RX490 Speculation - Vega 10, not Polaris.

Ieldra said:
Why do FE prices matter?
Click to expand...
Normally I would say they don't, but we can clearly see FE cards are the cheapest cards on Newegg. So, by being so damn persistent of keeping them around, they have skewed prices upwards. At this point, FE cards are probably the best "average" you have.
 
Daniel_Chang said:
Actually, my comparison was 100% apples to apples. The x60 lineup has targetted the mainstream since before it was the x60 lineup. Back when it was the GF 4 ti4200, replaced by the FX 5600/5700, followed by the 6600/7600/8600/9600 before we hit the more modern naming scheme, the GTX 260/460/560/660/760/960 all the way to the current 1060.

The x60 lineup has always been about maintaining roughly x60 pricing while mimicking the prior generation's x70 performance, give or take. The 1060 exceeds the 970's performance, but also exceeds the 960's price. Apologists like yourself follow Nvidia's narrative, and by doing so you condone their price hikes. You blame the retailers, but you're wrong there too! Nvidia set the pricing. It's $299/$449/$699 for FE models, and aside from the 1060, these cards have all launched at or above those prices.

Next up, you state that the 380x offers the same performance as the 480. Nope, not even close. The 380x was an in-between card, between the 380 and the 390. The 480 is a 380 replacement. Despite that, the 480 outperforms the 380x. In terms of performance, it's 390x > 480 > 390 > 380x > 380. Unless, of course, you are arguing that the 380x is also faster than the 390.

In summary, your post was essentially:

Click to expand...


So your definition of price gouging means it doesn't matter you're getting more performance for less money, rather the price went up because inter-generational model numbers say so? Ummm yeah....I'm done here.

Do yourself and quit @##@$#@ until your price : performance ratio increases.
 
DigitalGriffin said:
So your definition of price gouging means it doesn't matter you're getting more performance for less money, rather the price went up because inter-generational model numbers say so? Ummm yeah....I'm done here.

Do yourself and quit @##@$#@ until your price : performance ratio increases.
Click to expand...

I don't think he fully understands generational replacements and the correlation behind direct comparisons.

One does not compare a C6 Corvette to a Gen5 Camaro because they perform similarly. You compare a C6 Corvette against a C7 Corvette. The price and performance went up from generation to generation. Which is why we compare the 980 to the 1080.

The 1080 launched at $700. The 980 launched at $550. That's almost a 30% increase in price.

If we follow Daniel's logic if the 1180 is 40% faster than the 1080 then it's price should be, what, $980 at launch?

GTFO
 
DigitalGriffin said:
So your definition of price gouging means it doesn't matter you're getting more performance for less money, rather the price went up because inter-generational model numbers say so? Ummm yeah....I'm done here.

Do yourself and quit @##@$#@ until your price : performance ratio increases.
Click to expand...


So, rather than address my post, you'd rather put words into my mouth and address that? Yes, I agree, you're done here.

Riccochet said:
I don't think he fully understands generational replacements and the correlation behind direct comparisons.

One does not compare a C6 Corvette to a Gen5 Camaro because they perform similarly. You compare a C6 Corvette against a C7 Corvette. The price and performance went up from generation to generation. Which is why we compare the 980 to the 1080.

The 1080 launched at $700. The 980 launched at $550. That's almost a 30% increase in price.

If we follow Daniel's logic if the 1180 is 40% faster than the 1080 then it's price should be, what, $980 at launch?

GTFO
Click to expand...

Why is this content addressed at me? I think you confused DigitalGriggin and me. Let me break it down.

Riccochet said:
One does not compare a C6 Corvette to a Gen5 Camaro because they perform similarly. You compare a C6 Corvette against a C7 Corvette. The price and performance went up from generation to generation. Which is why we compare the 980 to the 1080.
Click to expand...

Agreed. I was comparing 960 to 1060, 970 to 1070, and 980 to 1080. DG was comparing 1070 to 980ti, and 1060 to 980. You got us confused.

