 

[tential]

The 1080 is 9.0 TFlops and the 1070 is 6.5 TFlops, suggesting that the 1070 is equal to 72% of the full die. Roughly a 227mm^2 GP104 full equivalent.

This would suggest that a full Polaris 10 should be equivalent in perf/mm^2 to GP104 if it is equal in performance to the 1070.

Your calculation would be right if comparing to the 1080 variant, not the 1070.

What indication do you have that full P10 is equivalent to the 1070 though? All reports show it at a significant disadvantage.

 

[raghu78]

The 1080 is 9.0 TFlops and the 1070 is 6.5 TFlops, suggesting that the 1070 is equal to 72% of the full die. Roughly a 227mm^2 GP104 full equivalent.

This would suggest that a full Polaris 10 should be equivalent in perf/mm^2 to GP104 if it is equal in performance to the 1070.

It does not work that way. You cannot say GTX 1070 is 72% of the full die. GTX 1070 is likely to be clocked lower than GTX 1080. GTX 1070 is most likely going to be 2048 cuda cores at 1.6 Ghz boost clock (actual config yet to be seen). We don't know if GTX 1070 has any ROPs disabled (like GTX 970) though I think Nvidia would avoid a PR nightmare like the GTX 970. TMUs disabled will be proportionate to cuda cores. Anyway all this comparison has to wait for actual reviews of performance.

 

[maddie]

What indication do you have that full P10 is equivalent to the 1070 though? All reports show it at a significant disadvantage.

I was replying to this by RS. He started his statement by saying IF P10 = 1070. I was merely replying that the calc was wrong.

"If AMD's 232mm2 Polaris 10 is as fast as NV's 314mm2 1070, that means for next gen vs. next gen cards AMD would have pulled off a 35% increase in performance per mm2 over Pascal. No chance I'd believe that."

 

[tential]

I was replying to this by RS. He started his statement by saying IF P10 = 1070. I was merely replying that the calc was wrong.

"If AMD's 232mm2 Polaris 10 is as fast as NV's 314mm2 1070, that means for next gen vs. next gen cards AMD would have pulled off a 35% increase in performance per mm2 over Pascal. No chance I'd believe that."

Oops. Lol ya k that makes more sense

 

[el etro]

Node sizes needs to be normalized, being Samsung 14LPE having a 9-11% density advantage over TSMC 2nd gen FF process(16FF+), where both Pascal and Apple A9 are manufactured.

A 14LPP 232mm2 chip would equal a 16FF+ >250mm2 chip.

 

[RussianSensation]

What indication do you have that full P10 is equivalent to the 1070 though? All reports show it at a significant disadvantage.

Working backwards:

290X = 294W / 2.5 perf/watt = 118W for Polaris 10 ~ 290X (Best case scenario!)

1070 supposedly beats Titan X. Let's just say it = Titan X = 288% (1440p)

Assigned 1440p performance rating:

290X = 207% (118W equivalent Polaris 10)
1070 Projected minimum Titan X level = 288% (150W TDP for 1070)

1070 is ahead of our projected Polaris 10 by 288% / 207% = 39%

Polaris 10 would need a 40% increase in GPU clocks / performance to catch up.

Polaris 10 = 118W x 1.39% => New 164W TDP

Major caveats:

1. Assumes increasing performance with Polaris 10 scales 100% linearly with TDP. We know this isn't true for almost any GPU architecture. Firstly, we know that almost no GPU scales 100% linearly with increased clock speeds. Secondly, we know that to increase GPU clocks 40% without increasing voltage or running into exponential increase in power usage is unrealistic.

2. Assumes AMD's 2.5X claim over 290X is 100% correct in almost all cases, not just best case scenarios. If not, the 118W Polaris 10 ~ 290X would mean the power usage of Polaris needs to rise above 118W to match a 290X. For example, if AMD's real world result is 2X perf/watt over 290X, Polaris 10 would be now using 147W to match a 290X. This just shows how much variation there is in AMD's claim of Polaris 10 delivering "up to 2.5X perf/watt" or "2x perf/watt" etc.

