(Updated) AMD Navi GPU to Offer GTX 1080 Class Performance at ~$250 Report Claims

[Valentyn]

22 minutes ago, leadeater said:

It's not my segment. In actual fact Titan was never intended for gaming at all anyway, consumers just usurped it's purpose because Nvidia kept withholding the fully unlocked dies from the standard GTX lineup. So rather than waiting the people with the money just brought the Titans.

 

Simplest solution is to just ignore their existence, I know I do heh.

I couldn’t do that sadly

 

I miss getting the full die for a 6th or th  price they currently are. My GTX 580 is still running at least.

[leadeater]

18 minutes ago, Valentyn said:

I couldn’t do that sadly

 

I miss getting the full die for a 6th or th  price they currently are. My GTX 580 is still running at least.

Well I mean the 1080 Ti just comes later now, getting the Titan is just an early access pass to the unlocked die.

[Humbug]

LOL. I think it's deluded to think that the GTX 1060 is budget and the GTX 1080 mid-range. Even at MSRP...

[Valentyn]

39 minutes ago, leadeater said:

Well I mean the 1080 Ti just comes later now, getting the Titan is just an early access pass to the unlocked die.

Early Access with a price attached. Even the the Ti costs more than the old full dies did. 

If only AMD had  another great arch to really drive into NV and lower prices again. 

[Razor01]

Look the tiers come from one place, JPR

 

The way they list them are

 

0-149 value

150-299 mainstream

300-599 performance

600 and up enthusiast

 

Now if you think gp 104 is a mid range DIE, yeah its not the top level sku.

 

But its performance is just as good as last gens enthusiast performance.  They are set at performance to price.  There is no denying any of this.  Just because nV was able to create a smaller die chip that gives more performance doesn't mean its a mid range chip (mainstream). 

 

Prior to Maxwell there was NO enthusiast segment.  So we can't even compare to older gens prior to this.  The gp104 is now the same as the old gx100 chips.  nV made one tier above that.  Just because the name is gp104, its not a mid range (mainstream), its based on performance they deliver and that tells the market where its priced at.

 

Consumers don't need to care about die size lol, if a GPU the size of a penny beats out V100, its going to be priced higher than a V100.

[inmytaxi]

This is not a big deal. The 1080 was down to under $500 last year, and would down even further now if it was not for the mining. The natural price of the 1080 should approach the $300 range by the end of this year (absent mining), and of course new tech should be cheaper. New tech should always be a better deal per dollar. Once we put mining into the equation, there is no way to see what the actual street price will be, but it won't be as great a deal as people are saying. $250 for a 1080 in mid 2019 is not a such a big deal.

[AresKrieger]

1 hour ago, Humbug said:

LOL. I think it's deluded to think that the GTX 1060 is budget and the GTX 1080 mid-range. Even at MSRP...

Well tbf when this Navi gpu comes out the old 80 series will likely be around 60 series performance so it will be mid range then  

 

But yeah the 1080 is by no means mid range 1080ti is enthusiast 1080 is high, 1070 high mid, 1060 is low mid, and 1050/ti is budget or low, people just keep listening to that idiot at adored who is an AMD fanboy.

[Deus Voltage]

10 minutes ago, AresKrieger said:

Well tbf when this Navi gpu comes out the old 80 series will likely be around 60 series performance so it will be mid range then  

 

But yeah the 1080 is by no means mid range 1080ti is enthusiast 1080 is high, 1070 high mid, 1060 is low mid, and 1050/ti is budget or low, people just keep listening to that idiot at adored who is an AMD fanboy.

He is also the same individual whom, as I argued in a previous forum post, posted countless videos critiquing AMD and a healthy amount giving credit when credit is due to Nvidia. Yea at this point, people are hating on him without making substantive arguments. 

 

 

 

[Mira Yurizaki]

Again with this "GP104 is midrange"

 

Can I call a Core i7-8700K entry level due to its paltry offering of six cores when Intel has 28 core processors available? Maybe Ryzen 7 is entry level too when AMD has 32-core processors?

[GoldenLag]

Why is this thing about the 1080 being mid-range or not still going? Going by Adored the 1080 is a step-down card that can be argued to be called a mid-ranged GPU. What are we arguing? When does something become me low-end or High-end? 

 

Say, when does a battlecruise become a battleship? When is the Battleship a battlecruiser? Does anyone care?

