AMD Launching Polaris 10 400 Series GPU June 1st At Computex. $299 (rumor)

its crazy. the shader numbers keep going down. I don't understand why AMD would chose not to launch a product with 40cu's but if 2048 shaders are performing faster than 390x with half the wattage. That is even more impressive. I could easily match the 1070 or above if they have one with 2560 shaders. May be they are trying really hard to stay under a certain price. Sort of confusing but o well.

Well, with 2048 versus 2560, they can fit 15-20% MORE chips on a wafer. That makes them much cheaper, and allows AMD to sell them for a sweet price while maintaining a decent margin.
 
Well Computex for AMD could be great or a disappointment.....
I mentioned before that the Computex invitation was sending out mixed messages as the header mentioned Polaris announcement but the details suggested it was more of an introduction to 2016 products and rather with the official announcement launch for the A-Series APU.
Well, it looks like that header was only a watermark......
This is the actual invitation:

amd-event.png



So no idea how AMD is handling their messaging on this, and as I said could either be a great Computex or rather disappointing, but I do not think we can now assume Polaris is launching at this event as without that added header/watermark (I assume by ArabHardware) the context of this invitation subtly changes.
Link to update: (Update) AMD to launch Polaris 10 at Computex | VideoCardz.com

Cheers
 
They could be launching at E3 again. It was always a possibility.
True but everyone is assuming Computex because of that invitation and the incorrect watermark applied to it.
The launch may still be Computex, but oh man that invitation+watermark that has spread through most news publications is causing an expectation, with the invitation being a messaging mess as well.
Now nothing is certain.
Cheers
 
Even if AMD launches Polaris by late June that's totally fine as long as they can score ODM victories for notebooks. A lot of notebook enthusiasts are pretty pissed at NVIDIA and have been waiting for AMD to offer them an alternative. If Polaris can beat the 980 "desktop" GPU in notebooks, then they will have a solid foothold for the enthusiast market. The problem is I can see NVIDIA quickly answering w/a cut down 1070 as well. But nonetheless, AMD has been absent in gaming notebooks for years so they will gain market share regardless.
 
True but everyone is assuming Computex because of that invitation and the incorrect watermark applied to it.
The launch may still be Computex, but oh man that invitation+watermark that has spread through most news publications is causing an expectation, with the invitation being a messaging mess as well.
Now nothing is certain.
Cheers
We've already grown accustomed to being disappointed by AMD. At this point I expect nothing less.
 
Yes it did, but how much marketshare did it grab? not to mention it was late....

So far I think they didn't grab much at all...

If its $300 yeah I think its going to be more like Nano to FuryX level.

at 5-10% slower, I think they would put it at $325, they have room to play and still make their card look like a better buy than the 1070 if its at this kind of performance.
Was gonna mention this on a post you made a few up but seeing you mentioned it again...

Given parity at the performance level between AMD and Nvidia, most users will still likely buy Nvidia as they have so far. AMD will have to be far cheaper than Nvidia to intice those same users to buy AMD instead. You see it all the time. Even when AMD leads posters will comment that the price needs to be lower yet it is already lower than the competition. The 1070 AMD counterpart would have to be $50 less to get the masses to buy it.
 
Was gonna mention this on a post you made a few up but seeing you mentioned it again...

Given parity at the performance level between AMD and Nvidia, most users will still likely buy Nvidia as they have so far. AMD will have to be far cheaper than Nvidia to intice those same users to buy AMD instead. You see it all the time. Even when AMD leads posters will comment that the price needs to be lower yet it is already lower than the competition. The 1070 AMD counterpart would have to be $50 less to get the masses to buy it.


The features that they poured into Crimson and the aggressive bug patching for their drivers is making a very positive image for their new cards. They may have to sell cheaper for awhile until consumers are convinced that all is good software wise if they purchase an AMD card. When you watch a Youtube review of AMD vs Nvidia in a certain price bracket, and the AMD is faster, it makes for a very positive image.

It doesn't have to be cheaper for a notebook PC. The sum of the parts inside the notebook need to be top notch so that the notebook can be viewed as an elite piece of hardware. Then AMD will have a much more favorable view in the minds of retailers and consumers. They can still have the cut down version of the notebook on the shelf for the lower budget people. But there needs to be an elite setup on display also. At that point AMD can charge the same premium as Intel and Nvidia.


