Anybody else unimpressed with new midrange Nvidia GPUs, and much higher MSRP?

[Silverforce11]

VR World rumor says that Polaris 10 will fit in both 480 and 490 series, with Tweaktown saying non-x 480 packs already 390x performance. Benchlife says it is a ~115W ACP card, Wccf says it performs like a TitanX on FStrike at far less than the 170W card TDP. And the rumor mill says it packs perf/mm2 and perf/w lead.

So is not too much optimism to expect Polaris 10 beatimg GTX 1070.

The problem is those sites are very unreliable. VR World just before the 1080 launch, was claiming 100% sure on 1920 CC.. which I gave reasons why it was BS, it's 2560.

A lot of rumors come in, you have to use your logic to filter out the clearly wrong ones. VR World editors have less logic than some of us forum warriors so don't believe anything from them in the future.

The rest are just repeating each other, an echo chamber.

There's really only a few solid sources, they are involved with or work for AIBs. MSI in particular is very leaky.

At this point, we can only go by facts released in the public from AMD.

Polaris 11 and 10 are replacing their entire lineup until VEGA. Take from that what you will.

 

[el etro]

The problem is those sites are very unreliable. VR World just before the 1080 launch, was claiming 100% sure on 1920 CC.. which I gave reasons why it was BS, it's 2560.

In the truth it makes all sense even if info is not tottaly right on both cases. Tweaktown said R9 480 packs 390x performance thinking that this SKU would be top bin P10, when Wccf says full P10 performs around TitanX on Firestrike Ultra. All in-line for now.<br />


The VR world rumor may seen very wrong, but one thing is to note: 1920SM to 2560SM equals proportionally the difference between 1070 and 1080 claimed tflops. Around 1900SM is what is expected from a 6.5Tflop card being cut down of a 9Tflop card.

 

[Bacon1]

especially when you consider the addition of FinFETs take die space,

Why do you say that? Isn't the whole point of finfet to save space?

 

[MrTeal]

Why do you say that? Isn't the whole point of finfet to save space?

No, it's to reduce leakage.

 

[tviceman]

GP100 in a GTX form, should be ~25% faster than the 1080.

1080 has 2/3 of the GP100's FP32 CC, and none of it's hindrances such as 1:2 FP64 & NVLink. P100 Tesla already hit the 300W barrier, as a cut-down part with 1480mhz boost clocks. Don't expect the GTX variant to be running 1.86ghz.

This is why some people think there's a gaming all FP32 GP102 chip somewhere.

I was among the crowd that believed a GP102 existed; but not anymore. There is no room between GP100 and GP104 to comfortably exist, and the R&D/cost to make a ~475-500mm2 SP focused chip that would overlap in performance with GP100 in all segments except DP doesn't justify it's existence.

Nvidia went all out on DP with GP100 because - and this is a guess - Nvidia is going to live with GP100 in the professional space for as long as they did with GK110/210 because Volta may end up being SP focused like Maxwell and they wanted to be able to counter Intel better through the life of GP100. Why will GP100.

It looks like Nvidia is adopting a tick-tock strategy of their own, not unlike Intel. Their architecture updates since Fermi have been more focused and refined. Kepler wasn't a massive departure from Fermi except for cutting DP out of all of it's non-flagship chips. Maxwell was a refinement of Kepler. Pascal is a refinement of Maxwell. I expect Volta to come on 16nmFF and be similar to Pascal what Maxwel was to Kepler.

 

[arandomguy]

Nvidia went all out on DP with GP100 because - and this is a guess - Nvidia is going to live with GP100 in the professional space for as long as they did with GK110/210 because Volta may end up being SP focused like Maxwell and they wanted to be able to counter Intel better through the life of GP100. Why will GP100.

Volta already has announced super computer design wins. No DP would be rather problematic for all parties involved.

I think in their literature they've also shown it will scale up in DP over Pascal.

 

[provost]

How is the founders edition worse for consumers? A Nvidia is charging as high price as possible for new gpus since they have NO COMPETITION ?

They have every right to charge these prices and if people continue to buy then Nvidia clearly has the right move.

AMD is insanely idiotic to cede the high end to Nvidia. They deserve whatever Nvidia has in store for them this then as they brought it on them selves. The 1070 will repeat the 970 success and Nvidia has not only increased prices but hyper segmented again with founders editions and leaving their 1080ti til later to again get consumers to upgrade.

Nvidia just spoiled the amd parade. If you want to play blind go ahead.

