AMD Polaris 10 and Polaris 11 Specifications leak

[-BirdiE-]

I'm really intrigued to see what these new Polaris/Vega cards will do. I know there's not a ton of difference between 16nm and 14nm, but it doesn't take much of an edge these days for a card to "win" a segment. I guess we'll see when ALL the benchmarks are out within a month.

[Enderman]

2 minutes ago, Prysin said:

you know you can adjust firestrike settings? You also know that none of the slides are validated.

Thus we both know this is just smoke being blown up our asses.

lol sure lets just set the resolution to 240p just to make a biased graph to show to everyone!

 

the benchmarks fit in pretty well with what nvidia claims and with what all the other GPUs can do (eg. 970 is similar to 390)

also GPU benchmark leaks so close to the release date have pretty much always been true in the past, so expect the official reviews from linus, pcper, etc.. to be almost exactly the same

 

if you dont want to believe it then that's your problem, but you're pretty much lying to yourself at that point

[Memories4K]

2 minutes ago, Enderman said:

lol sure lets just set the resolution to 240p just to make a biased graph to show to everyone!

 

the benchmarks fit in pretty well with what nvidia claims and with what all the other GPUs can do (eg. 970 is similar to 390)

also GPU benchmark leaks so close to the release date have pretty much always been true in the past, so expect the official reviews from linus, pcper, etc.. to be almost exactly the same

 

if you dont want to believe it then that's your problem, but you're pretty much lying to yourself at that point

AMD's not above unethical and corporate business tactics.

 

I still think Polaris 10's going to do very well, but i'm not backing AMD itself.

[ivan134]

1 minute ago, Enderman said:

lol sure lets just set the resolution to 240p just to make a biased graph to show to everyone!

 

the benchmarks fit in pretty well with what nvidia claims and with what all the other GPUs can do (eg. 970 is similar to 390)

also GPU benchmark leaks so close to the release date have pretty much always been true in the past, so expect the official reviews from linus, pcper, etc.. to be almost exactly the same

 

if you dont want to believe it then that's your problem, but you're pretty much lying to yourself at that point

I'm not going to comment on how the 1080 actually performs since no one knows, but Videocardz is not a reliable source. They're even worse than WCCF.

[Enderman]

2 minutes ago, Memories4K said:

AMD's not above unethical and corporate business tactics.

 

I still think Polaris 10's going to do very well, but i'm not backing AMD itself.

polaris is mid-low end

you can tell by how it perform similar to last gen and just uses far less power
 

vega (high end) is getting pushed to Q4 2016 because of the 1070 and 1080 release

problem is that AMD is having problems producing good vega chips

1 minute ago, ivan134 said:

I'm not going to comment on how the 1080 actually performs since no one knows, but Videocardz is not a reliable source. They're even worse than WCCF.

yeah it not reliable, doesnt mean they cant be correct like they have in the past

[ivan134]

Just now, Enderman said:

polaris is mid-low end

you can tell by how it perform similar to last gen and just uses far less power
 

vega (high end) is getting pushed to Q4 2016 because of the 1070 and 1080 release

problem is that AMD is having problems producing good vega chips

yeah it not reliable, doesnt mean they cant be correct like they have in the past

I've never seen it happen. In recent memory, they were the source of all the junk rumors that were put out prior to the release of AMD's 300 series that caused people to accuse AMD of overhyping and not releasing anything.

[Memories4K]

1 minute ago, Enderman said:

polaris is mid-low end

you can tell by how it perform similar to last gen and just uses far less power
 

vega (high end) is getting pushed to Q4 2016 because of the 1070 and 1080 release

problem is that AMD is having problems producing good vega chips

yeah it not reliable, doesnt mean they cant be correct like they have in the past

New architecture on a new size.

I'll hold my judgments, but i'm still optimistic form what i've seen.

 

A broken clock is right twice a day, yes...but only twice a day

[Enderman]

1 minute ago, Memories4K said:

New architecture on a new size.

I'll hold my judgments, but i'm still optimistic form what i've seen.

optimistic for actually efficient GPUs from AMD? yup

optimistic for polaris to perform like a 1070? lol no

[Citadelen]

These look like they'll be some pretty good mobile chips, AMD has done well here, especially considering the clock speed because this is still GCN.

