Are we assuming Vega is the 490? Or is it possibly the 500 series?

J

jamesgalb

Gawd
Joined
Feb 11, 2015
Messages
565
Any information on this?

Im excited to see what is to come. If AMD could push a Tonga architecture to Fury levels with 14nm, various architecture improvements, clock boost, and more/faster vram, then I cant wait to see what happens when they do it with Vega which I assume will be a rework of the Fiji architecture with the same improvements and 8/16gb HBM2...
 
idk what other people are assuming but yeah at this point the 490 and 490X are probably Vega.
 
That adoredtv youtuber had an insane theory, that because the slide for the cf 480 vs 1080 chips

AMD-RX-480-CF-1-1200x658.jpg


Showed better performance AND efficiency, that the two 480s were each drawing a maximum of around 90w (since a 1080 might be drawing around 170w and if the cf was more efficient EVEN FOR CF, then each individual card must be using much less than the max allowed power usage of 150w)

His theory was that amd tuned the 480 to be an ultra low power draw card to push back against their negative image as power hungry, but that amd would release a less power efficient polaris variant, perhaps with the full 40 CUs that would typically run closer to the 150w power draw limit, pushing up clocks being much closer to the 1070 for around 300 dollars.

If an 8GB 480 will really cost only around 230, and Lisa Su was truthful in talking about a range for polaris from 100 to 300 dollars, there seems like there ought to be something else higher up in the ranks before vega hits.


Now, there are a LOT of assumptions in these theories, a LOT, I don't know that the 480 really draws that little power, but who knows. You can see his insane theory here:

 
For above^^
Buying into marketing never turns out well, after all it could mean "cost efficiency"!
 
Lisa Su said they would release Polaris GPUs filling the $100 to $300 tiers, common sense dictates there will be at least two more: One around $300, and one or two below $200.
Also when was the last time AMD released one SKU from a GPU? I believe even Raja himself said there were more coming this summer.

There's still room for a potential 480X, 490/X, and the 470/X. I don't need a YouTube video to tell me the entire Polaris line-up isn't solely the RX 480, and I don't think anybody is that stupid.

AdoredTV made that video to ensure the 1070 gets pandered against, gotta make sure people are aware AMD is going to release something faster than the 480 because they were too incompetent to announce it themselves.
 
I would have loved for his channel to exist last year during Pirate Islands hype. He would have made 20 videos over the course of 6 months with all sorts of detailed analysis about how the Fury X was going to destroy the 980, then oops the Titan X comes out, but don't worry that's a $1000 part. Okay then the 980 Ti comes out but don't worry Fury X uses 8 GB HBM, and look this AMD marketing guy said "it's the fastest GPU in the world", it's going to be amazing, don't worry.

He's the one guy carrying a megaphone in a room full of people who already learned AMD OVERHYPES THEIR PRODUCTS. His last several videos about Pascal were effectively: "Don't believe Nvidia's 1080 marketing hype" and "Don't believe Nvidia's 1070 marketing hype" and then he spends the rest of his videos going through AMD's marketing material with a fine-tooth comb. I just can't take it seriously.

Even better I would have loved to see his pre-Bulldozer videos.

If the "490" is going to trade blows with the 1070 then perhaps AMD should have focused on that rather than how efficient their "75W" mid-range Polaris chip is. We shouldn't need AdoredTV to fill AMD's marketing gaps.
 
TaintedSquirrel said:
Lisa Su said they would release Polaris GPUs filling the $100 to $300 tiers, common sense dictates there will be at least two more: One around $300, and one or two below $200.
Also when was the last time AMD released one SKU from a GPU? I believe even Raja himself said there were more coming this summer.

There's still room for a potential 480X, 490/X, and the 470/X. I don't need a YouTube video to tell me the entire Polaris line-up isn't solely the RX 480, and I don't think anybody is that stupid.

AdoredTV made that video to ensure the 1070 gets pandered against, gotta make sure people are aware AMD is going to release something faster than the 480 because they were too incompetent to announce it themselves.
Click to expand...
To be honest, i have my doubts we will actually see higher SKUs than 480 in retail until autumn, no matter what Lisa Su said (it's not like she was saying that to investors), because 480 really has that Tonga smell to it. Heck, it even uses same underlying command set as Tonga (source: Linux OpenGL stuff)!