I haven't read all the posts, but I don't see anything earth shattering here from Nvidia, unfortunately. A $700 mid range card with a driver prioritization period of less than a year (going by history and track record here) and a supposedly lower end card at $499/$449 or whatever that price is clocked to the moon with an older memory structure.
So what if it beats the Titan x; with Nvidia's driver prioritization approach, any newer card can beat whatever is not being driver prioritized. And to answer OP's question, the new release seems as unimpressive and flat as the presentation (I only caught a few minutes of it)

 

[Silverforce11]

I was among the crowd that believed a GP102 existed; but not anymore. There is no room between GP100 and GP104 to comfortably exist, and the R&D/cost to make a ~475-500mm2 SP focused chip that would overlap in performance with GP100 in all segments except DP doesn't justify it's existence.

Nvidia went all out on DP with GP100 because - and this is a guess - Nvidia is going to live with GP100 in the professional space for as long as they did with GK110/210 because Volta may end up being SP focused like Maxwell and they wanted to be able to counter Intel better through the life of GP100. Why will GP100.

It looks like Nvidia is adopting a tick-tock strategy of their own, not unlike Intel. Their architecture updates since Fermi have been more focused and refined. Kepler wasn't a massive departure from Fermi except for cutting DP out of all of it's non-flagship chips. Maxwell was a refinement of Kepler. Pascal is a refinement of Maxwell. I expect Volta to come on 16nmFF and be similar to Pascal what Maxwel was to Kepler.

Yes, each chip costs several hundred millions (about half of NV's profits in 2015!) in R&D to reach production stage. GP102 cannot justify it's existence as a consumer SKU because as a gaming focused chip, it cannot command the impressive margin/profit that P100 Teslas carry.

Not only that, it will be a large chip and it has to compete with 16nm wafers from TSMC. Wafers that NV would MUCH prefer to go into Teslas that sell for $127K.

A tick-tock strategy is going to be necessary because we may be on 14/16nm for awhile. TSMC always overhype their nodes, it's actually behind schedule on the 16nm volume ramping according to their projection from 2015.

As for Volta, probably too early to say, but a safe bet is an evolution of Pascal with a hardware scheduler making a return. They need it for true DX12/Vulkan Multi-Engine support (aka Async Compute, which is a very misleading term). GV100 will undoubtedly be a FP64 beast, optimized for a refined 16nm FF node, much higher clocks at similar TDP.

 

[Silverforce11]

Good news for NV owners who likes taking screenshots to import it to VR:

http://www.geforce.com/hardware/technology/ansel/supported-gpus

Ansel supports Kepler & Maxwell as well.

A software solution that's uarch agnostic.

 

[tential]

I haven't read all the posts, but I don't see anything earth shattering here from Nvidia, unfortunately. A $700 mid range card with a driver prioritization period of less than a year (going by history and track record here) and a supposedly lower end card at $499/$449 or whatever that price is clocked to the moon with an older memory structure.
So what if it beats the Titan x; with Nvidia's driver prioritization approach, any newer card can beat whatever is not being driver prioritized. And to answer OP's question, the new release seems as unimpressive and flat as the presentation (I only caught a few minutes of it)

The wait and see how well are cards hold up has done what for amd? The 1070 will eat amd alive the average gamer after hearing this launch has pledged their money to the 1070. Most won't even bother to wait for Polaris 10,and I don't blame em.

If Polaris 10 is competitive with the Nvidia launch then vega with a further improved arch would be a monster.

 

[Silverforce11]

The wait and see how well are cards hold up has done what for amd? The 1070 will eat amd alive the average gamer after hearing this launch has pledged their money to the 1070. Most won't even bother to wait for Polaris 10,and I don't blame em.

If Polaris 10 is competitive with the Nvidia launch then vega with a further improved arch would be a monster.

You can't get the 1070 until June 10th. That's the $449 Founders Edition. Who knows when custom versions will arrive and whether availability will be any good.

MSI's GP104 die shot is a QA sample they got from NV, manufactured in April from TSMC. How many months do you think QA and volume building takes for a retail launch?

If I had to guess, expect July for good availability of non-Founders Ed.

 

[JDG1980]

The wait and see how well are cards hold up has done what for amd? The 1070 will eat amd alive the average gamer after hearing this launch has pledged their money to the 1070. Most won't even bother to wait for Polaris 10,and I don't blame em.

Current rumor is that AMD will announce the Polaris release at Computex (which starts May 31). If it's an immediate hard launch (optimistic but possible), then it might actually beat GTX 1070 to market, as the latter isn't coming out until June 10.

