[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3

[Glo.]

Its not going to be 249$. Put that out of your heads, guys. We are looking at 270-280$ GPU.

And the GPU in question will be right in between Vega56 and GTX 1070 Ti.

 

[jpiniero]

Steam survey has nothing to do with statistics and it's not relevant to any discussion.

If the survey was done correctly, it would be reasonably accurate for actual PC gaming usage when you factor in that Steam is so dominant. You don't need 100% surveyed, that's how statistical sampling is done. Unit sales includes things like mining as I mentioned which helps the bottom line but doesn't reflect what gamers are actually using.

Now I don't know if Valve or developers are actually taking action based upon what the survey results return but it wouldn't surprise me.

 

[Veradun]

If the survey was done correctly, it would be reasonably accurate for actual PC gaming usage when you factor in that Steam is so dominant. You don't need 100% surveyed, that's how statistical sampling is done. Unit sales includes things like mining as I mentioned which helps the bottom line but doesn't reflect what gamers are actually using.

Now I don't know if Valve or developers are actually taking action based upon what the survey results return but it wouldn't surprise me.

I know how statistics are done. Steam doesn't statistically sample, it's just random. Infact it doesn't even remotely reflect what JPR say (that is not an esteem, but an exact value), proving how wrong is steam hw survey at representing the world.

 

[GodisanAtheist]

I read the 505mm^2 die size part but wasn't sure if they were talking about Navi or a Zen 2 APU.

I'm assuming that the Navi 20 series includes Ray Tracing hardware. Double the hardware (CU's, ROP's, etc) and adding RT hardware but balancing that with a 20% density improvement, I believe that die size. I wonder how much they've improved the CU's and stuff over the Navi 10 series, or if the only difference will be in the addition of RT?

I hope they get the N7+ process pretty reliable. At 505mm^2 per die it wont take too many imperfections to ruin an entire wafer's worth.

- I know rumors are rumors but there is no universe where AMD releases a Zen2 APU at 505mm2. It would be too low production/margin part to go into a console and if that's not happening then PC is definitely out. By comparison, AMD's most current "Picasso" APU (in the 3400G) is ~200mm2 on 12nm.

The 2080TI is monstrous at ~750mm2 on 12nm. Assuming (and its one hell of an assumption) that NV follows AMD's density scaling of roughly 1.45x from 14/12 nm down to 7nm, the 2080TI would clock in at ~515mm2 on the current 7nm process.

505mm2 on the N7+ process would really be a performance juggernaut, but never count against AMD botching the job... they're trying to work miracles on a shoestring budget (compared to the competition).

 

[Dribble]

I know how statistics are done. Steam doesn't statistically sample, it's just random. Infact it doesn't even remotely reflect what JPR say (that is not an esteem, but an exact value), proving how wrong is steam hw survey at representing the world.

Random is a perfectly valid way of sampling. I suspect the real problem with steam is it doesn't say what you want to hear.

 

[lifeblood]

I know how statistics are done. Steam doesn't statistically sample, it's just random. Infact it doesn't even remotely reflect what JPR say (that is not an esteem, but an exact value), proving how wrong is steam hw survey at representing the world.

Steam survey sucks but it's all we really have. Until someone provides something better then it at least provides some semblance of being a valid source. It provides a glimpse at what gamer's who use steam are using. You cannot say product X is selling better than product Y when the steam survey says otherwise and you have no better source.

 

[Mopetar]

Random is a perfectly valid way of sampling. I suspect the real problem with steam is it doesn't say what you want to hear.

Random sampling is preferred assuming you have no way of getting a representative sample, which in the case of something like Steam you don’t. However, Steam isn’t actually random since it asks if you want to participate which is a potential source of bias. Does that mean it’s actually biased? No, but you’d want to account for it in some way and we have no real way of doing that. At best you can match it to other sales reports that can have their own error estimated and use that to get a better idea.

 

[Veradun]

Random is a perfectly valid way of sampling. I suspect the real problem with steam is it doesn't say what you want to hear.

I suspect you have no clue on how statistics are done.

Simply put, statistics are models of reality. When reality (total shipments) is this different from statistics (steam hw survey) then said statistics are just wrong.

Your problem might be this borked statistics say what you want to hear so you ignore hard facts to elevate that shit to truth.