Riccochet said:
The 1080 launched at $700. The 980 launched at $550. That's almost a 30% increase in price.
Click to expand...

My point exactly!

Riccochet said:
If we follow Daniel's logic if the 1180 is 40% faster than the 1080 then it's price should be, what, $980 at launch?
Click to expand...

Nope, that's DG's logic. My point was that the 1080 should launch around the 980's price, and thus so should the 1180 eventually. The 1070 should be around the 970's price, but instead was $120 more.

Yea, you definitely got us confused.
 
So is there any hope this thing comes out this year? It seems like the hbm2 memory is the main thing preventing it from being released.
 
Pricing for this gen has been all messed up because of the FE cards, gen to gen 980 to 1080 there should have been only a 49 buck difference that didn't happen as AIB partners took advantage of the FE pricing and priced their stock versions above MSRP. EVGA has a $620 stock card, I'm sure you will see cards like that from PNY and others at similar prices in the future.

Right now you can get out of the box overclocked cards for less than the FE cost, so unless you are comparing FE prices and using those as base lines, assumptions like that aren't really where the pricing is at since there was no such thing as FE card pricing in the past. Now if you want to compare a stock $550 980 vs a what then? I would say the best way to go is compare it to stock 1080, or overclocked 980 to a overclocked 1080 of the same overclock bracket. In EVGA's line if you look at a SC from the 980 look at a SC for the 1080.

That $49 price increase from $550 to $600, 980 to 1080 respectively should have been able to cover the margins of the cost increase of the process change, but AIB, retailers are selling for even more than that because they can.

nV doesn't getting the entire cost of the cards, that $49 extra is still split among its board partner and then the retailer.

xSneak said:
So is there any hope this thing comes out this year? It seems like the hbm2 memory is the main thing preventing it from being released.
Click to expand...

Back to topic lol, little hope, I would say the earliest would be some time end of q4.

HBM2 mass production from Hynix is Q3 this year, so given that AMD possible can go into production and release boards with sampled HBM2 would be the only way to get it out before Q4 which I don't think they will do, cause there will be a huge shortage of cards and that won't disappear no matter what till HBM2 manufacturing ramps up.
 
Last edited:
razor1 said:
Pricing for this gen has been all messed up because of the FE cards, gen to gen 980 to 1080 there should have been only a 49 buck difference that didn't happen as AIB partners took advantage of the FE pricing and priced their stock versions above MSRP. EVGA has a $620 stock card, I'm sure you will see cards like that from PNY and others at similar prices in the future.

Right now you can get out of the box overclocked cards for less than the FE cost, so unless you are comparing FE prices and using those as base lines, assumptions like that aren't really where the pricing is at since there was no such thing as FE card pricing in the past. Now if you want to compare a stock $550 980 vs a what then? I would say the best way to go is compare it to stock 1080, or overclocked 980 to a overclocked 1080 of the same overclock bracket. In EVGA's line if you look at a SC from the 980 look at a SC for the 1080.

That $49 price increase from $550 to $600, 980 to 1080 respectively should have been able to cover the margins of the cost increase of the process change, but AIB, retailers are selling for even more than that because they can.

nV doesn't getting the entire cost of the cards, that $49 extra is still split among its board partner and then the retailer.



Back to topic lol, little hope, I would say the earliest would be some time end of q4.

HBM2 mass production from Hynix is Q3 this year, so given that AMD possible can go into production and release boards with sampled HBM2 would be the only way to get it out before Q4 which I don't think they will do, cause there will be a huge shortage of cards and that won't disappear no matter what till HBM2 manufacturing ramps up.
Click to expand...


Exactly what I have said since FE launched. I was predicting that the AIB will have no reason to price MSRP since Nvidia's reference card is 699. So instead of playing with paper price they are playing with actual reference price 699. That is where nvidia sort of gouged the price up in a sense. AIBs will justify this all day long until nvidia gets rid of this founders edition crap. Reference cards were always MSRP prices, until nvidia made founders edition a 100 increase and made them reference cards and announced a paper msrp. I guess when you are on the top you can justify anything and that is whats going on. They will milk it until AMD has something to compete on the top end, and when vega comes out watch founders edition drop down to 599 and now we have all the cards between 500-600 in. They are milking it for the time being and we can hate em but we can't blame em. After all cant fault them for making profit thats what they are in business for.
 
