(GURU3D RUMOR) AMD Polaris 10 GPU To Offer Near 980 Ti Performance For 299 USD?

[Silverforce11]

It might be as fast in the right DX12 games though. In legacy, non-optimized games Polaris won't be as good. GCN 1.0. Rinse. Repeat.

In my opinion, most certainly.

It should be comfortably faster than Hawaii, on paper.

Definitely, it's a new uarch so there's bound to be some driver slips in some games. I would not rule out Polaris capable of matching Fury X though.

 

[MrTeal]

One time should not make a trend.

Do many realize that the Fury X and 980Ti will NOT be the big cards from both camps soon?

It's not one time. When the 5870 and 5850 launched at $380 and $260 respectively, they were a decent step up in performance from the 4890. Even the 5850 was 25% faster than a 4890, so it was 20% or so increase in performance/$ if you look just at launch prices. If you compare it to the $180 that the 4890 was actually selling for at the time though, there wasn't an increase in perf/$ the week before the launch of 40nm cards to the week after it.

The trend towards less savings in cost per transistor in new nodes is likely not something that's going away, so we're going to be stuck with these long nodes and reduced improvement per generation. GPU's are still doing pretty good relative to the pace of CPUs, but I fear the days of 100% increases in perf/$ with a new node are long since over.

 

[guskline]

Maybe I'll sit pat and buy a year old GTX980TI for $300 since it will be so cheap and couple it with my other one to run in SLI! That should handle any game I play even at 3440x1440.

 

[Mopetar]

Or that most humans have long memories and recall the miserly increase in perf/$ the last time we had a new node.

I think that had to do with several external factors including a trend towards gaming at larger resolutions which ate up some of the performance gains along with increased demand on chip manufacturers in the form of the growing smart phone industry.

Also the move from 28 nm to 14 nm is essentially two node jumps at once, so even with higher costs and the move towards ever larger resolutions, the performance gains over 28 nm are going to be large enough for the new cards to have plenty of value.

 

[maddie]

It's not one time. When the 5870 and 5850 launched at $380 and $260 respectively, they were a decent step up in performance from the 4890. Even the 5850 was 25% faster than a 4890, so it was 20% or so increase in performance/$ if you look just at launch prices. If you compare it to the $180 that the 4890 was actually selling for at the time though, there wasn't an increase in perf/$ the week before the launch of 40nm cards to the week after it.

The trend towards less savings in cost per transistor in new nodes is likely not something that's going away, so we're going to be stuck with these long nodes and reduced improvement per generation. GPU's are still doing pretty good relative to the pace of CPUs, but I fear the days of 100% increases in perf/$ with a new node are long since over.

I would agree with you if the sales ratio was more even, but not in this situation.

We'll know soon enough.

 

[DarkKnightDude]

Don't believe anything till you actually see reviews and videos.

 

[Slaughterem]

I agree on the pricing. The Polaris will be priced in relation to its performance and availability. The 390X is $399 at Newegg give or take a few. So a new Polaris card that is faster, has more memory and feature set than a 390X is going to cost more. If it is even close to 980ti performance it will be way more than the $399 R9 390X

That could be true if they needed to use the same board, connectors, vrms etc. Remember it is a 175 TDP card. Probably the size of Nano. The gpu is rumored to be 232 mm2 compared to 438 mm2 for Grenada. Point being they will harvest more per wafer even if 14nm process is more costly than 28nm. Cooling of the card will also cost less. There is no way in telling what the cost will be when there is so many variables involved.

 

[Mopetar]

There is no way in telling what the cost will be when there is so many variables involved.

Somewhere around $300 seems to be the general thought at the moment. I'd expect a full die for about that, with a cut die for ~$250, which leaves room at $350-$375 for a binned version with GDDR5X later on to add some middle ground between Polaris and Vega.

Tack on another $25 across the board if the performance is better than expected.

 

[Phynaz]

This reminds me of the Fury thread when there were all the hopeful rumors.