 

[Bouowmx]

505 mm^2: Using Navi 10's density of 41 M/mm^2, that area gets 20.7 B transistors. Shouldn't have an issue against NVIDIA TU102 (18.6 B).

 

[Shmee]

There are many platforms of PC gaming other than steam. Anyways this isn't really relevant to the discussion...

 

[happy medium]

There is a word for this. Starts with "B", has "S" in the middle and ends with "T".

RX 5700 and RX 5700 XT are the reason why in two quarters AMD went from 18% marketshare to over 30%. If anything, RX 5700 and 5700 XT are selling on the same level as RTX 2070 and 2070S.

Wheres your link?
Nvidia sells more GPU's in every performance tier compared to AMD.
You dont need to be rocket scientist to know this.
And before you say it,its not my job to prove your statement.

 

[Glo.]

Wheres your link?
Nvidia sells more GPU's in every performance tier compared to AMD.
You dont need to be rocket scientist to know this.
And before you say it,its not my job to prove your statement.

None of you EVER posted any links proving that Nvidia RTX 2060S and 2070S sold more GPUs than RX 5700 and 5700XT.

It is ONLY everyones assumption over this. So suddenly assumption, despite the fact that two AMD GPUs made vast spike in GPU marketshare for AMD when they were released, is taken as a fact that Nvidia outsold AMD mainstream GPUs. Ok.

Where are your links proving that RTX 2060S and 2070S sold more GPUs than RX 5700 and 5700 XT, eh?

 

[Head1985]

Its not going to be 249$. Put that out of your heads, guys. We are looking at 270-280$ GPU.

And the GPU in question will be right in between Vega56 and GTX 1070 Ti.

More like 300usd for better aib.

 

[Glo.]

More like 300usd for better aib.

Well, PowerColor's GPUs are usually 10$ over MSRP, so if MSRP is 270-280 PowerColor's models may be 280-290$ for Red Dragon, for example.

 

[AtenRa]

AMD increased its marketshare YoY in Q3 2019 vs Q3 2018 according to JJP.
Also, AIB shipments increased 6.2% YoY. This means that AMD not only took marketshare from NVIDIA but also increased its overall shipments substantially YoY in the third Quarter.

Add-in Board report – a report on the Graphics Add-in Board market

Add-in Board report – a report on the Graphic Add-in Board market The Add-in Board report is a quarterly report that focuses on the market activity of PC graphics controllers for desktop, workstation, and data center display and computing. The report provides an in-depth look at the PC graphics...

www.jonpeddie.com

Q3 2018 AIB Marketshare

AMD = 25.72%
NVIDIA = 74.28%


Q3 2019 AIB Marketshare

AMD = 27.08%
NVIDIA = 72.92%

 

[insertcarehere]

There are many platforms of PC gaming other than steam. Anyways this isn't really relevant to the discussion...

Even discounting that people who mainly use Origin/Uplay/Epic stuff are also very likely to have Steam installed anyways. In order to discount Steam surveys as information, there needs to be good evidence that GPU sales of PC Gamers who use steam (and thus included in survey) are not representative of PC gamers (the sort which buy gpus) as a whole .

 

[DeathReborn]

AMD increased its marketshare YoY in Q3 2019 vs Q3 2018 according to JJP.
Also, AIB shipments increased 6.2% YoY. This means that AMD not only took marketshare from NVIDIA but also increased its overall shipments substantially YoY in the third Quarter.

Add-in Board report – a report on the Graphics Add-in Board market

Add-in Board report – a report on the Graphic Add-in Board market The Add-in Board report is a quarterly report that focuses on the market activity of PC graphics controllers for desktop, workstation, and data center display and computing. The report provides an in-depth look at the PC graphics...

www.jonpeddie.com

Q3 2018 AIB Marketshare

AMD = 25.72%
NVIDIA = 74.28%


Q3 2019 AIB Marketshare

AMD = 27.08%
NVIDIA = 72.92%

While that is useful general data it doesn't tell what cards sell, it could be just the seemingly never ending fire sale on Polaris that is boosting sales. The Q-Q (42.2% boost Q2>Q3) results aren't so rosy for AMD so while compared to last year they are up, compared to the last 3 months not so much.