That's before we even got to Polaris 10 vs. 1070's overclocking headroom. We are already seeing 2.1Ghz on air from GP104, with rumors of 2.3-2.4Ghz on air and 2.5Ghz with an AIO CLC! AMD's cards in the last 2 generations weren't great overclockers. This once again should give 1070 a big edge.

Since NV claims 1070 > Titan X, and using AMD's own claims of 2.5X perf/watt are most likely too optimistic, Polaris 10 has no chance of beating 1070 in perf/watt using math and neither will it match 1070 FE, nevermind 1070 AIB. More likely than not, NV will also have lower GPU Boost on 1070 than 1080 to create an artificial gap that justifies the 1080's price premium. This means that 1070 should have even bigger overclocking headroom in % terms than 1080 does since they are the same GP104 silicon.

Conclusion:

Using the process of deduction, Polaris 10 will not match 1070 on average even at 150W TDP. It will need 165-175W TDP at least but then even if some AIBs make Polaris 10 cards with max 175W TDP headroom, so will NV's AIB partners who will add 15-20% higher GPU Boost clocks on 1070.

I was replying to this by RS. He started his statement by saying IF P10 = 1070. I was merely replying that the calc was wrong.

"If AMD's 232mm2 Polaris 10 is as fast as NV's 314mm2 1070, that means for next gen vs. next gen cards AMD would have pulled off a 35% increase in performance per mm2 over Pascal. No chance I'd believe that."

How is my calculation wrong?

If 1070 and Polaris 10 are equal in performance, it means a 314mm2 chip (35% larger) is only as fast as a 232mm2 chip (Polaris 10). This means AMD would have pulled off a 35% increase in performance /mm2 over Polaris 10. I don't believe this is realistic.

> If we use die sizes, Polaris 10 loses
> If we use 2.5X perf/watt claims from AMD, Polaris 10 still loses

There is no realistic math that can get us to the point where a Polaris 10 chip is as fast as a 1070. In all my calculations, I also gave AMD the best possible chance and I low-balled 1070's performance by equating it to Titan X.

I just want to remind you once again that before 1070/1080 came out and we heard that Vega is a late 2016/early 2017 launch, no one really contested that Polaris 10 ~ 390X or Fury at 110-130W. Now that we saw how good 1070/1080 are, and people realized AMD will cede the entire $350+ market segment for 6 months+, suddenly Polaris 10's performance keeps going up and up and up. It's becoming a joke to try and fill in performance gap that AMD has when Polaris 10 is Pitcairn HD7850/7870 successor (clearly), while GP104 with > 300mm2 die size and GDDR5X are straight up GTX670/680 successors.

No one would have expected 7850/7870 to be competing with 670/680 during the start of that generation. So why now are people expecting Polaris 10 -> Pitcairn lineage to suddenly start competing with 1070/1080? It's wishful thinking because AMD is way behind with Vega 11 and 10.

AMD's own comments = Polaris 10 = mainstream + high-end laptops (ONLY)
AMD's own slides show Polaris 10 as a mid-range chip at best
AMD's own comments paraphrased = "We aren't concerned about NV's Maxwell since they are targeting high end"

Just face it, AMD is 6-8 months behind GP104 and they got caught off-guard about how much faster 1080 is + its overclocking headroom vs. what they expected. They also didn't expect 1070 to cost $379, likely expecting it to be $450 minimum.

Even their invite to Computex focuses on APUs, not Polaris 10.

"During this event, AMD executives and special guests will introduce new, comprehensive details on many of the 2016 products, including the official launch of the 7th Generation AMD A-Series APU."

This is a big red flag. Sounds like they may reveal details of Polaris 10/11 cards but still no hard launch in June.
http://www.amdcomputex.com.tw/

Node sizes needs to be normalized, being Samsung 14LPE having a 9-11% density advantage over TSMC 2nd gen FF process(16FF+), where both Pascal and Apple A9 are manufactured.

A 14LPP 232mm2 chip would equal a 16FF+ >250mm2 chip.

Not conclusive. Rumour had it that TSMC didn't have much time to optimize A9 for 16FF. Besides, you cannot have it both ways -- if you argue that AMD's chip is more dense, then you need to assign a GPU clock penalty for Polaris 10 over GP104. This is something we've seen in the last 4 years. AMD's denser designs have lower GPU clocks and overclock less than NV's.