And then the launch price was so high even the fanboys jaws hit the ground.

I don't think this is going to be any different. AMD isn't a charity.

 

[VulgarDisplay]

I would love if they pulled another 5870 with the Polaris 10. That was a legendary GPU that got me to upgrade from my previous flagship, and leave nvidia behind for 3 generations. The price to performance got me to switch to ati after owning 5 straight nvidia GPUs. It also showed me that the ati/AMD driver FUD was all from the past.

I don't need more performance than my r9 290, but 980ti performance at $299-$350 and I'd buy one for sure.

 

[maddie]

This reminds me of the Fury thread when there were all the hopeful rumors.

And then the launch price was so high even the fanboys jaws hit the ground.

I don't think this is going to be any different. AMD isn't a charity.

Wow, look at that, seriously, are you sure? The dentists must be happy then.

 

[MagickMan]

More like "slightly-overclocked 980 performance for $300", which is still pretty good. :hmm:

 

[IllogicalGlory]

More like "slightly-overclocked 980 performance for $300", which is still pretty good. :hmm:

For a firestrike Ultra score of 3760, you need a 1521/2065 980. That's not "slight" IMO.

 

[Phynaz]

Wow, look at that, seriously, are you sure? The dentists must be happy then.

Do you have something to contribute, or do you post just to insult?

 

[Makaveli]

http://www.guru3d.com/news-story/am...ffer-near-980-ti-performance-for-299-usd.html

Imagine if true. I will be starting a 980ti owners therapy group. To join the therapy group, reply to this thread and send me a PM to confirm your membership. We will be meeting anonymously from this point forward.

lol this is actually a good lesson for those dropping $800-$2000 on gpu's just to have them outdated in a generation or two.

And that lesson is spend your money wisely!

 

[moonbogg]

Maybe I'll sit pat and buy a year old GTX980TI for $300 since it will be so cheap and couple it with my other one to run in SLI! That should handle any game I play even at 3440x1440.

I guess you could do a lot worse than dual 980TI's

 

[ElFenix]

This reminds me of the Fury thread when there were all the hopeful rumors.

And then the launch price was so high even the fanboys jaws hit the ground.

did you even bother reading the thread?

Do you have something to contribute

one could ask you the same

 

[frozentundra123456]

This reminds me of the Fury thread when there were all the hopeful rumors.

And then the launch price was so high even the fanboys jaws hit the ground.

I don't think this is going to be any different. AMD isn't a charity.

Or the Intel thread thinking 10 core BW-E would be priced like the current 8 core.

But yea, I agree. Seems every thread about AMD new products expects revolutionary performance at prices much lower than the competition. It could happen, but recent pricing on items like the 9590 and Fury X, and even high end APUs like the 7850k when it came out, don't point to that.

Also, I think that new die shrinks save much less cost than previous ones, so there is less cost saving to pass on to the consumer.

 

[kraatus77]

Or the Intel thread thinking 10 core BW-E would be priced like the current 8 core.

But yea, I agree. Seems every thread about AMD new products expects revolutionary performance at prices much lower than the competition. It could happen, but recent pricing on items like the 9590 and Fury X, and even high end APUs like the 7850k when it came out, don't point to that.

Also, I think that new die shrinks save much less cost than previous ones, so there is less cost saving to pass on to the consumer.

So you ignored dozens of low/better priced products and assume new products will have high prices because of those 3 high priced products? lol

I guess only ngreedia is allowed to price 1k for their short lived/ cut down gpus. but amd can't even price 650$ for technically/build wise superior products which also performs for longer.

When nvidia did it with 970. it was second coming of jesus. but now when amd does it " meh, last year performance".

 

[JDG1980]

This reminds me of the Fury thread when there were all the hopeful rumors.

And then the launch price was so high even the fanboys jaws hit the ground.

I don't think this is going to be any different. AMD isn't a charity.