TaintedSquirrel said:
I would have loved for his channel to exist last year during Pirate Islands hype. He would have made 20 videos over the course of 6 months with all sorts of detailed analysis about how the Fury X was going to destroy the 980, then oops the Titan X comes out, but don't worry that's a $1000 part. Okay then the 980 Ti comes out but don't worry Fury X uses 8 GB HBM, and look this AMD marketing guy said "it's the fastest GPU in the world", it's going to be amazing, don't worry.

He's the one guy carrying a megaphone in a room full of people who already learned AMD OVERHYPES THEIR PRODUCTS. His last several videos about Pascal were effectively: "Don't believe Nvidia's 1080 marketing hype" and "Don't believe Nvidia's 1070 marketing hype" and then he spends the rest of his videos going through AMD's marketing material with a fine-tooth comb. I just can't take it seriously.

Even better I would have loved to see his pre-Bulldozer videos.

If the "490" is going to trade blows with the 1070 then perhaps AMD should have focused on that rather than how efficient their "75W" mid-range Polaris chip is. We shouldn't need AdoredTV to fill AMD's marketing gaps.
Click to expand...
I mean, looks like AMD this time legitimately tried to keep expectations low, according to Sky they were briefed with Polaris being mainstream product as far back as November '15.

P. S. Damn, the Reddit hype train power, the guy went from 100500th letsplayer to AMD's hype guy in literally a week.
 
As I said a few days ago, the $300 970 represents 50% more of the market than the $200 960 (according to Steam).
AMD would have been much better served to rollout the $300 480X / 490 first, since it will probably be 980 Ti / Fury X in performance. And being up to $100 cheaper than the 1070 puts it in a perfect spot to eat into Nvidia's sales.

So there's something happening here: Either the 480 is being overhyped, the 480X/490 is nowhere near ready, or AMD's marketing department made a serious mistake. AdoredTV is just performing damage control to buy time downplaying the GTX 1070 until AMD has a response. "The RX 490 will launch in July at $300 and compete with the GTX 1070" is a great way to stop people from buying GTX 1070's. It's too bad AMD didn't say that.

We'll probably get more info about the rest of the Polaris chips on June 29th when all the NDA's end.
 
As for the naming scheme, at 2560 SP the next chip would be better suited as a "480X". If they end up calling it "490" it will just be a much smaller gap than the 380 --> 390 or 280 --> 290.

If it's 2816+ SP then all bets are off, because nobody saw that coming.
 
TaintedSquirrel said:
I would have loved for his channel to exist last year during Pirate Islands hype. He would have made 20 videos over the course of 6 months with all sorts of detailed analysis about how the Fury X was going to destroy the 980, then oops the Titan X comes out, but don't worry that's a $1000 part. Okay then the 980 Ti comes out but don't worry Fury X uses 8 GB HBM, and look this AMD marketing guy said "it's the fastest GPU in the world", it's going to be amazing, don't worry.

He's the one guy carrying a megaphone in a room full of people who already learned AMD OVERHYPES THEIR PRODUCTS. His last several videos about Pascal were effectively: "Don't believe Nvidia's 1080 marketing hype" and "Don't believe Nvidia's 1070 marketing hype" and then he spends the rest of his videos going through AMD's marketing material with a fine-tooth comb. I just can't take it seriously.

Even better I would have loved to see his pre-Bulldozer videos.

If the "490" is going to trade blows with the 1070 then perhaps AMD should have focused on that rather than how efficient their "75W" mid-range Polaris chip is. We shouldn't need AdoredTV to fill AMD's marketing gaps.
Click to expand...

For the sake of devil's advocate though.....how interesting would it be if he's right though. 75watts, wowza.
 
Vercinaigh said:
For the sake of devil's advocate though.....how interesting would it be if he's right though. 75watts, wowza.
Click to expand...
That's actually possible though, we're already pretty sure it's below 100W.
Maybe the 150W TDP is throwing people off.
 