Apple's WWDC starts June 13, and will almost certainly include some products with Polaris. Of course, if early yields are too low, it's possible that Apple might absorb all of AMD's production capacity and AIB parts won't be available until later on. I think we will be seeing at least some AIB Polaris cards in early June, though - even if not the full die.

 

[Azix]

I am optimistic about AMDs launch because they demonstrated the GPUs long ago. We just saw functioning pascal yesterday.

Also, polaris 11 is pretty much going to rule everything in the lower mid to low end. We already saw that demonstrated as well. Pretty much every oem will want that in their budget systems with their whatisthisnonamecrap 300W PSUs. To be honest its a huge deal that pascsal is launching at 1070. That leaves a massive market unaddressed. People looking to upgrade from 280x cards, 270x cards, their old 670s, 660s etc etc. and dont want to go past $300.

 

[Silverforce11]

At this point, NV would know about Polaris 10's performance. They are pricing the 1070 at $379 ($449 early access) for a good reason. Early rumors from Taiwan had it at $499.

Expect Polaris 10 to come close to the 1070 but not beating it (unless DX12 AMD titles).

 

[railven]

So wait. The Founder's edition is paying a premium for early access of NV's reference card right?

Their blog suggests the normal variety or AIB cards will come later.

NV's cooler is quite good up to 225W, it should be great for the 1070 and 1080.

On the blog that's the wording and my interpretation. Pay $699 for early access of 1080, or later $599 for custom models. Would you pay more for reference models though vs something like an MSI or EVGA SC?

Dunno if you responded to my other post (still reading through this thread) but this is what I took away from the Founder's thing. Pay more, get it first.

But without full confirmation, who knows. Either way if on May 27th you can buy @ $600, I'm in.

 

[dark zero]

At this point, NV would know about Polaris 10's performance. They are pricing the 1070 at $379 ($449 early access) for a good reason. Early rumors from Taiwan had it at $499.

Expect Polaris 10 to come close to the 1070 but not beating it (unless DX12 AMD titles).

I feel that the order might be like this (in order to best to worst, green are confirmed and red are speculated):
- GTX 1080 at USD 600 (considering that there might be a shortage)
- R9 490 at USD 500
- GTX 1070 at USD 450 (considering that there might be a shortage) = R8 480X at USD 400
- R8 480 at USD 320
- (OEM) GTX 1060 = R7 470X at USD 250
- R7 470 at USD 210
- R6 460X at USD 175
- R6 460 at USD 140

 

[piesquared]

Only if it can do >60 FPS @4K in all the major titles, with all settings maxed out. Otherwise, no.

There will be a lot more people buying in Polaris' price range than Pascal. 4K @ 60 FPS maxed out won't determine what cards are most popular, and the majority won't be at 4K resolution anyway. Further, if that is the criteria, Pascal won't be running all major titles maxed out at 4K, but I think you could anticipate that AMD is well positioned for all major titles releasing with Vulkan and DX12 APIs. Since they've been headed in that direction for multiple generations, and have handed Mantle to Kronos.

 

[frozentundra123456]

Hard to know what to think of this release until we see what AMD brings to the table. However, I always thought the expectations of previous gen high end performance at midrange prices were grossly over-rated, just like AMD cpu fans are expecting Haswell E performance at 300.00. Sales are flat, and die shrinks are becoming increasingly difficult and expensive. We pay for performance, not just die die size. I think the days of big performance jumps per dollar with new generations are sadly over.

 

[railven]

Aight, read through the thread. Seems a lot of people have been ignoring the price shift that happened with Kepler.

$600 is a decent price for me, since I paid $650 for the 980 Ti. Card is faster, uses less power, hopefully has good OC headroom, and seems to address Maxwell's DX12 deficenies. I can swing.

And since I'm selling my 980 Ti with all the upgrades for a good penny, I can't complain

With all the rebates, free games, and promos I got owning my 980 Ti, it cost me roughly $620. Cost of owner ship for a year is <$50 after I sell it.

All smiles from me

 

[jlee]

Aight, read through the thread. Seems a lot of people have been ignoring the price shift that happened with Kepler.

$600 is a decent price for me, since I paid $650 for the 980 Ti. Card is faster, uses less power, hopefully has good OC headroom, and seems to address Maxwell's DX12 deficenies. I can swing.