AMD's best hope lies in them targetting as many AAA developers as they can and locking in DX12 games/game bundling with Polaris 10.

For example, offer Polaris 10 at $299 but throw in Total War Warhammer with it.

 

[maddie]

Working backwards:

290X = 294W / 2.5 perf/watt = 118W for Polaris 10 ~ 290X (Best case scenario!)

1070 supposedly beats Titan X. Let's just say it = Titan X = 288% (1440p)

Assigned 1440p performance rating:

290X = 207% (118W equivalent Polaris 10)
1070 Projected minimum Titan X level = 288% (150W TDP for 1070)

1070 is ahead of our projected Polaris 10 by 288% / 207% = 39%

Polaris 10 would need a 40% increase in GPU clocks / performance to catch up.

Polaris 10 = 118W x 1.39% => New 164W TDP

Major caveats:

1. Assumes increasing performance with Polaris 10 scales 100% linearly with TDP. We know this isn't true for almost any GPU architecture. Firstly, we know that almost no GPU scales 100% linearly with increased clock speeds. Secondly, we know that to increase GPU clocks 40% without increasing voltage or running into exponential increase in power usage is unrealistic.

2. Assumes AMD's 2.5X claim over 290X is 100% correct in almost all cases, not just best case scenarios. If not, the 118W Polaris 10 ~ 290X would mean the power usage of Polaris needs to rise above 118W to match a 290X. For example, if AMD's real world result is 2X perf/watt over 290X, Polaris 10 would be now using 147W to match a 290X. This just shows how much variation there is in AMD's claim of Polaris 10 delivering "up to 2.5X perf/watt" or "2x perf/watt" etc.

3. Assumes Polaris 10 will arrive as a fully unlocked chip.

That's before we even got to Polaris 10 vs. 1070's overclocking headroom.

Since NV claims 1070 > Titan X, and using AMD's own claims of 2.5X perf/watt, Polaris 10 has no chance of beating 1070 in perf/watt using math.

Conclusion:

Using the process of deducation, Polaris 10 will not match 1070 on average even at 150W TDP. It will need 165-175W TDP at least.

I notice you took the peak gaming power as the value for 290X. Do you think this is equivalent to the 1070 150TDP value? I bolded the relative sentence.

From TPU charts:

• Average: Metro: Last Light at 1920x1080 because it is representative of a typical gaming power draw. The average of all readings (12 per second) while the benchmark was rendering (no title/loading screen) is used. In order to heat up the card, the benchmark is run once without measuring power consumption.
• Peak: Metro: We use Last Light at 1920x1080 as it produces power draw typical to gaming. The highest single reading during the test is used.

If the average gaming power of 258W is used, then P10 best case in your calcs become 103W and you end up with this value instead.

Polaris 10 = 103W x 1.39% => New 143W TDP

edit:
In that same TPU chart we see the peak gaming power of the 970 as 191W, and we all know this is way above the official Nvidia value, leading me to believe that the 150W for Nvidia 1070 is the average gaming value.

 

[el etro]

Not conclusive. Rumour had it that TSMC didn't have much time to optimize A9 for 16FF. Besides, you cannot have it both ways -- if you argue that AMD's chip is more dense, then you need to assign a GPU clock penalty for Polaris 10 over GP104. This is something we've seen in the last 4 years. AMD's denser designs have lower GPU clocks and overclock less than NV's.

AMD's best hope lies in them targetting as many AAA developers as they can and locking in DX12 games/game bundling with Polaris 10.

For example, offer Polaris 10 at $299 but throw in Total War Warhammer with it.

Size assumptions are based from values of CPP amd M1 Pitch given in the semiwiki article, hardly they are wrong. If TSMC had density parity with Samsung they would call the process 14nm too, no?

 

[selni]

Nope. I remember R9 280X vs. 770 generation and every time someone was building a mid-range rig, it was impossible to recommend the 770. 770 2GB first dropped from $400 to $380 and it took 8-10 months before it went down to $330-340 range.