[GoldenLag]

2 minutes ago, M.Yurizaki said:

Again with this "GP104 is midrange"

 

Can I call a Core i7-8700K entry level due to its paltry offering of six cores when Intel has 28 core processors available? Maybe Ryzen 7 is entry level too when AMD has 32-core processors?

In some sence you can. If you have the entire range of products in front of at your offer you could call them entry level. In the consumer/prosumer market you can say the 16 core threadripper are high end and the R7 is mid rabge and so fourth. Im calling the 16 core high end due to that is whats readily avavible to consumers

[Mira Yurizaki]

Just now, GoldenLag said:

In some sence you can. If you have the entire range of products in front of at your offer you could call them entry level. In the consumer/prosumer market you can say the 16 core threadripper are high end and the R7 is mid rabge and so fourth. Im calling the 16 core high end due to that is whats readily avavible to consumers

NewEgg sells the 32-core EPYC processor. So by your rules, Threadripper is now midrange.

 

So at this point, calling something low-end, midrange, high-end is a matter of opinion. I recall there was some kerfluffle on PC Gamer when they called an $800 build a "budget" build and everyone was like "that's too expensive." They argued it's really a matter of opinion. Which I can't argue with that. If you live paycheck to paycheck, sure $800 is not "budget", but for middle class America, that's certainly within the realm of being a budget build.

[GoldenLag]

4 minutes ago, M.Yurizaki said:

NewEgg sells the 32-core EPYC processor. So by your rules, Threadripper is now midrange.

It shure is. Now i wouldnt call them rules. They are more like guidelines as long as you dont go ahead calling the r3 1200 high end. At that point we know youre trolling 

[Mira Yurizaki]

Just now, GoldenLag said:

It shure is. Now i wouldnt call them rules. They are more like guidelines as long as you dont go ahead calling the r3 1200 high end. At that point we know youre trolling 

But it's still a matter of perception and where your baselines are. If your baseline starts at something like an Atom, then the R3 1200 starts looking pretty high end.

 

If anything, we should really avoid the performance tier monikers. If I can afford $1000 video cards and processors any time I want, then anything lower starts to look like peasant tier.

[GoldenLag]

3 minutes ago, M.Yurizaki said:

But it's still a matter of perception and where your baselines are. If your baseline starts at something like an Atom, then the R3 1200 starts looking pretty high end.

 

If anything, we should really avoid the performance tier monikers. If I can afford $1000 video cards and processors any time I want, then anything lower starts to look like peasant tier.

The reasom i used the 1200 was because its the lowest performing Zen Chip

[Razor01]

47 minutes ago, Deus Voltage said:

He is also the same individual whom, as I argued in a previous forum post, posted countless videos critiquing AMD and a healthy amount giving credit when credit is due to Nvidia. Yea at this point, people are hating on him without making substantive arguments. 

 

 

 

 

 

Look if he isn't a fanboy of AMD he wouldn't have hyped the shit out of Polaris lol, yeah 980ti performance levels

 

Masterplan video series?  How about those?  What changed in his view of 1 year ago?  Look we know the tech industry moves fast, but 1 year for a multi billion dollar company to change direction from corning the market to getting crushed, doesn't happen that way.  It was years in the making for things like that happen.  If anyone was paying attention to how nV pushed CUDA and marketed to colleges and have hired people for specific programming tasks related to HPC, knows CUDA was what the market was looking for lol.  You don't see mid three digit salaries on Open CL programmers, on CUDA though you do.   This started in around 2007, its been a decade since then.

 

Gaming, wasn't it obvious how much trouble AMD was in with Maxwell.  I saw it with the 750ti, that card even though low end and first generation of Maxwell, showed so much potential.  I'm sure at that point AMD knew they were in deep shit.  When did we ever see 50% changes in Perf/watt on the same node?  The last time it was close to that never reached that much was g80.  Before that, never.  We have never seen this on the CPU side either.

 

Pascal video where he mentions it won't be much more over Maxwell because of process change, nV won't be able to do custom transistor layouts like they did with Maxwell?

 

People with any base knowledge of how chips are designed and manufactured will cringe at the things he has stated.  There was no way he could support Vega, even the most diehard fans know its not very good at gaming when it comes to everything it has shown vs the competition.

 

After that he couldn't say anything about Vega because of how wrong he was about his "forecasting", its not even forecasting its throwing a dart blind folded with a red blind fold on.   So GPU side he had to tone it down a bit and go the other way. 