AMD Q2 19 = 32.08%
NVIDIA Q2 19 = 67.92%

Fact is that there is no reliable and detailed method of determining what card outsold which, just a general trend which can be based on anything (except 2080 Ti outselling any other card ).

 

[mohit9206]

More like 300usd for better aib.

Well, PowerColor's GPUs are usually 10$ over MSRP, so if MSRP is 270-280 PowerColor's models may be 280-290$ for Red Dragon, for example.

My body is so ready to crap on AMD once they announce the prices. 300 USD for 6GB graphics card in 2020..i hope they know what they're doing.

 

[Glo.]

My body is so ready to crap on AMD once they announce the prices. 300 USD for 6GB graphics card in 2020..i hope they know what they're doing.

Uhhh, what about RTX 2060, then? It still exists.

 

[uzzi38]

Uhhh, what about RTX 2060, then? It still exists.

And it's $349 too

 

[mohit9206]

Yeah screw that card. Whoever bought that card would also be screwed within a couple of years. It can't do rtx because its too weak, and only 6gb. Nobody should buy that card when 5700 is available for less.

 

[Glo.]

And it's $349 too

Which was exactly my point .

Yeah screw that card. Whoever bought that card would also be screwed within a couple of years. It can't do rtx because its too weak, and only 6gb. Nobody should buy that card when 5700 is available for less.

It was released months before RX 5700. So it was best thing you could buy at that price range.

 

[tajoh111]

While that is useful general data it doesn't tell what cards sell, it could be just the seemingly never ending fire sale on Polaris that is boosting sales. The Q-Q (42.2% boost Q2>Q3) results aren't so rosy for AMD so while compared to last year they are up, compared to the last 3 months not so much.

AMD Q2 19 = 32.08%
NVIDIA Q2 19 = 67.92%

Fact is that there is no reliable and detailed method of determining what card outsold which, just a general trend which can be based on anything (except 2080 Ti outselling any other card ).

Yes, the Q2 to Q3 numbers show a better pictures of what is outselling what. Not the Q3 over Q3 yearly numbers. Using q3/q3 year on year makes little sense. Particularly if we are going to demonstrate Turing vs Navi sales since neither card was available in q3 2018 and thus has no relevance. This as a result is a red herring. AMD particularly low marketshare in 2018 q3 were the result of mining glut and and an oversupply of polaris cards. Add in the used market and people were just not interested in buying polaris. Nvidia had this problem but to a lesser extent. Nvidia was insulated a bit due to being the more popular design in laptop and system integrators. Both suffered but more so AMD which resulted in AMD having to do more severe pricecuts to sell their cards. Overall for a Company, AMD numbers didn't fall as much because of the CPU division which were not impacted by the mining collapse.

Q3 2019 numbers are the quarter where both cards are available and are thus the only relevant numbers we should look at(q4 eventually too). As was shown by the numbers, Nvidia gained marketshare. Not only that, Nvidia financial number got way better in q3 vs q2 beating expectations. Particularly the gaming revenue segment. Nvidia beat analysts predictions by 120 million dollars. AMD on the other hand didn't quite meet revenue expectations on their earnings.

https://www.investors.com/news/technology/amd-earnings-q3-2019-amd-stock/

Considering Q3 was the quarter, Ryzen 2 and Navi came out, this should have been a quarter where AMD crushed expectations. But they didn't. Look at AMD CPU and Discrete's revenue 1.276 billion in q3. Very good no doubt compared to earlier quarters but significantly less than the 1.66 billion from Nvidia gaming division. Analyst were expecting Nvidia to sell 1.54 billion in their gaming division. So why did analyst get it wrong. The answer is Navi did not provide the competition expected.

120 million added to AMD GPU discrete division would likely boost their GPU revenue numbers by atleast 33% if we assume most of the CPU and GPU division revenue is coming from CPU which is the right assumption, the strength of Ryzen/Ryzen 2 vs Intel, and the total addressable market of the CPU market(10 billion plus quarterly)vs GPU(2 billion Quarterly). So what is happening? This should have been in AMD's pocket not Nvidia's if AMD was taking marketshare from Nvidia.