May 30, 2013 = $420
February-March 2014 = $340
April-June 2014 = $330

http://camelcamelcamel.com/EVGA-GeF...-P4-2774-KR/product/B00CZIQXBA?context=browse

http://camelcamelcamel.com/GTX770-D...OC-selected/product/B00D3F7CFK?context=browse

Even on the best deals, it took 770 at least 6 months before it dropped to $330
http://camelcamelcamel.com/EVGA-GeF...KR-Graphics/product/B00CZIQZ40?context=browse

The comparison to 770 2GB is also strange considering we all knew that 2GB cards are DOA for anyone cross-shopping 280X vs. 770 around that time. That means the proper comparison in hindsight for those who kept defending 2GB cards was 280X vs. 770 4GB.

For a card that cost more throughout its entire life-cycle, 770 was VRAM gimped or too expensive and aged far worse.
https://www.techpowerup.com/reviews/ASUS/GTX_950/23.html

Do you also remember the R280X was released in October 2013 and the 770 price cut came that same month? Why are you comparing pricing in May 2013? Yes it's a 7970 and that'd been out for a while, but it was a much cheaper 7970.

The 4GB 770 was a very bad purchase after that drop too as the 4GB version wasn't discounted - $120 extra for VRAM on a card that as you've pointed out aged badly anyway.

 

[Abwx]

Now that we saw how good 1070/1080 are,

Personaly i did see nothing, or have you a link of a review, with perfs and power comsumption..?..

 

[el etro]

AMD hardware is already awesome. To take the lead from Nvidia they need better power management/IC design(thing they are shooting with Artic Islands),much better software stack, improving game driver support(includes crossfire performance) and worldwide agressivity on shopping deals, they need to sell better(higher margins and better sales) worldwide, not just in US, having better price/performance ratio on every place.

 

[monstercameron]

AMD hardware is already awesome. To take the lead from Nvidia they need better power management/IC design(thing they are shooting with Artic Islands),much better software stack, improving game driver support(includes crossfire performance) and worldwide agressivity on shopping deals, they need to sell better(higher margins and better sales) worldwide, not just in US, having better price/performance ratio on every place.

-Power management isn't an issue for amd. Zerocore and associated tech handles that just fine.
-Dx12 should render they backend tweaking less efficacious going forward in regards to drivers and 3d engine, front loading the responsibility on devs.
-AMDs 200 and 300 are technically the same yet the 300 series raised msrp. Strats are in effect.

 

[raghu78]

Working backwards:

290X = 294W / 2.5 perf/watt = 118W for Polaris 10 ~ 290X (Best case scenario!)

1070 supposedly beats Titan X. Let's just say it = Titan X = 288% (1440p)

Assigned 1440p performance rating:

290X = 207% (118W equivalent Polaris 10)
1070 Projected minimum Titan X level = 288% (150W TDP for 1070)

1070 is ahead of our projected Polaris 10 by 288% / 207% = 39%

Polaris 10 would need a 40% increase in GPU clocks / performance to catch up.

Polaris 10 = 118W x 1.39% => New 164W TDP

Major caveats:

1. Assumes increasing performance with Polaris 10 scales 100% linearly with TDP. We know this isn't true for almost any GPU architecture. Firstly, we know that almost no GPU scales 100% linearly with increased clock speeds. Secondly, we know that to increase GPU clocks 40% without increasing voltage or running into exponential increase in power usage is unrealistic.

2. Assumes AMD's 2.5X claim over 290X is 100% correct in almost all cases, not just best case scenarios. If not, the 118W Polaris 10 ~ 290X would mean the power usage of Polaris needs to rise above 118W to match a 290X. For example, if AMD's real world result is 2X perf/watt over 290X, Polaris 10 would be now using 147W to match a 290X. This just shows how much variation there is in AMD's claim of Polaris 10 delivering "up to 2.5X perf/watt" or "2x perf/watt" etc.

That's before we even got to Polaris 10 vs. 1070's overclocking headroom. We are already seeing 2.1Ghz on air from GP104, with rumors of 2.3-2.4Ghz on air and 2.5Ghz with an AIO CLC! AMD's cards in the last 2 generations weren't great overclockers. This once again should give 1070 a big edge.