 

Look at the conlake video did that really come to light?  We have those lower end boards out right now, I can't see that video as realistic.  I even pointed this out when it was posted here, can't make judgements and conclusions based across reviews like he did.  In one review with controlled settings, there is a mean deviation and error percentile values, if you start using different reviews, different data sets, you mean deviation and error percentiles are so large its not valid anymore, even for comparison sake they will be way out of bounds.

 

The pcper thing, that was all BS, he had issues with PCper calling him out in the past.  If someone wants to find errors in what another does, yeah every one can do that.  Because being perfect is extremely hard if not impossible to do. 

 

Lets not take away ALL the mistakes this guy did vs what PC per did by not putting up a disclaimer OK?

 

Big differences in errors.  One is lack of knowledge and passing it off as he has some, which is ignorance and damages his audience, vs a person that has the knowledge and does a good job getting that across but not telling us where and how they got the tests to being with.

[Valentyn]

2 hours ago, Razor01 said:

Look the tiers come from one place, JPR

 

The way they list them are

 

0-149 value

150-299 mainstream

300-599 performance

600 and up enthusiast

 

Now if you think gp 104 is a mid range DIE, yeah its not the top level sku.

 

But its performance is just as good as last gens enthusiast performance.  They are set at performance to price.  There is no denying any of this.  Just because nV was able to create a smaller die chip that gives more performance doesn't mean its a mid range chip (mainstream). 

 

Prior to Maxwell there was NO enthusiast segment.  So we can't even compare to older gens prior to this.  The gp104 is now the same as the old gx100 chips.  nV made one tier above that.  Just because the name is gp104, its not a mid range (mainstream), its based on performance they deliver and that tells the market where its priced at.

 

Consumers don't need to care about die size lol, if a GPU the size of a penny beats out V100, its going to be priced higher than a V100.

The 1080 = Mid range comes from the price segment being based on GPU die

 

Before Kepler and Maxwell the Fully unlocked die the the GF110 = GTX 580 for $500

Cutdown GF110 = GTX 470

The mid range die GF114 was the GTX 460

 

With Pascal GP100 = Tesla and Quadro

Cutdown GP102 = Titan and 1080Ti

Mid chip GP104 = 1080

Small chip = GP106

 

With Kepler that all changed, when the GTX 680 using GK104 was launched at previous unlocked die costs.

Then the Titan hit at $999 but was a GK100.

 

Pricing shifted as a result; but many old farts like me reminisce about getting a fully unlocked Die aka Titan for far cheaper. 

Overall though Generational performance jumps have mostly stayed in the same 40-60% when comparing each Architecture full die to the next;

Going from 7900 > 8800 > 285 > 580 > Titan Black > Titan X > Titan Xp > Titan V

Just pricing and naming has changed; with the entrance of a new segment.

Most of that can be thanks to AMD dropping the ball, the 7970 wasn't a big enough leap over the GTX 580, and NVIDIA could easily compete with the 680 which was much easier to produce and cheaper than the full Kepler die they used for the Titan/Black

 

AMD really need to shift away from GCN, and get a new architecture out there; they've hit a wall, and underlying arch is just not cutting it. It costs too much and delivers too little.

 

I can't say I'm helping either; I keep buying the top end whenever I can; even the current one.
 

[Deus Voltage]

On 4/14/2018 at 5:00 PM, Razor01 said:

Look if he isn't a fanboy of AMD he wouldn't have hyped the shit out of Polaris lol, yeah 980ti performance levels

Masterplan video series?  How about those?  What changed in his view of 1 year ago?  Look we know the tech industry moves fast, but 1 year for a multi billion dollar company to change direction from corning the market to getting crushed, doesn't happen that way.  It was years in the making for things like that happen.

Polaris was actually not bad at all before the mining craze, specifically when you factor in the chill functionality from the software side. 

Quote

Pascal video where he mentions it won't be much more over Maxwell because of process change, nV won't be able to do custom transistor layouts like they did with Maxwell?

Remember when GPU's in the mid-early 2000's used to have huge percentage bumps in performance? True, they were basically space heaters, but the performance increase with each generation was substantial. Nowadays it is somewhere between 30 to 45%. We don't see huge increase anymore, mainly I would argue because Moor's law lost much of its steam. But there is a good argument to be made with regards to monopolistic practices as well. And yes, I do believe AMD would take advantage of consumers just as much as Nvidia would if they were well positioned.