We all know what happened. I think AMD always wanted to do a hard launch with these cards with the least amount of info leaked. Nvidia paper launched Pascal earlier and gave out all the specs. I think AMD had to do something but they still gave potential specs and max power draw. We are still guessing at specs. They gave the price for the lowest model of RX 480 4gb and not even clock speeds, just expect it to be greater than 5 Tflop and not consume more than 150w lol.

I think its obvious there will be another version. We know there is 8gb version we just dont know where they clock it or if there will be one with full 40Cu's at 299.
 
TaintedSquirrel said:
That's actually possible though, we're already pretty sure it's below 100W.
Maybe the 150W TDP is throwing people off.
Click to expand...

Well you are right lol. They said >5tflop and 150w tdp. So they pretty much said we are not going to tell you exact clocks you gotta wait for that but no matter where we clock it it wont use more than 150w lol. So yea I don't expect this thing to use a whole lot of wattage if its clocked at 5tflops.
 
its highly unlikely we will see a higher version at launch, not until EOL of Fury parts are done. There is no price point that isn't covered right now all the way up to performance segment. Fury parts have to drop in price, which they have to go lower than the current 1070, so Fury parts will cover the 300 buck range.


And hell no rx480 will not be lower than 100 watts, if anything it will be closer to 150 watts than 100, if it was anything lower than 115 watts, it would have a peft/watt similarity to a 1070 or even higher.

They would have used that to their marketing advantage, they did not that tells you they can't match the perf/watt advantage of the current Pascal's.
 
razor1 said:
its highly unlikely we will see a higher version at launch, not until EOL of Fury parts are done. There is no price point that isn't covered right now all the way up to performance segment. Fury parts have to drop in price, which they have to go lower than the current 1070, so Fury parts will cover the 300 buck range.


And hell no rx480 will not be lower than 100 watts, if anything it will be closer to 150 watts than 100, if it was anything lower than 115 watts, it would have a peft/watt similarity to a 1070 or even higher.

They would have used that to their marketing advantage, they did not that tells you they can't match the perf/watt advantage of the current Pascal's.
Click to expand...
Their confusing Ashes slide did mention better efficiency.
 
That slide Raja explained better efficiency in the context of GPU utilization based on mGPU's actual inefficiencies lol, its a marketing ploy of shifting numbers around to show something that is not there.

NKD, we are not guessing at specs anymore, enough of the leaks after the launch and what was stated at the launch, matches up well with many things AMD stated in the past and the more plausible rumors that aligned with AMD marketed prior to launch, all of that stuff are pretty much confirmed. Really the only questions left, what are they doing to do with the Fury line and what are their next steps for Vega.

ALU, clocks, Tflops, we know where this card will land.

Actually another question too, what clocks on GDDR5 memory are they going to use. That will give us the last bit of information where its going to land and if there will be a possibility of a higher clocked rx480 or more units on a rx480 coming out later. Gotta remember if this card doesn't go up against a 1070 directly, why would they need the same bandwidth as the 1070? Does it not have the same compression capabilities so a lower performant card needs the same amount of bandwidth as the 1070 or it will be handicapped by it?
 
Last edited:
AMD launched their mainstream cards first. This tells me a few things.

They recognized that the mainstream makes them the most money.
They need to get the same version of GCN for all ranges of cards. No more GCN 2.0, 3.0, and 4.0 coexisting. Easier to write drivers and release updates for one architecture. Plus since the earlier versions of GCN share tech with the newest version, those customers can get boosts and keep the people who don't upgrade happy. Ask Nvidia if this works well.
They probably don't have a vast inventory of Fury, Fury X, and Nano cards on the market since the 390 series was what got them market share. Tax write off.
They no longer care about super high end sales. If I were Lisa Su I would target Vega at the $500 market and let Nvidia have the 501 - infinity market to themselves. Target something between the 1080 - 1080ti and skip the rest. As long as it is cheaper than the card that is faster than it who cares?
 
point 1, mainstream is not the most anymore
point 2, GCN 1.0, 1.1, 1.2, all were out at the same time, why would you think there can't be overlap with GCN 1.3?
point 3, you can't take a tax write off unless you are actually making money lol, so taking a write off would kinda be useless unless you are trying to make it the bottom line look worse than it would be so the rebound would look better.
point 4, and you believe that after they are putting two GPU's on a card to go up against the 1080? And that is just a performance card not the enthusiast. They know the importance of halo products, they just can't do anything about it immediately. Near future possibly.