Yes, I know I'm taking your post out of context as you were talking about desktop cards only. :) Hope you don't mind.
 
Was gonna mention this on a post you made a few up but seeing you mentioned it again...

Given parity at the performance level between AMD and Nvidia, most users will still likely buy Nvidia as they have so far. AMD will have to be far cheaper than Nvidia to intice those same users to buy AMD instead. You see it all the time. Even when AMD leads posters will comment that the price needs to be lower yet it is already lower than the competition. The 1070 AMD counterpart would have to be $50 less to get the masses to buy it.


Thats not true at all, when you have a card that draws double the power usage at similar performance you have to price it cheaper. Fermi (480) did loose market share even though it was the faster card
 
Well Computex for AMD could be great or a disappointment.....
I mentioned before that the Computex invitation was sending out mixed messages as the header mentioned Polaris announcement but the details suggested it was more of an introduction to 2016 products and rather with the official announcement launch for the A-Series APU.
Well, it looks like that header was only a watermark......
This is the actual invitation:

amd-event.png



So no idea how AMD is handling their messaging on this, and as I said could either be a great Computex or rather disappointing, but I do not think we can now assume Polaris is launching at this event as without that added header/watermark (I assume by ArabHardware) the context of this invitation subtly changes.
Link to update: (Update) AMD to launch Polaris 10 at Computex | VideoCardz.com

Cheers


I want to drive with someone on that map. I want to be able to tell the person I am riding with to keep their eyes on the Fuxing Road...


That said, I think this could end up interesting given all the shenanigans over the 1080 launch coming to light (thank you Kyle for that recent editorial detailing some of it). There was so much hype, and yes, some of the 1080 performance is real, but we are also learning a lot of what was rumored, or even what we were shown/told in that presser was total bullshit.

As such, I really have no presumptions any more about what a 1070 really is because frankly, 1) we haven't seen it yet in reviews and 2) half of what NVidia said about the 1080 was bullshit anyway.

As such, the 1070 may indeed be close to Polaris in price, but in terms of the purported performance, I am not believing the rumors any more - not for a New York second.
 
  • Like
Reactions: c3k
like this
I want to drive with someone on that map. I want to be able to tell the person I am riding with to keep their eyes on the Fuxing Road...


That said, I think this could end up interesting given all the shenanigans over the 1080 launch coming to light (thank you Kyle for that recent editorial detailing some of it). There was so much hype, and yes, some of the 1080 performance is real, but we are also learning a lot of what was rumored, or even what we were shown/told in that presser was total bullshit.

As such, I really have no presumptions any more about what a 1070 really is because frankly, 1) we haven't seen it yet in reviews and 2) half of what NVidia said about the 1080 was bullshit anyway.

As such, the 1070 may indeed be close to Polaris in price, but in terms of the purported performance, I am not believing the rumors any more - not for a New York second.

So what are you saying? That the overclocked 1080 [H] sample at 2.08ghz at 62C it's a BS?.. that at stock clocks perform on average 30% faster than 980TI is BS? And due that same reason you believe the 1070 will be slower than the Polaris price equivalent?

Aside from that shit about founder edition, What exactly nvidia said is bullshit.? What matter really, The performance is there, the overclocking is there just like they said..
 
So what are you saying? That the overclocked 1080 [H] sample at 2.08ghz at 62C it's a BS?.. that at stock clocks perform on average 30% faster than 980TI is BS? And due that same reason you believe the 1070 will be slower than the Polaris price equivalent?