Daniel_Chang said:
So, rather than address my post, you'd rather put words into my mouth and address that? Yes, I agree, you're done here.
Click to expand...

Look are you or are you not going to agree you get more performance for less money then the previous generation? You have ZERO right to complain if they are giving you more for less. Especially when there isn't competition. And this you are reluctant to admit because it would destroy your argument.

There's being a fan and then there is fanboy-ism. I've owned 4 AMD cards (Rage, 8500, 9800 pro, 7970), 3 NVIDIA cards (TNT2, MX200, ?460?). I buy that which gives me the best bang for my buck. Anybody who does otherwise isn't making the best purchase decision.

Nobody wants more than me to see AMD give NVIDIA a kick in their complacency. And when they do, I'll step up to the plate and buy one. But there is certainly no price gouging here. Your price to performance ratio has gone down and no matter how you try to spin it, that is a good thing. And with the 1050 comes out it will likely again give you more for less money compared to AMD. (Overall) There is ZERO gouging other than what the AIB vendors are doing on their pricing and BOTH sides are guilty of this.
 
This generation really feels like the 6xx series, except I doubt AMD has anything in Vega to counter, unless its some half-assed science project like the Fury line was. So we are faced with gouging on both sides of the fence over essentially middle of the road and low end parts.
 
DigitalGriffin said:
Look are you or are you not going to agree you get more performance for less money then the previous generation? You have ZERO right to complain if they are giving you more for less. Especially when there isn't competition. And this you are reluctant to admit because it would destroy your argument.

There's being a fan and then there is fanboy-ism. I've owned 4 AMD cards (Rage, 8500, 9800 pro, 7970), 3 NVIDIA cards (TNT2, MX200, ?460?). I buy that which gives me the best bang for my buck. Anybody who does otherwise isn't making the best purchase decision.

Nobody wants more than me to see AMD give NVIDIA a kick in their complacency. And when they do, I'll step up to the plate and buy one. But there is certainly no price gouging here. Your price to performance ratio has gone down and no matter how you try to spin it, that is a good thing. And with the 1050 comes out it will likely again give you more for less money compared to AMD. (Overall) There is ZERO gouging other than what the AIB vendors are doing on their pricing and BOTH sides are guilty of this.
Click to expand...
Just to be devils advocate... Explain the 980Ti from the 780Ti... Huge price difference in favor of the 980Ti. I mean I get slightly higher prices this generation due to new more expensive nodes, new architecture and so on but it seems a bit excessive. Based on the 1080 and Titan X(P) prices then the 1080Ti (if it ever sees the light of day) will be what? $899?
 
JustReason said:
Just to be devils advocate... Explain the 980Ti from the 780Ti... Huge price difference in favor of the 980Ti. I mean I get slightly higher prices this generation due to new more expensive nodes, new architecture and so on but it seems a bit excessive. Based on the 1080 and Titan X(P) prices then the 1080Ti (if it ever sees the light of day) will be what? $899?
Click to expand...

The entire 7-series was an outlier (I hope). It had the prices raised at nearly every segment.

GTX 460 - $199/$229
GTX 560 - $199 (560ti at $249)
GTX 660 - $229 (660ti at $299)
GTX 760 - $249
GTX 960 - $199/229

GTX 470 - $349
GTX 570 - $349
GTX 670 - $399
GTX 770 - $399
GTX 970 - $329

GTX 480 - $499
GTX 580 - $499
GTX 680 - $499
GTX 780 - $649 (later reduced when the 780ti came out)
GTX 980 - $549


GTX 480ti - didn't exist
GTX 580ti - didn't exist (GTX 590 was $699)
GTX 680ti - didn't exist (GTX 690 was $999)
GTX 780ti - $699
GTX 980ti - $649

You can see trends here.