Come on man think it through.
 
razor1 said:
That slide Raja explained better efficiency in the context of GPU utilization based on mGPU's actual inefficiencies lol, its a marketing ploy of shifting numbers around to show something that is not there.

NKD, we are not guessing at specs anymore, enough of the leaks after the launch and what was stated at the launch, matches up well with many things AMD stated in the past and the more plausible rumors that aligned with AMD marketed prior to launch, all of that stuff are pretty much confirmed. Really the only questions left, what are they doing to do with the Fury line and what are their next steps for Vega.

ALU, clocks, Tflops, we know where this card will land.

Actually another question too, what clocks on GDDR5 memory are they going to use. That will give us the last bit of information where its going to land and if there will be a possibility of a higher clocked rx480 or more units on a rx480 coming out later. Gotta remember if this card doesn't go up against a 1070 directly, why would they need the same bandwidth as the 1070? Does it not have the same compression capabilities so a lower performant card needs the same amount of bandwidth as the 1070 or it will be handicapped by it?
Click to expand...
Well, Kyle claims the 8Ghz effective clock on review samples and 7Ghz on retail cards. Obviously it could be a Polaris 10 version of 3.5/4 joke, but i don't think there's any point not having GDDR5 clocked below 7Ghz nowadays.

By the way, looks like Fiji/Tonga front end deficiency was completely intentional, if i understand Bridgman right. And that rises a red herring for P10's front end abilities for me.

razor1 said:
point 1, mainstream is not the most anymore
point 2, GCN 1.0, 1.1, 1.2, 1.3 all were out at the same time, why would you think there can't be overlap with GCN 1.4?
point 3, you can't take a tax write off unless you are actually making money lol
point 4, and you believe that after they are putting two GPU's on a card to go up against the 1080?

Come on man think it through.
Click to expand...
GCN 1.3 is Polaris, man.
 
lolfail9001 said:
Well, Kyle claims the 8Ghz effective clock on review samples and 7Ghz on retail cards. Obviously it could be a Polaris 10 version of 3.5/4 joke, but i don't think there's any point not having GDDR5 clocked below 7Ghz nowadays.

By the way, looks like Fiji/Tonga front end deficiency was completely intentional, if i understand Bridgman right. And that rises a red herring for P10's front end abilities for me.


GCN 1.3 is Polaris, man.
Click to expand...
ah sorry was just editing lol

Can ya link me to the Bridgman info?
 
razor1 said:
ALU, clocks, Tflops, we know where this card will land.
Click to expand...

Yes only from rumors and sisoft and 3dmark benches. If we are to believe that the card is closer to 5.8-5.9 Tflop. So yea then I ready to believe it will creep up to 130-140, but most people seem to not grasp that and are assuming its hitting 150W at 5tflop and claiming it is doom and gloom. We know it can't go above 150w since it only has 1 6pin connector. So It has to be below that no matter where they clock it, I doubt they are going to make it 150 exactly, lol. I agree with the rumors specs and wattage but the slide seems to be getting blown up as this card is the worst performance per watt than 1070. I believe it to use around 130w cuz we have to believe it will stay under the 150w mark since thats the limit here.
 
well the perf/watt is not as good as pascal, otherwise they would have shown that or talked about perf/watt in current games, which they were talking about before, now they have mysteriously omitted that from their presentation and talked about pref/watt with VR......

Marketing only changes strategies when something is better or worse than originally expected (in this case Pascal came out), they prefer not to change if at all possible because they want to keep a unified message. Confusing end consumers is never good a thing.
 
razor1 said:
well the perf/watt is not as good as pascal, otherwise they would have shown that or talked about perf/watt in current games, which they were talking about before, now they have mysteriously omitted that from their presentation and talked about pref/watt with VR......