Aside from that shit about founder edition, What exactly nvidia said is bullshit.? What matter really, The performance is there, the overclocking is there just like they said..
Everyone is a couch lawyer in this forum Lol. Sucks neither nvidia or amd pays anyone. You all need to relax lol
 
Everyone is a couch lawyer in this forum Lol. Sucks neither nvidia or amd pays anyone. You all need to relax lol

Right? And I mentioned the article Kyle just put up on [H] - whether intentional or not, a lot of what was said was bullshit. The dude freaking out could start there, and if he still isn't convinced, then I got a bridge in Brooklyn I can sell him that's been rendered by NVidia ;)
 
Well Computex for AMD could be great or a disappointment.....
I mentioned before that the Computex invitation was sending out mixed messages as the header mentioned Polaris announcement but the details suggested it was more of an introduction to 2016 products and rather with the official announcement launch for the A-Series APU.
Well, it looks like that header was only a watermark......
This is the actual invitation:

amd-event.png



So no idea how AMD is handling their messaging on this, and as I said could either be a great Computex or rather disappointing, but I do not think we can now assume Polaris is launching at this event as without that added header/watermark (I assume by ArabHardware) the context of this invitation subtly changes.
Link to update: (Update) AMD to launch Polaris 10 at Computex | VideoCardz.com

Cheers


Well if its an announcement for another announcement for a launch at a later date, that would be funny, now they need 2 announcements before they launch something, bring on the hype.......

We shall see soon enough.
 
Even if AMD launches Polaris by late June that's totally fine as long as they can score ODM victories for notebooks. A lot of notebook enthusiasts are pretty pissed at NVIDIA and have been waiting for AMD to offer them an alternative. If Polaris can beat the 980 "desktop" GPU in notebooks, then they will have a solid foothold for the enthusiast market. The problem is I can see NVIDIA quickly answering w/a cut down 1070 as well. But nonetheless, AMD has been absent in gaming notebooks for years so they will gain market share regardless.

That's an extremely NICHE market. I would hardly call that a big win compared to desktop. (And I remember AMD trying a similar approach to their marketing of their APU/CPU's...didn't work out too well in the long run)
 
if people can remember WAY back when AMD pulled a coup with the Radeon 4850/4870 claiming the best they had was the 4850 it made Nvidia adjust their prices instantly when they realized via marketing BS and playing the "nothing to see here card" that AMD was basing numbers on 4850 so not a slow card and def not that expensive, Nvidia caught with pants down when 4870 was actually a chunk higher performance level and the "proper" performance, AMD did great twist of facts with this one.

Now IMO 390/390x level of performance from something more then likely a good deal shorter/smaller while using like 1/2 the power and costing maybe say $320 CAN is not a bad deal considering the Nvidia card (1070) would be ~$460 or some such here, no brainer, unless oyu are omg fan boy, the Radeon will sell more based on cost alone if "rumored" performance is as its claimed to be.

Also a specific "round about" level of performance/power only means so much, there are so many other factors of a gpu, avg power, how many amps actually being thrown through that cards poor VRM, point is, it would not at all surprise me if it(Polaris 11/10) have had actual performance figures skewed to catch Nvidia off gaurd and surprise the crap out of everyone.

Also considering when I bought my 7870 years ago now, it was just shy of $340 shipped, the 2nd one cost me $280, if I can get something that is WAY more powerful, using a similar or hopefully less power envelope at hopefully no more then $275 shipped, I am going to buy it, downclock/downvolt it, and be happy as a pig in mud(need build top to bottom new PC once Zen/AM4 released, if I can afford it)

Long story short, folks are basing opinion on rumors nothing more right now, same thing happend with 4k series launch, basing expected level of perofrmance at 4850 yet 4870 was AMD "master plan" which they pulled off VERY well, and, beyond anything else, we simply do not know what kind of secret sauce they put in, guesstimated performance/power means nothing, I still personally think, 390/90x level of performance for less $ less power will make Nvidia drop their pricing as well, win win for everyone :)
 
Interesting bit of stats :

Interestingly, JPR found that only 15% of all GPUs sold are priced at $349 or higher. And of that 15% only 3% is priced at $449 or higher. So AMD’s Polaris GPUs are aimed at market segments where the volumes are inherently significantly larger compared to Nvidia’s GTX 1080 and GTX 1070.
 
Yes it did, but how much marketshare did it grab? not to mention it was late....

So far I think they didn't grab much at all...

If its $300 yeah I think its going to be more like Nano to FuryX level.

at 5-10% slower, I think they would put it at $325, they have room to play and still make their card look like a better buy than the 1070 if its at this kind of performance.