 

Another thing to keep in mind is that Nvidia is hiding a lot of firepower. AMD being admittedly not very competitive is the reason why Nvidia isn't bringing "the good stuff" to the consumer market.  

Quote

People with any base knowledge of how chips are designed and manufactured will cringe at the things he has stated.  There was no way he could support Vega, even the most diehard fans know its not very good at gaming when it comes to everything it has shown vs the competition.

I'm not informed or educated enough to comment on that. 

Quote

After that he couldn't say anything about Vega because of how wrong he was about his "forecasting", its not even forecasting its throwing a dart blind folded with a red blind fold on.   So GPU side he had to tone it down a bit and go the other way. 

 

He did not just tone down a bit. He provided some very interesting criticisms with regards to the Radeon group.

Quote

Look at the conlake video did that really come to light?  We have those lower end boards out right now, I can't see that video as realistic.  I even pointed this out when it was posted here, can't make judgements and conclusions based across reviews like he did.  In one review with controlled settings, there is a mean deviation and error percentile values, if you start using different reviews, different data sets, you mean deviation and error percentiles are so large its not valid anymore, even for comparison sake they will be way out of bounds.

That is actually a valid point I haven't considered. Different reviewers used different test benches and methodologies. I concur with your statement.

 

Edit: That's actually a bit misleading. There is a reason why a site like "silicon lottery" exists. Once you remove GPU bottleneck and are able to consistently cool down the CPU then it doesn't matter what other people's methodology is. That's why people, in theory, watch multiple reviews and are to make their purchase. 

Quote

 

The pcper thing, that was all BS, he had issues with PCper calling him out in the past.  If someone wants to find errors in what another does, yeah every one can do that.  Because being perfect is extremely hard if not impossible to do. 

 

Lets not take away from ALL the mistakes this guy did vs what PC per did by not putting up a disclaimer OK?

Yea I agree with that. He may be correct, but he should have been more professional in retrospect. Somewhat similar to what CTS labs did to AMD. 

Edited April 16, 2018 by Deus Voltage

[Razor01]

Just now, Valentyn said:

The 1080 = Mid range comes from the price segment being based on GPU die

 

Before Kepler and Maxwell the Fully unlocked die the the GF110 = GTX 580 for $500

Cutdown GF110 = GTX 470

The mid range die GF114 was the GTX 460

 

With Pascal GP100 = Tesla and Quadro

Cutdown GP102 = Titan and 1080Ti

Mid chip GP104 = 1080

Small chip = GP106

 

With Kepler that all changed, when the GTX 680 using GK104 was launched at previous unlocked die costs.

Then the Titan hit at $999 but was a GK100.

 

Pricing shifted as a result; but many old farts like me reminisce about getting a fully unlocked Die aka Titan for far cheaper. 

Overall though Generational performance jumps have mostly stayed in the same 40-60% when comparing each Architecture full die to the next;

Going from 7900 > 8800 > 285 > 580 > Titan Black > Titan X > Titan Xp > Titan V

Just pricing and naming has changed; with the entrance of a new segment.

Most of that can be thanks to AMD dropping the ball, the 7970 wasn't a big enough leap over the GTX 580, and NVIDIA could easily compete with the 680 which was much easier to produce and cheaper than the full Kepler die they used for the Titan/Black

 

AMD really need to shift away from GCN, and get a new architecture out there; they've hit a wall, and underlying arch is just not cutting it. It costs too much and delivers too little.

 

 

 

Its not an locked die, the gp104 is a full die lol. Its not the biggest die in the family though.

 

That is what changed.

 

Keplar, that Titan was a totally different class of chip, it had HPC components and warranted that price.  Even years after that Titan's resale value was still as high as the MSRP or close to it.

 

nV saw the success of that line and rolled out the enthusiast level tier.

[Razor01]

Just now, Deus Voltage said:

Polaris was actually not bad at all before the mining craze, specifically when you factor in the chill functionality from the software side.

 

 

Polaris was bad and still is, when comparing to a gtx 1060 a card that has 30% less flops, 25% less bandwidth, 25% less die size, 20% less ram @ 20% less power usage at stock lets not even talk about overclocking ok.  There is no competition.  And He hyped to 980ti levels of performance!