[Sakkura]

17 minutes ago, -BirdiE- said:

I'm really intrigued to see what these new Polaris/Vega cards will do. I know there's not a ton of difference between 16nm and 14nm, but it doesn't take much of an edge these days for a card to "win" a segment. I guess we'll see when ALL the benchmarks are out within a month.

These lithography numbers are pretty imprecise anyway. It's not like Samsung/GloFo 14nm is the same size as Intel 14nm, for example.

10 minutes ago, Enderman said:

polaris is mid-low end

you can tell by how it perform similar to last gen and just uses far less power
 

vega (high end) is getting pushed to Q4 2016 because of the 1070 and 1080 release

problem is that AMD is having problems producing good vega chips

If Vega is being moved up from Q1 2017 to Q4 2016, if anything that indicates they're doing better than expected in terms of producing good Vega chips.

[Memories4K]

7 minutes ago, Sakkura said:

If Vega is being moved up from Q1 2017 to Q4 2016, if anything that indicates they're doing better than expected in terms of producing good Vega chips.

It seems more like desperation or surprise to me.

Notice how that news came out only after Nvidia showed off the 1080/1070.

 

It could also be that AMD's been pretty confident this entire time and has only been trying to gauge Nvidia's own progress on Pascal in relation to Vega (meaning, yes, Vega is progressing much quicker than expected) but i can't really say.

[Enderman]

10 minutes ago, Sakkura said:

If Vega is being moved up from Q1 2017 to Q4 2016, if anything that indicates they're doing better than expected in terms of producing good Vega chips.

They're moving it up because of competition

 

And no, they aren't doing great in terms of good chips apparently

http://www.pcper.com/news/General-Tech/Say-it-aint-so-AMD-Polaris-may-not-shine-Computex

[RagnarokDel]

36 minutes ago, Enderman said:

http://www.overclock3d.net/articles/gpu_displays/gtx_1080_overclocked_3dmark_performance_leaked/1

Overclock a 980 TI and you'll see that it's much closer to the 1080 as you might think. Of course it's going to be more performant right off the bat when it boosts at 1733mhz.

[That Norwegian Guy]

GPU upgrade isn't as critical this year as CPU upgrades, unless you're early adopter of VR or you skipped the previous generation. Bring on ZEN!

 

That said, I won't mind having a look at what a cut-down Vega around the $500 price point can do.

[Memories4K]

1 minute ago, That Norwegian Guy said:

GPU upgrade isn't as critical this year as CPU upgrades, unless you're early adopter of VR or you skipped the previous generation. Bring on ZEN!

Zen's not exciting anything within me anymore.

Last i checked we're looking at Haswell IPC.

Maybe the Octacores and Hexacores will do well if they perform similarly to Broadwell-E at a much lower price, but i'm not impressed this late in the game.

[Citadelen]

Just now, Memories4K said:

Zen's not exciting anything within me anymore.

Last i checked we're looking at Haswell IPC.

Maybe the Octacores and Hexacores will do well if they perform similarly to Broadwell-E at a much lower price, but i'm not impressed this late in the game.

They've bumped it up to greater than 40% IPC gain, so that is/or almost at Broadwell levels if they're correct.

[That Norwegian Guy]

1 minute ago, Memories4K said:

Zen's not exciting anything within me anymore.

Last i checked we're looking at Haswell IPC.

I'll be happy with Ivy Bridge IPC. What matters is the number of cores at an acceptable price point, and how much it overclocks.

 

Because IPC matters not if, as with the three previous Intel generations, the overclocking headroom went down so much that the IPC didn't matter in end performance over Ivy/ Sandy.

 

Sandy -> Ivy, 6% IPC improvement, 5% average overclocking headroom lost

Ivy -> Haswell, 7.04% IPC improvement, 12% overclocking headroom lost

Haswell -> Broadwell, nobody cares about Broadwell

Broad -> Skylake, 6.5% IPC improvement, 3% overclocking headroom regained

 

TLDR, Ivy Bridge still end performance champion

[Starelementpoke]

13 minutes ago, Enderman said:

They're moving it up because of competition

 

And no, they aren't doing great in terms of good chips apparently

http://www.pcper.com/news/General-Tech/Say-it-aint-so-AMD-Polaris-may-not-shine-Computex

Blocked for me being at school, what do it say?