AMD Takes More GPU Market Share From Nvidia In Q1 2016 - Builds Momentum Ahead Of Polaris Launch

Looks like it did grab more Marketshare....Wow AMD is shocking me.

I guess people do love those 390/390x's

"In desktops the company grew its share to 22.7% a 1.8 percentage points gain. In notebooks the company managed to take a much more impressive chunk of the market from its rival. it now sits at 38.7% a 7.3 percentage points improvement from the last quarter."
 
AMD Takes More GPU Market Share From Nvidia In Q1 2016 - Builds Momentum Ahead Of Polaris Launch

Looks like it did grab more Marketshare....Wow AMD is shocking me.

I guess people do love those 390/390x's

"In desktops the company grew its share to 22.7% a 1.8 percentage points gain. In notebooks the company managed to take a much more impressive chunk of the market from its rival. it now sits at 38.7% a 7.3 percentage points improvement from the last quarter."

flat,800x800,070,f.jpg
 
AMD Takes More GPU Market Share From Nvidia In Q1 2016 - Builds Momentum Ahead Of Polaris Launch

Looks like it did grab more Marketshare....Wow AMD is shocking me.

I guess people do love those 390/390x's

"In desktops the company grew its share to 22.7% a 1.8 percentage points gain. In notebooks the company managed to take a much more impressive chunk of the market from its rival. it now sits at 38.7% a 7.3 percentage points improvement from the last quarter."


JPR numbers they took 3% total for the quarters since the r3xx line and fury lines were released, much less than expecting in the desktop segment. And that is included the best quarter of the year Q3.......

They need to do better, but its a slow gain uphill.

They got a good amount back in desktop which is a good start though.
 
I just hope on june 1st they dont get stupid and announce the new cards are launching in july. If i ran AMD and we had very capable cards, i would be sending them out for reviews yesterday to try and steal some nvidias new card sales. Send [H] cherry picked cards if need be and give people a reason to wait
 
JPR numbers they took 3.0% total for the quarters since the r3xx line and fury lines were released, much less than expecting in the desktop segment. And that is included the best quarter of the year Q3.......

They need to do better

They got a good amount back in desktop which is a good start though.

Hey baby steps for AMD, There was rough times for them. So any Marketshare for AMD is good. We need competition to drive down prices.
 
I think it's more about how well these crossfire together because I could run 2 Polaris 10's and use the same amount of power as my single 290x or 390x and that is really going to showcase how much more performance pre watt AMD managed at 250 to 300watts .. as two of them may be faster then the 1080 at about the same amount of power.
 
I think it's more about how well these crossfire together because I could run 2 Polaris 10's and use the same amount of power as my single 290x or 390x and that is really going to showcase how much more performance pre watt AMD managed at 250 to 300watts .. as two of them may be faster then the 1080 at about the same amount of power.
well best case..is one out of 3 games cf/sli friendly? then theres ALWAYS that one game that doesnt ever get support and never will...this will never change even with amd's best case scenario so just remember that
 
yep and crossfire and sli are like 1% of 1% of the market, and this is why support isn't top priority.
 
I would like to see AMD support bridgeless crossfire for all Polaris cards.

Would be pretty cool to see tri-fire 470x's LOL. Although I hate tri/quad sli/crossfire. Would be interesting to see.

Just because at 60w TDP. You could use all PCI-E to power them. Fill all the slots up.
 
Last edited:
NVIDIA GeForce GTX 980
0.91% , 0.92%, 0.95% , 1.04% , 1.03%

NVIDIA GeForce GTX 980 Ti
0.56% , 0.66%, 0.75% , 0.89% , 0.94%

These are the steam hardware results from December 2015 through April 2016. Can you please explain why you are so obsessed with the top 2% of the market as if that is going to make or break AMD's success for the rest of 2016?

That includes multiple generations of GPUs from multiple manufactures as well as integrated graphics. How about total 900 series addin cards from steam survey:
GTX 970 4.91%
GTX 960 3.14%
GTX 980 0.99%
GTX 980Ti 0.91%
GTX 950 0.52%

So high/mid/low:
980 980Ti 18.1%
970 46.8%
950 960 34.9%
 
For people who think Polaris 10 is only 390X performance, they will be in for a rude awakening.