The x60 lineup was typically $199-$229. The 760 raised this to $249. The 960 reduced this back to the norm, and then the 1060 brought it back up. I suspect that the supposed 3GB 1060 will hit that $199 price point, once again leaving the 7-series as the outlier.

The x70 series has traditionally been $349-$399. The 1070 hitting $449 is an outlier, but seems worse in the face of the $329 970. In reality, once cards reach availability, the 1070 should fall within the norms.

The x80 is interesting. It was traditionally a $499 part, but the 780 launched at a ridiculous $649 supposedly due to ZERO competition. When AMD countered with the 290x, the 780 magically hit the $499 price again (and NV countered by releasing a 780ti at the exorbitant $699 price). You can argue that these were the Titan's prior to the actual Titan - no competition from AMD so priced accordingly. The 980 launched at $549, and the 1080 of course again went high ($699 official, $599 paper).

Basically, prices seem to be consistent when there's competition, partly why I think the 1060 launched at MSRP unlike the 1070/1080. The price experimentation in the x70/x80 series is WHY we need AMD to get back into the game.
 
This has been the case, Nvidia will charge more when there is no competition. That is reality of any business if there is no direct competition expect to pay more. AMD needs to make Vega around 400-450. If they can knock out gtx 1080 performance for that price than that will be a disruptive product like the older gens.

This is why Vega needs to hit the spot. If it doesn't say good by to low prices and Nvidia will keep the founders edition crap going. This was a major slap in the face to MSRP that I have eve seen. Never seen a paper price of 599 and then a company launching their reference product at 699.99. Truly one of a kind. I think we will see another 50-100 dollar tax on next gen if AMD doesn't get their shit together.

They seriously even had the balls to charge 50 premium for gtx 1060. lol.
 
NKD said:
AMD needs to make Vega around 400-450. If they can knock out gtx 1080 performance for that price than that will be a disruptive product like the older gens.
Click to expand...

That won't be disruptive. If AMD offers GTX 1080 performance at $400-$450, then Nvidia simply drops the 1080 price to compete, and releases a 1080ti. AMD needs to release a product where NV can't just compete with existing or a held back product. NV can afford to cut margins if it hurts AMD. AMD cannot afford to keep cutting margins.
 
Daniel_Chang said:
That won't be disruptive. If AMD offers GTX 1080 performance at $400-$450, then Nvidia simply drops the 1080 price to compete, and releases a 1080ti. AMD needs to release a product where NV can't just compete with existing or a held back product. NV can afford to cut margins if it hurts AMD. AMD cannot afford to keep cutting margins.
Click to expand...

Exactly this. They'd better hurry the fuck up so they can release at least 5 months prior to Volta. However rumours of interposer issues on their GP100 was certainly interesting. The Pitan was telling to me in this way - perhaps that was the 'answer' they had planned (as a Ti) for Vega which was delayed, assuming Vega would beat 1080?
 
cageymaru said:
AMD VEGA (Radeon RX 490) Launch Imminent ?
AMD VEGA (Radeon RX 490) Launch Imminent ?

Chris Hook This is the Vega launch venue. Shh, don't tell the press.....:)

Reddit thread.

Looks like he is trolling to me. What do you think? :)
Click to expand...

That would just be my luck. I thought we weren't getting anything from AMD until next year. So I bought myself an Nvidia card and now they mention Vega. Good one.
 
Maddness said:
That would just be my luck. I thought we weren't getting anything from AMD until next year. So I bought myself an Nvidia card and now they mention Vega. Good one.
Click to expand...

That's AMD. Par for the course. I've waited too many times for AMD just to be disappointed. Anymore I just buy the most logical when I need it, AMD or nVidia. Although I might wait for nVidia if launch is imminent.
 
It's weird, but parts of that pic look rendered to me, like its an in-game scene rather than a random picture.
 
exlink said:
I'd be happy with an October release for a card from AMD that provides 1080 performance. Itching for an upgrade.
Click to expand...