[WereCat]

32 minutes ago, Enderman said:

optimistic for actually efficient GPUs from AMD? yup

optimistic for polaris to perform like a 1070? lol no

GTX 1070 - 6.5 TFlop - $379 MSRP

based on this topic Polaris 10 - 5.5 TFlop -$??? MSRP if it will be $299 it will be sitting on the right spot in price/performance compared to 1070.

 

Also there is another factor, how well will performance scale when we apply OC both on GTX 1070 and Polaris 10.

[TidaLWaveZ]

2 minutes ago, Starelementpoke said:

Blocked for me being at school, what do it say?

"There is a rumour floating around this morning and it is not good.  Guru of 3D translated a story over at Nordic Hardware which suggests AMD has stated they will not have any new working cards to show at Computex."

[Starelementpoke]

1 minute ago, WereCat said:

GTX 1070 - 6.5 TFlop - $379 MSRP

based on this topic Polaris 10 - 5.5 TFlop -$??? MSRP if it will be $299 it will be sitting on the right spot in price/performance compared to 1070.

 

Also there is another factor, how well will performance scale when we apply OC both on GTX 1070 and Polaris 10.

Assuming all the leaked ¨benchmarks¨ are to be believed for the 1080, and assuming that those were done with the 2000 mhz oc on it, well I´m thinking for the 1070?

[Starelementpoke]

1 minute ago, TidaLWaveZ said:

"There is a rumour floating around this morning and it is not good.  Guru of 3D translated a story over at Nordic Hardware which suggests AMD has stated they will not have any new working cards to show at Computex."

Hopefully a rumor and nothing more.

[Enderman]

8 minutes ago, Starelementpoke said:

Blocked for me being at school, what do it say?

Say it ain't so AMD; Polaris may not shine at Computex?

Subject: General Tech | May 12, 2016 - 05:56 PM | Jeremy Hellstrom
Tagged: rumour, Polaris, computex, amd

There is a rumour floating around this morning and it is not good.  Guru of 3D translated a story over at Nordic Hardware which suggests AMD has stated they will not have any new working cards to show at Computex.  The only Polaris hardware they will have ready will be equivalent to the current R9 390 and 390X, albeit at a lower price point.  The rumoured problem is that the new flagship cards simply won't hit 850 MHz reliably, which in turn means high end GPUs are right out. 

This had better not be true or AMD may find themselves shoeless and GPU enthusiasts will be as disappointed as White Sox fans back in 1919, albeit for different reasons.  Those in the NVIDIA camp would do well to remember this has an effect on them as well; why would NVIDIA lower the price on those shiny new 1080's or 1070's when there is nothing in the market to compete with them? 

This is a rumour from an anonymous source at an AMD partner, so be sure to take it with a grain of salt and hope that it is completely unsubstantiated; or that a silicon-based miracle happens in the coming months if there is some substance to this.

"Okay before I start on this news-item, I really need to state that this is based on a vague rumor, nothing has been confirmed or denied otherwise. Here's the story, some reports say Polaris 10 can't hit 850 MHz reliably and that availability will be pushed back to October. I sincerely hope the rumor is not true."

 

 

 

basically TLDR rumours say it cant hit 850 MHz reliably

hopefully not true

[WereCat]

4 minutes ago, Starelementpoke said:

Hopefully a rumor and nothing more.

AMD has already debunked this rumor saying that they will be at Computex. 

 

[MEC-777]

Not sure how many of you know this, but I believe AMD is planning on going a different route with their high-end GPUs. Instead of using bigger GPU dies, which cost more due to lower yields and higher defects (per die), they might be going a different direction and start using multiple smaller GPU dies, per card. Think of it like 2 or 4 Polaris 10 chips on one card that work together as one larger single GPU chip. The result would be a more powerful card but potentially at a lower cost due to higher yields and lower defects (per die). 

 

I think this is what we might see with Vega. AdoredTV made a video on this and it actually makes a lot of sense, IMO.