4 minutes ago, Deus Voltage said:

Remember when GPU's in the mid-early 2000's used to have huge percentage bumps in performance? True, they were basically space heaters, but the performance increase with each generation was substantial. Nowadays it is somewhere between 30 to 45%. We don't see huge increase anymore, mainly I would argue because Moor's law lost much of its steam. But there is a good argument to be made with regards to monopolistic practices as well. And yes, I do believe AMD would take advantage of consumers just as much as Nvidia would if they were well positioned.

 

On the same node, never been done before sorry just didn't happen.  That was why we had the slow down in performance increases.  The node didn't change, nV changed tactics took them time but they were able to figure out architectural changes that will accomplish what they were doing before.  This is where R&D came in, AMD doesn't have that, Adored should have figure that out.  It would have shot down his masterplan videos even before they began. 

[Valentyn]

12 minutes ago, Razor01 said:

 

 

Its not an locked die, the gp104 is a full die lol. Its not the biggest die in the family though.

 

That is what changed.

 

Keplar, that Titan was a totally different class of chip, it had HPC components and warranted that price.  Even years after that Titan's resale value was still as high as the MSRP or close to it.

 

nV saw the success of that line and rolled out the enthusiast level tier.

Never said it was locked, it's just the smaller cut down one.

 

The GTX 580 also had those components; though; just like the Kepler Titan. The Titan was just faster at FP64; but overall Fermi had decent FP 64 compute also. More so than most Kepler parts outside of the Titan line.

 

 

FP 64 performance

GTX 580 1/8 

GTX 680 1/24 

780Ti 1/24

Titan 1/3 

Titan X 1/32  

Titan Xp 1/32 

Titan V 1/2

 

NVIDIA introduced a new segment under a new name; but it already existed as the x80 in the past. Silly buggers like myself just craving top performance kept buying them anyway; fueling it all.

[Drak3]

2 minutes ago, Razor01 said:

Polaris was bad and still is, when comparing to a gtx 1060 a card that has 30% less flops, 25% less bandwidth, 25% less die size, 20% less ram @ 20% less power usage at stock lets not even talk about overclocking ok.  There is no competition.  And He hyped to 980ti levels of performance!

Comparison wise, it was always 980 for a single 480, 980Ti/1080 for Crossfire. Outside of gaming and CUDA, the 480 is quite better than the 1060.

[Mira Yurizaki]

6 minutes ago, Valentyn said:

-Snip-

The thing with FP64 performance is... It's useless in consumer applications. And I'd argue that 64-bit data types in general are hardly ever used in consumer applications. Unless you actually need that amount of precision or number range, you get no benefit.

 

So NVIDIA gimping FP64, at least for consumer cards, doesn't really affect any actual performance on applications consumers are likely to use. 

[Valentyn]

5 minutes ago, M.Yurizaki said:

The thing with FP64 performance is... It's useless in consumer applications. And I'd argue that 64-bit data types in general are hardly ever used in consumer applications. Unless you actually need that amount of precision or number range, you get no benefit.

 

So NVIDIA gimping FP64, at least for consumer cards, doesn't really affect any actual performance on applications consumers are likely to use. 

All true, but the Titan and previous full dies had all that anyway. Just the pricing and names changed; and in many cases prices moved up a slot.

Getting the big die now is €3100; where it use to be €500. 

So I can easily understand the argument that the GTX 1080 even at the moment is a "mid-range die"; as by the old naming and pricing schemes that die should be around €250 at the moment.

It's all down to competition not being there sadly. AMD rarely won even back then; but they were cheaper; cooler, and offered within 10% performance.
Now their top card is hotter, uses more power, and doesn't come close enough

 

It's also damn smart Business from NVIDIA; means they make a lot more money; and don't waste the top dies they can ship in Tesla and Quadros.

Even now the Titan V has less ROPs than the Quadro GV100; so it's still not the fully unlocked die also for prosumers

[Mira Yurizaki]

1 minute ago, Valentyn said:

Getting the big die now is €3100; where it use to be €500. 

So I can easily understand the argument that the GTX 1080 even at the moment is a "mid-range die"; as by the old naming and pricing schemes that die should be around €250 at the moment.

I understand that sentiment too, but I'll just keep pulling my "Core i7 is an entry level processor" card out. Even Xeons didn't have way more cores than the consumer version until recently-ish