Except that Roy Taylor has specifically said that one of the goals of Polaris is to bring VR-capable performance to a lower price point. VR-capable is defined as R9 290X or GTX 970-class performance.

Remember, this is going to be a Pitcairn-sized chip with a moderate TDP and a 256-bit memory bus. The cards should be no more expensive to manufacture than Tonga (currently $189-$239) and much cheaper to make than Hawaii. Moreover, while Nvidia has the advantage of a bunch of diehard fanboys who won't even consider AMD, most AMD buyers are price-oriented and less attached to the brand. AMD has to provide a good deal, or the cards won't sell. That is why I am expecting MSRPs of $249 for the cut-down card with roughly 390/970 levels of performance (this would be the cheap VR card Roy Taylor was referring to), and $349 for the full SKU with performance maybe 10% or so above 390X/980. Those prices also match what we saw with Pitcairn (7850/7870) when it first came out in 2012. For buyers it will offer better perf/$ than anything currently on the market (not counting closeouts/refurbs), and for AMD it will still be considerably higher profit margins than they're getting now.

Nvidia wants to rob all the benefits of the node shrink for itself. AMD doesn't have that luxury - they need to compete hard for market share.

 

[Phynaz]

If this gets anywhere near 980 performance, let alone 980ti, for $300 and 150w power consumption (have I rounded up all the rumors?) I'll buy one on launch day.

My wallet will be safe.

 

[Phynaz]

did you even bother reading the thread?



one could ask you the same

Yes.

I made my contribution by observing and reporting the historical recurrence of the same posting style as recent AMD product launches. Do you have a problem with that?

 

[AtenRa]

If this gets anywhere near 980 performance, let alone 980ti, for $300 and 150w power consumption (have I rounded up all the rumors?) I'll buy one on launch day.

My wallet will be safe.

Take off the rumors, what do you believe Polaris 10 performance will be relatively to todays GPUs and at what price ??

 

[3DVagabond]

299USD for their most expensive GPU in their 2016 line up? Suuuure.

Maybe the Polaris 10 PRO, but not the XT!
Even the PRO I would imagine around 349 and the XT at 449USD, if they match 980ti performance...

Nvidia will announce their line up first and until then I'm guessing AMD is in stand by to look at their numbers which should beat AMD polaris in performance, and then position their solution as the best value...

I'm not very positive about nVidia launching first. The die shot that just surfaced of GP104 was an early QA sample and we haven't heard anything about GDDR5X being available yet. Which was on that board.

It's not one time. When the 5870 and 5850 launched at $380 and $260 respectively, they were a decent step up in performance from the 4890. Even the 5850 was 25% faster than a 4890, so it was 20% or so increase in performance/$ if you look just at launch prices. If you compare it to the $180 that the 4890 was actually selling for at the time though, there wasn't an increase in perf/$ the week before the launch of 40nm cards to the week after it.

The trend towards less savings in cost per transistor in new nodes is likely not something that's going away, so we're going to be stuck with these long nodes and reduced improvement per generation. GPU's are still doing pretty good relative to the pace of CPUs, but I fear the days of 100% increases in perf/$ with a new node are long since over.

I'm thinking 5800 pricing myself. AMD needs to climb out of the market share hole they're in ATM. That would do it, assuming the performance rumors are true.

This reminds me of the Fury thread when there were all the hopeful rumors.

And then the launch price was so high even the fanboys jaws hit the ground.

I don't think this is going to be any different. AMD isn't a charity.

I thought we were expecting it to cost more but when nVidia launched the 980 ti @ $650 it forced AMD to reduce prices.

 

[showb1z]

If this gets anywhere near 980 performance, let alone 980ti, for $300 and 150w power consumption (have I rounded up all the rumors?) I'll buy one on launch day.

My wallet will be safe.

We're keeping you to that.

980 is 180W.
980 is way overpriced but 970 is $300 for ~10-15% less performance.

And you honestly think AMD can't do better than that on 14nm? Your bias is blinding you.