Simple basics, current GCN has a high TFlops on paper, lots of shaders, but its under-utilized especially in DX11. With some DX12 that's capable of of fully utilizing GCN, we see funny things like the R290X/390 = 980Ti. That's because their TFlops rating are actually similar at full through-put.

Guess what AMD is working on with Polaris?

Everything about it is about improving utilization and shader efficiency.

Great vid that covers this topic in more details:



Finally, the 32 CU 2048 SP part leak was originally from Videocardz, they said in their article these are NOTEBOOK SKUs.

AMD's current notebook SKUs are cut down to 32 CU as well with lower clocks.

Guess what the full Polaris 10 with 40 CU running at even higher clocks will do? Over 7TFlops, or ~20% higher TFlops than the 290X/390X.

If what AMD says about GCN 4 is true, those paper TFlops will be better utilized and so it's performance will be more in-line with it's true processing capabilities. We're talking a beast of a performer in a very small chip.

Suddenly for $299, doesn't sound so awful anymore.
 
Hmm is that what I stated about where they will end up? really need a video for that...... And don't need to go into all the complex explanations either , he actually just read off marketing material, he made some mistakes about IPC and throughput, its not as straight forward as he stated, many more parts of the chip come into play.

Front end improvements give around ~18% perf/watt, don't expect all of that to go to performance.... I would say around 10% to performance and 8% to wattage reduction.

30/70 split from architecture and node respectively, to give a 2.5x perf/watt advantage over today's high end chips (r390x)

Now if we go by 2.0 perf per watt from Tonga, that will be even less....

They actually gave us all the numbers to figure out what the performance will be for a set wattage.

If they are at ~100 watts, they will get Tonga performance, if they are at ~120 watts they get 390 to 390x performance, if they are at ~150 watts 390x performance Fury X performance


AMD Polaris 10 Has 390/390X Performance

P10
If its Fury X level performance I think 300 is a good price.
If its 390x performance 250 bucks
if its tonga performance (which its not going to be thats pretty much 100%) so lets not even talk about that.
 
Last edited:
Any tech head who frequently reads white papers and partake in tech forum discussion will see that Polaris 10 has a lot of potential.

The leak 32 CU @ ~1.3ghz being a notebook chip with that kind of TFlops? Come on, not hard to see a desktop variant with higher clocks, full chip, blow the socks off the 390X and match Fury X.

Fury X for $299 with 8GB vram? It may even match or beat the heavily cut 1070.

AMD hurry up and take my money!!

Btw, for anyone who still thinks Polaris 10 is only between 390 and 390X. Think about this scenario. Take a 390 or 390X, run it at 40% higher clocks. There's no way this chip is going to be 390/X speed UNLESS GloFo fail hard and they can't clock it higher than current GCN.
 
You guys keep forgetting 1 thing. AMD has talked about being efficient. Right now we have no idea how much IPC was gained with Polaris over Hawaii.

For all we know a 2048SP Polaris could match a 390x. Which has 2816? (im tired and don't want to look it up).

Another thing too, we know a 2048SP Polaris is cut down. So do not rule out AMD.

Sometimes IPC > GHZ.
 
Here are the rumors for 470x 1280 shaders. If we go by previous launches. My little if then proof statements below. Now if you really are honest with yourself you know all previous series had similar specs.

if 470x has 1280 shaders then 480x will be 2048 shaders 32 CUs

Now if rumors are any indicaitons of Polaris starting out at 40 CUs and then 36 and then 32 (32 is definitely 480x if 470x rumors are true which seems pretty legit as its same amount of shaders as 370x)

Now if 32 CUs is the highest card then stop right here, But I doubt it so lets keep going.

if they release anything higher it would have to be 490, and if there is 490 then there has to be 490x

Now 36 CUs 2304 shaders make sense for 490 and 2560 40 CUs make sense for the 490x (May be 490x card with gddr5x)?


If they release 480x with performance faster than 390x or the same there is no point for 390 and 390x and if they get rid of those then they have to replace them. if that happens expect 490 and 490x, and than vega to replace fury and fury x soon after.