And since I'm selling my 980 Ti with all the upgrades for a good penny, I can't complain

With all the rebates, free games, and promos I got owning my 980 Ti, it cost me roughly $620. Cost of owner ship for a year is <$50 after I sell it.

All smiles from me

I don't really understand why so many people are so upset -- a card under $600 that easily tops a Titan X (which is how much, again?) is ripping everyone off.....? What?

A mid-range card is a GTX 960, mayyyybe a 970 if you want to stretch it that far (but they're still generally ~$300).

I'll be really tempted to get a 1080 on release and replace my 970.

 

[IEC]

Aight, read through the thread. Seems a lot of people have been ignoring the price shift that happened with Kepler.

$600 is a decent price for me, since I paid $650 for the 980 Ti. Card is faster, uses less power, hopefully has good OC headroom, and seems to address Maxwell's DX12 deficenies. I can swing.

And since I'm selling my 980 Ti with all the upgrades for a good penny, I can't complain

With all the rebates, free games, and promos I got owning my 980 Ti, it cost me roughly $620. Cost of owner ship for a year is <$50 after I sell it.

All smiles from me

$600 is not a decent price for a 980 equivalent. I say that after spending $750 on a 34" Ultrawide, so it's not that I can't afford it, I just don't see it as a good value to pay increasingly more for less of a performance jump, especially considering this time there's a node change...

 

[railven]

I don't really understand why so many people are so upset -- a card under $600 that easily tops a Titan X (which is how much, again?) is ripping everyone off.....? What?

A mid-range card is a GTX 960, mayyyybe a 970 if you want to stretch it that far (but they're still generally ~$300).

I'll be really tempted to get a 1080 on release and replace my 970.

I get the feeling a lot of people are stuck in the old days when we'd get huge gains with a new node. Unfortunately, NV seems to rather want to make money. And if they don't have to put out giant performance gains to beat the competition - they won't.

This also isn't a "mid-range" card as someone want to claim it is.

$600 is not a decent price for a 980 equivalent. I say that after spending $750 on a 34" Ultrawide, so it's not that I can't afford it, I just don't see it as a good value to pay increasingly more for less of a performance jump, especially considering this time there's a node change...

Unfortunately this is the current trend. We saw it before. Let's see if AMD is willing to shake things up akin to the 4870 days or they'll just follow NV and try to get as much as they possibly can.

As a consumer, of course I don't want to pay more. But, these are my options /shrug.

 

[Head1985]

I don't really understand why so many people are so upset -- a card under $600 that easily tops a Titan X (which is how much, again?) is ripping everyone off.....? What?

Who cares about TITANX when there is 650USD 980TI with same performance?
1080 is nothing more than GTX560TI in fermi generation.
560TI cost 250USD
1080 cost 600-700USD

5 years in future the new 560TI will cost how much?1000USD?PC gaming is pretty much dead if this continue.Soon there will be only few elitist who can afford new GPU.

 

[Silverforce11]

Unfortunately this is the current trend. We saw it before. Let's see if AMD is willing to shake things up akin to the 4870 days or they'll just follow NV and try to get as much as they possibly can.

As a consumer, of course I don't want to pay more. But, these are my options /shrug.

AMD isn't going to go cheap on Vega with HBM2. The 4870 and 5870 days a long gone.

Offering 90-100% of the performance of the GTX 280 Ultra for half the price didn't do AMD much good besides marketshare %, they didn't even profit much from that entire "victory" of two generations with the 4800 and 5800 series.

Because the market doesn't reward AMD when they hit success. At best, they are looking at 50% marketshare by selling superior GPUs at bargain break even prices and that was back in the days. NV's mindshare has undoubtedly grown since then, proof with their Founder's tax hike. If AMD strikes a home-run, how much % marketshare do you think they can get? 35%? Should they price it dirt cheap or go high and profit...

Polaris will be priced well because it's small and GDDR5 is cheap. Vega will bring the performance crown and it's gonna be hella-expensive.

 

[3DVagabond]

So much for 1070 having no external power connector. lol

also my prediction on availability was spot on.

On topic, price aren't really bad for (380$) 1070 but that 700$ for reference cooler 1080 is a ripoff for 1X4 chip. seems like early adopter fee with hype for extra oc.

Btw does anyone know how many CC does 1070 have ?

None of those prices are real prices except for the top pricing. The others are, "Partners are going to offer cards between X and Y. They'll probably subsidize EVGA to offer a bottom price tier card. I'm also fairly certain that there will be cards even more expensive like Lightnings and the top EVGA cards, just like we always have now.