Since NV claims 1070 > Titan X, and using AMD's own claims of 2.5X perf/watt are most likely too optimistic, Polaris 10 has no chance of beating 1070 in perf/watt using math and neither will it match 1070 FE, nevermind 1070 AIB. More likely than not, NV will also have lower GPU Boost on 1070 than 1080 to create an artificial gap that justifies the 1080's price premium. This means that 1070 should have even bigger overclocking headroom in % terms than 1080 does since they are the same GP104 silicon.

Conclusion:

Using the process of deduction, Polaris 10 will not match 1070 on average even at 150W TDP. It will need 165-175W TDP at least but then even if some AIBs make Polaris 10 cards with max 175W TDP headroom, so will NV's AIB partners who will add 15-20% higher GPU Boost clocks on 1070.



How is my calculation wrong?

If 1070 and Polaris 10 are equal in performance, it means a 314mm2 chip (35% larger) is only as fast as a 232mm2 chip (Polaris 10). This means AMD would have pulled off a 35% increase in performance /mm2 over Polaris 10. I don't believe this is realistic.

> If we use die sizes, Polaris 10 loses
> If we use 2.5X perf/watt claims from AMD, Polaris 10 still loses

There is no realistic math that can get us to the point where a Polaris 10 chip is as fast as a 1070. In all my calculations, I also gave AMD the best possible chance and I low-balled 1070's performance by equating it to Titan X.

I just want to remind you once again that before 1070/1080 came out and we heard that Vega is a late 2016/early 2017 launch, no one really contested that Polaris 10 ~ 390X or Fury at 110-130W. Now that we saw how good 1070/1080 are, and people realized AMD will cede the entire $350+ market segment for 6 months+, suddenly Polaris 10's performance keeps going up and up and up. It's becoming a joke to try and fill in performance gap that AMD has when Polaris 10 is Pitcairn HD7850/7870 successor (clearly), while GP104 with > 300mm2 die size and GDDR5X are straight up GTX670/680 successors.

No one would have expected 7850/7870 to be competing with 670/680 during the start of that generation. So why now are people expecting Polaris 10 -> Pitcairn lineage to suddenly start competing with 1070/1080? It's wishful thinking because AMD is way behind with Vega 11 and 10.

AMD's own comments = Polaris 10 = mainstream + high-end laptops (ONLY)
AMD's own slides show Polaris 10 as a mid-range chip at best
AMD's own comments paraphrased = "We aren't concerned about NV's Maxwell since they are targeting high end"

Just face it, AMD is 6-8 months behind GP104 and they got caught off-guard about how much faster 1080 is + its overclocking headroom vs. what they expected. They also didn't expect 1070 to cost $379, likely expecting it to be $450 minimum.

Even their invite to Computex focuses on APUs, not Polaris 10.

"During this event, AMD executives and special guests will introduce new, comprehensive details on many of the 2016 products, including the official launch of the 7th Generation AMD A-Series APU."

This is a big red flag. Sounds like they may reveal details of Polaris 10/11 cards but still no hard launch in June.
http://www.amdcomputex.com.tw/

Not conclusive. Rumour had it that TSMC didn't have much time to optimize A9 for 16FF. Besides, you cannot have it both ways -- if you argue that AMD's chip is more dense, then you need to assign a GPU clock penalty for Polaris 10 over GP104. This is something we've seen in the last 4 years. AMD's denser designs have lower GPU clocks and overclock less than NV's.

AMD's best hope lies in them targetting as many AAA developers as they can and locking in DX12 games/game bundling with Polaris 10.

For example, offer Polaris 10 at $299 but throw in Total War Warhammer with it.

RS as I have said earlier you believe AMD's perf/watt gains are against older GPUs while I can argue that AMD's comparisons are against their more recent GPUs like Tonga/Fiji. If you take AMD's Capsaicin perf/watt slide it shows 28nm GPUs in late 2014/2015 timeframe. For eg: If you apply the 2.5x perf/watt gains to these 2 GPUs you end up with a R9 380X perf at 76w. (190/2.5 = 76). Incidentally HD 7770 had a 80W TDP. So it would not surprise me to see a Polaris 11 at 80w which matches or beats R9 380X. Doubling the GPU from Polaris 11 to Polaris 10 (like HD 7770 vs HD 7870) will get you another 80-85% perf which puts Polaris 10 on par with GTX 980 Ti or slightly ahead. If AMD has a flagship GDDR5X based SKU coming which I would say is very likely then there is a good chance that the perf could get even better.