You really can't say this is me speculating. It just makes sense. They said their top of line won't come out until vega, but I do believe they will replace the entire 300 series line up before vega.

With all this said I dont know if these cards will launch in july according to nvidia calling june 1st "polaris updates" among other things. They will probably announce the specs and cards with a end of june may be july launch.
 
Last edited:
Here are the rumors for 470x 1280 shaders. If we go by previous launches. My little if then proof statements below. Now if you really are honest with yourself you know all previous series had similar specs.

if 470x has 1280 shaders then 480x will be 2048 shaders 32 CUs

Now if rumors are any indicaitons of Polaris starting out at 40 CUs and then 36 and then 32 (32 is definitely 480x if 470x rumors are true which seems pretty legit as its same amount of shaders as 370x)

Now if 32 CUs is the highest card then stop right here, But I doubt it so lets keep going.

if they release anything higher it would have to be 490, and if there is 490 then there has to be 490x

Now 36 CUs 2304 shaders make sense for 490 and 2560 40 CUs make sense for the 490x (May be 490x card with gddr5x)?


If they release 480x with performance faster than 390x or the same there is no point for 390 and 390x and if they get rid of those then they have to replace them. if that happens expect 490 and 490x, and than vega to replace fury and fury x soon after.

You really can't say this is me speculating. It just makes sense. They said their top of line won't come out until vega, but I do believe they will replace the entire 300 series line up before vega.

It makes sense for them to die harvest as many Polaris 10 as possible to fill in the gaps in performance, power and price.

GDDR5X is a potential but I don't think they will have it for this coming launch. Because Micron only started mass production and their partnership with NVIDIA pretty much means NV gets priority access on the chips.

And it may not even make any sense, because architecture changes (PreFetch/Cache improvements and better memory compression tech) may result in GDDR5 8Gbps being enough bandwidth to meet the performance target.

I was worried about their ability to get some nice clock gains, but this notebook leak at ~1.3ghz is a good indicator they can clock higher. At the webminar (mostly useless rehashing), they also noted AIBs can focus on raw clocks instead of AMD's reference design which focuses on perf/w.
 
It makes sense for them to die harvest as many Polaris 10 as possible to fill in the gaps in performance, power and price.

GDDR5X is a potential but I don't think they will have it for this coming launch. Because Micron only started mass production and their partnership with NVIDIA pretty much means NV gets priority access on the chips.

And it may not even make any sense, because architecture changes (PreFetch/Cache improvements and better memory compression tech) may result in GDDR5 8Gbps being enough bandwidth to meet the performance target.

I was worried about their ability to get some nice clock gains, but this notebook leak at ~1.3ghz is a good indicator they can clock higher. At the webminar (mostly useless rehashing), they also noted AIBs can focus on raw clocks instead of AMD's reference design which focuses on perf/w.

True, not expecting gddr5x just hoping lol. But I do believe we will see 470x 480x 490/490x atleast 4 sku's launched. I am pretty confident with all the different specs I have heard for polaris they must be different sku's.
 
True, not expecting gddr5x just hoping lol. But I do believe we will see 470x 480x 490/490x atleast 4 sku's launched. I am pretty confident with all the different specs I have heard for polaris they must be different sku's.

AMD did say Polaris 10 supports GDDR5X at the recent partner meeting. If it does, I don't think there would be anything stopping AIB suppliers in shipping their own version of Polaris 10 with GDDR5X, even if AMD themselves didn't release one. In fact, the whole thing sounded like AMD was going to be very reserved with their own releases and allow their suppliers a lot of room to release very different versions.
 
Last edited:
I'm seeing some optimistic speculation here, and as somewhat of an AMD fan I hope that it pans out. Because as of right now, the rumors aren't looking good.

390/390X performance means -10% to +20% performance compared to a GTX 970, depending on the game.
Up to 150W means same power draw as a GTX 1070, and 5W more than a reference-specced GTX 970
$299 means $30 cheaper than a GTX 970 (MSRP)

Essentially, we'd be getting a slightly faster, slightly cheaper, 2-year old Nvidia card. These rumors need to be wrong, or AMD will be failing hard.
 
Back
Top