Could very well be possible. Don't believe some of the trolls that misread the roadmap on purpose.


 
Rvenger said:
Could very well be possible. Don't believe some of the trolls that misread the roadmap on purpose.


Click to expand...
Only down side is Vega will not match 1080 in dx11/open gl games if i had to guess.....But i bet dx12 and vulkan it does just fine. Fuck it i like the newer api's anyway!
 
primetime said:
Only down side is Vega will not match 1080 in dx11/open gl games if i had to guess.....But i bet dx12 and vulkan it does just fine. Fuck it i like the newer api's anyway!
Click to expand...

I think that people who upgrade annually care more for DX11/OGL performance than DX12/Vulkan. It's going to be another year or two before those APIs make up the majority of new releases. By then I 100% guarantee you that I won't be using my current GPU, and many on this board feel the same way.
 
Fun little fact, that "venue" that was posted is the same factory that was used when AMD and Valve announced their partnership for Half-Life 2...
 
So the day I go to buy a 1070, Newegg ends the sale early, and now this... Keeping waiting and possibly see price drops on 1070's or......
 
Daniel_Chang said:
And herein lies the problem. If AMD launches a card that is faster than the 1080, why does it also have to be $100 cheaper? "AMD, I see that your card is faster, but hey it's the same price so I'll be taking the slower Nvidia card. Better luck next time ol' chap."



First off, there are zero cards (1070/1080) at the prices your quoted. Just because Nvidia announced an MSRP ($379 for the 1070, btw) doesn't mean someone is going to adhere to it.

Secondly, you're comparing to the wrong cards. By comparing the 1060 to the 970/980 you're doing what Nvidia hoped you would do, which allows them to get away with jacking up prices.

The GTX 960 was $199/$229. The 1060 that replaces it is $249+.
The GTX 970 was $329. The GTX 1070 that replaces it launched at $449 and has yet to see a model sell below $399.
The GTX 980 was $549. The GTX 1080 that replaces it launched at $699 and has yet to see a model sell below $629 (EVGA briefly teased a $609 model but never sold it);

Nvidia jacked the prices of those cards $20-$50, $120, and $150 at launch, respectively. And the fact that you're arguing that it's a good thing just shows what camp you're in.
Click to expand...

Here in Vietnam I bought my Palit 1070 for $390, discounting VAT (10%) and import tax (10%). With a custom cooler. And 3 year warranty.

::Palit Products - GeForce® GTX 1070 Dual ::

Just saying...
 
Maddness said:
That would just be my luck. I thought we weren't getting anything from AMD until next year. So I bought myself an Nvidia card and now they mention Vega. Good one.
Click to expand...

I would not worry about it to much, these days not buying at launch might be better considering that prices are higher and options to buy something with decent cooler is usually a few weeks later anyway.
 
Pieter3dnow said:
I would not worry about it to much, these days not buying at launch might be better considering that prices are higher and options to buy something with decent cooler is usually a few weeks later anyway.
Click to expand...

Yeah, you do make a lot of sense there. Plus the 1080 was on sale for 12% off.
 
Rvenger said:
New IP version may have better performance.
Click to expand...

From what I understand, the problem was not so much the IP (although it could use improvement...there are a LOT of transistors in there.) There also seems to be a problem in the 14nm node size which has a lot of transistor leakage despite FinFET.
 
Last edited by a moderator:
DigitalGriffin said:
From what I understand, the problem was not so much the IP (although it could use improvement...there are a LOT of transistors in there.) The problem likes in the 14nm node size which has a lot of transistor leakage despite FinFET.
Click to expand...

I think there is a reason they went with ip 9.0. May be newer libraries for higher clocks or what not. I guess it was probably needed because there are certainly things holding the architecture back and one is clock speeds.
 
I don't think the block IP has anything to do with anything, GCN 1.1 to 1.2, block IP changed and yet we didn't really see any change that would constitute a change in the way transistors and node were addressed.