What is happening is the Nvidia's super cards are not only outselling Navi, they are doing so significantly. Nvidia's q2 vs q3 growth nearly match AMD's but they are doing this purely with Discrete gaming GPU's. AMD has Ryzen 2 in the equation which is likely the source of most of this growth.

Still don't believe me? Look at how aggressive the pricing of Navi GPU are now and all the promo's being thrown around. When Navi launched, it had no game bundle was priced at 350/400 for the 5700/5700xt respectively. Now we see the 5700 series being priced at $290 regularly and a game bundle. Same with the 5700xt. AIB 5700xt can be found in the 350-380 range along with the game bundle. Compare this the super series where there discounts do not take them below MSRP and the cards don't even come with a game and you can tell AMD is having a bit of problems moving Navi Stock while Nvidia is having an easier time selling super cards.

One last thing to add is Nvidia numbers grew in terms of marketshare between q2 and q3 of 2019 with the latter being the quarter Navi was on the market. If Navi was selling so well, how would Nvidia gain significant marketshare in the quarter it became a factor?

 

[guachi]

What is happening is the Nvidia's super cards are not only outselling Navi, they are doing so significantly. Nvidia's q2 vs q3 growth nearly match AMD's but they are doing this purely with Discrete gaming GPU's. AMD has Ryzen 2 in the equation which is likely the source of most of this growth.

Which is odd considering the only Super cards that are worth getting over their AMD competition are the 1650 Super and 1660 Super, neither of which were available in the 3rd quarter.

The reviews of the 2080 Super were super harsh. According to HUB, the 2080S was 26% faster at 4k and 17% faster at 1440 than the 5700XT and the 2080S costs 75% more. That's awful value. The 2080S is 9% faster than a 1080Ti - an equal-priced card that debuted 2.5 years before the 2080S.

And the 5700 and 5700XT are better values than the 2060S and 2070S. I guess I could see someone getting either of those cards in the 3rd quarter considering stock was so low for the AMD cards, but why would either card be a FIRST choice? I mean, the 2060S somehow is actually a WORSE value in cost/frame than the 2060. Good job Nvidia!!!!

 

[tajoh111]

Which is odd considering the only Super cards that are worth getting over their AMD competition are the 1650 Super and 1660 Super, neither of which were available in the 3rd quarter.

The reviews of the 2080 Super were super harsh. According to HUB, the 2080S was 26% faster at 4k and 17% faster at 1440 than the 5700XT and the 2080S costs 75% more. That's awful value. The 2080S is 9% faster than a 1080Ti - an equal-priced card that debuted 2.5 years before the 2080S.

And the 5700 and 5700XT are better values than the 2060S and 2070S. I guess I could see someone getting either of those cards in the 3rd quarter considering stock was so low for the AMD cards, but why would either card be a FIRST choice? I mean, the 2060S somehow is actually a WORSE value in cost/frame than the 2060. Good job Nvidia!!!!

The problem is your ignoring several factors and are only looking at a price to performance perspective, the last of which AMD isn't providing a convincing enough argument for considering it is single digits better.

When AMD took marketshare from Nvidia, it was when their cards provided dramatically better performance per dollar. E.g the 4870 or the 290/290x. AMD was complacent with this launch and basically slid into the high pricing along with nvidia. This is shown because when the 5700 and 5700xt launched, they were providing single digit better performance per dollars vs Nvidia.

AMD Radeon RX 5700 Review

The Radeon RX 5700 is the second Navi card AMD launches today. Priced at $349, it is more affordable than anything NVIDIA has to offer in this segment, yet brings more performance to the table than the RX Vega 64. What really impressed us is the greatly improved power efficiency that's just as...

www.techpowerup.com

It didn't shake up the market like the 4870 or 290x did.

These small improvements to price to performance are something most consumers can ignore for the stronger brand. It is why companies can charge inproportionally more for the last 10% of performance. Combine this with Ray tracing, quieter cards, less heat and the Nvidia brand, if you look at this from the general public's perspective, it is easy to see why Nvidia outsold AMD heavily. This is partly AMD fault and not just the public because they botched the launch due to drivers and heat/noise woes. AMD needs to provided much better performance per dollar gains as AMD CPU market shows.

AMD is walloping Intel even more with Ryzen in terms of performance per dollar and the marketshare gains have been slow.