Now the other thing is 14LPP is atleast 10% denser than TSMC 16FF+(which is 2x the density of TSMC 28nm). 14LPP is a better process than 14LPE and has improvements to transistor performance, more mature process as it incorporates learning from 14LPE which was a early time to market process. This should allow denser layout than 14LPE. You have to look at Pitcairn vs Hawaii to see the increase in transistor density (2800/212 =13.2 million transistors per sq mm vs 6200/438 = 14.15 million transistors vper sq mm ) If you take a 232 sq mm 14LPP process and use a 2.2x transistor density that puts it at 232 *2.2 = 510 sq mm on TSMC 28nm. Now If you extrapolate based on Pitcairn density you would get around 6.75 billion transistors for Polaris 10 and based on Hawaii density you would get 7.2 billion transistors. I am saying an average of these is more likely as Hawaii came 18 months after Pitcairn while 14LPP is arriving 12 months after 14LPE. imo a 7 billion transistor Polaris 10 chip is likely with massive architectural improvements and thats why I think Polaris is going to be very competitive.

 

[el etro]

Cameron, whem i say power management i say few about power gating and other firmware power saving techniques, and more about improving the IC design, allowing cards to overclock better without much overvolt and decontrol of power consumption, making GPUs easier to set bins and increasing both gold cards(high Asic golden cards and high overclocking Golden cards) rate.

By worldwide sales agressivity i know what i wanted to say. Outside of US price/performance ratio of AMD cards is pretty subpar and it should not be, the MSRP/perf set by AMD is always better but this don't reflect im the price in the stores! Is very hard to recomend any AMD graphics card here in Brazil, as a example. Nvidia cards have always better price/performance here!

 

[kraatus77]

RS as I have said earlier you believe AMD's perf/watt gains are against older GPUs while I can argue that AMD's comparisons are against their more recent GPUs like Tonga/Fiji. If you take AMD's Capsaicin perf/watt slide it shows 28nm GPUs in late 2014/2015 timeframe. For eg: If you apply the 2.5x perf/watt gains to these 2 GPUs you end up with a R9 380X perf at 76w. (190/2.5 = 76). Incidentally HD 7770 had a 80W TDP. So it would not surprise me to see a Polaris 11 at 80w which matches or beats R9 380X. Doubling the GPU from Polaris 11 to Polaris 10 (like HD 7770 vs HD 7870) will get you another 80-85% perf which puts Polaris 10 on par with GTX 980 Ti or slightly ahead. If AMD has a flagship GDDR5X based SKU coming which I would say is very likely then there is a good chance that the perf could get even better.

Now the other thing is 14LPP is atleast 10% denser than TSMC 16FF+(which is 2x the density of TSMC 28nm). 14LPP is a better process than 14LPE and has improvements to transistor performance, more mature process as it incorporates learning from 14LPE which was a early time to market process. This should allow denser layout than 14LPE. You have to look at Pitcairn vs Hawaii to see the increase in transistor density (2800/212 =13.2 million transistors per sq mm vs 6200/438 = 14.15 million transistors vper sq mm ) If you take a 232 sq mm 14LPP process and use a 2.2x transistor density that puts it at 232 *2.2 = 510 sq mm on TSMC 28nm. Now If you extrapolate based on Pitcairn density you would get around 6.75 billion transistors for Polaris 10 and based on Hawaii density you would get 7.2 billion transistors. I am saying an average of these is more likely as Hawaii came 18 months after Pitcairn while 14LPP is arriving 12 months after 14LPE. imo a 7 billion transistor Polaris 10 chip is likely with massive architectural improvements and thats why I think Polaris is going to be very competitive.

This type of thinking is why AMD can't deliver and fulfill expectations even though this time they said it pretty clear about no availability of mid-high end stuff. you people just expect too much.

 

[Arachnotronic]

This type of thinking is why AMD can't deliver and fulfill expectations even though this time they said it pretty clear about no availability of mid-high end stuff. you people just expect too much.

agreed. It's not even that Fury X was a bad product, it's just that people hyped it up to be something in their minds that it couldn't possibly deliver on.