All we saw were feature changes like DC.
 
Vega sucks! Zen Sucks! AMD SUCKS! LOL.

But wait they are not out yet!
 
NKD said:
Vega sucks! Zen Sucks! AMD SUCKS! LOL.

But wait they are not out yet!
Click to expand...


Well from GCN 1.1 to 1.2, if anything power consumption vs performance got worse, just look at Tonga ;) (can't really talk about Fiji since it used HBM)

This is why I say, putting hopes on the IP block change is not good to draw speculation from.
 
Last edited:
razor1 said:
Well from GCN 1.1 to 1.2, if anything power consumption vs performance got worse, just look at Tonga ;) (can't really talk about Fiji since it used HBM)

This is why I say, putting hopes on the IP block change is not good draw speculation from.
Click to expand...

I think improvements need to be made on both fronts I would imagine. Stopping leakage would help with excess current. Also, RX480 has an obscene amount of transistors : GFLOPS/MHz ratio. While miners love FP32 support, how much of that market would they lose if AMD dropped some of that FP32 support off to gain MHz and higher IPC? Great FP32 is awesome, but not if you can't put it to use.

I'm hoping Vega will turn things around a bit. But the real telling tail is if AMD only shows benchmarks which are known for heavily favoring AMD (ie: AoTS). If they pull that again, I know they are in trouble.
 
NKD said:
I think there is a reason they went with ip 9.0. May be newer libraries for higher clocks or what not. I guess it was probably needed because there are certainly things holding the architecture back and one is clock speeds.
Click to expand...

Could be different libraries but new IP block revision doesn't necessarily mean libraries are changed

DigitalGriffin said:
I think improvements need to be made on both fronts I would imagine. Stopping leakage would help with excess current. Also, RX480 has an obscene amount of transistors : GFLOPS/MHz ratio. While miners love FP32 support, how much of that market would they lose if AMD dropped some of that FP32 support off to gain MHz and higher IPC? Great FP32 is awesome, but not if you can't put it to use.

I'm hoping Vega will turn things around a bit. But the real telling tail is if AMD only shows benchmarks which are known for heavily favoring AMD (ie: AoTS). If they pull that again, I know they are in trouble.
Click to expand...


People don't often consider this but having fewer ALUs on a chip running overall higher clocks is good because the whole chip is running at the higher clock, so you have all the units (geo, tex, pix) running at higher clocks, making the design more balanced as a whole.

Having a huge number of ALUs and the chip clocking low means all auxiliary units are slower
 
DigitalGriffin said:
I think improvements need to be made on both fronts I would imagine. Stopping leakage would help with excess current. Also, RX480 has an obscene amount of transistors : GFLOPS/MHz ratio. While miners love FP32 support, how much of that market would they lose if AMD dropped some of that FP32 support off to gain MHz and higher IPC? Great FP32 is awesome, but not if you can't put it to use.

I'm hoping Vega will turn things around a bit. But the real telling tail is if AMD only shows benchmarks which are known for heavily favoring AMD (ie: AoTS). If they pull that again, I know they are in trouble.
Click to expand...

I don't know even know if just droppigng FP32 support will do the trick lol,

but yeah node issues to drop leakage amounts and base transistor layout to increase frequencies without crazy power draw or hitting a frequency wall.

Now Vega has an extra 8 months of design work over Polaris maybe, that doesn't seem be to enough time to much, look at Pascal, nV had what close to 2 years of time to do what they did to modify and change Maxwell 2 to what Pascal could do, that is a solid 1 generation to make those types of changes even when they already had a good foundation to start from in Maxwell 2.
 
razor1 said:
I don't know even know if just droppigng FP32 support will do the trick lol,

but yeah node issues to drop leakage amounts and base transistor layout to increase frequencies without crazy power draw or hitting a frequency wall.
Click to expand...

Didn't the mention Polaris using their new power gating features ? How the hell did that go so wrong
 
You must log in or register to reply here.
Back
Top