Marketing only changes strategies when something is better or worse than originally expected (in this case Pascal came out), they prefer not to change if at all possible because they want to keep a unified message. Confusing end consumers is never good a thing.
Click to expand...

They talked about perf/watt at the Computex event.
 
variant said:
They talked about perf/watt at the Computex event.
Click to expand...

where?

if you are talking about 1.7 and 2.8 I just told you what the 2.8 was for VR (which could be in limited situation) Raja stated right after showing that slide and before going to next slide he hasn't talked about Gaming yet, in relationship with the perf/watt numbers.

And this is different than what they were talking about before.

first off they were talking about 2.5 (hawaii/grananda )than 2.0 (tonga), now its totally weird with 2.8 with VR.

if you want to guess some more, nV stated 3 times perf/watt in VR comparing to their older cards. See what I'm getting at?
 
Last edited:
Brackle said:
I have one thing to say.

Please no rebrands!

I hate that both companies do it!
Click to expand...


well sometimes its unavoidable, and new chips aren't needed all the time as long as they are competitive.
 
razor1 said:
where?

if you are talking about 1.7 and 2.8 I just told you what the 2.8 was for VR (which could be in limited situation) Raja stated right after showing that slide and before going to next slide he hasn't talked about Gaming yet, in relationship with the perf/watt numbers.

And this is different than what they were talking about before.

first off they were talking about 2.5 (hawaii/grananda )than 2.0 (tonga), now its totally weird with 2.8 with VR.

if you want to guess some more, nV stated 3 times perf/watt in VR comparing to their older cards. See what I'm getting at?
Click to expand...

So the 1.7 he said for FinFet has only to do with VR?
 
1.7 is lower than what Samsung, AMD, and GF stated 14nm would give. 60% drop in power usage would be..... 2.5 times increase in perf/watt. Of course they were saying same clocks, so a 20% increase in clocks is creating how much drop in perf/watt?

What you have to consider that 200 mhz increase is increasing how much voltage, to create that drop in perf/watt......

And we can't look and compare with nV's cards either, cause a 200 mhz increase in nV's cards, power usage goes up fairly linear, because voltage doesn't change too much. Once ya get to 300 mhz increase on their cards, then ya got start increasing voltage.
 
Last edited:
razor1 said:
1.7 is lower than what Samsung, AMD, and GF stated 14nm would give. 60% drop in power usage would be..... 2.0 times increase in perf/watt. Of course they were saying same clocks, so a 20% increase in clocks is creating how much drop in perf/watt?

What you have to consider that 200 mhz increase is increasing how much voltage, to create that drop in perf/watt......

And we can't look and compare with nV's cards either, cause a 200 mhz increase in nV's cards, power usage goes up fairly linear, because voltage doesn't change too much. Once ya get to 300 mhz increase on their cards, then ya got start increasing voltage.
Click to expand...
I don't they mentioned 2.8 was only for VR. Or I missed it somewhere. I remember his saying 1.7 from finfet and 2.8 from finfet+ architecture enhancements + amd technology. It had three pieces to it.
 
check as the slide is switching, Raja states he wasn't talking about gaming. So up till that point he was only talking about VR.

I think he specifically stated this "wait, we haven't talked about gaming yet"



Ok this is what he stated

"So wait I spoke about VR, but what about gaming, Redeon RX series also delivers power gaming"

So all the slides they had prior to that was about VR, which makes sense with the ATI technologies moniker.
 
Last edited:
I apologize for not reading the entire thread as I'm in a hury (got through the first 8 posts), but just in case no one pointed it out, the reason the Crossfire are so efficient is because the combined usage is only 51%, meaning each card is roughly only being 25% utilized; thus, a much lower power draw.

It's not so much marketing hype, as just reality.
 
Its utilization of the dual card set up isn't calculated that way, need more info about it though. AMD's Hallock stated some stuff, but didn't make much sense.
 
razor1 said:
1.7 is lower than what Samsung, AMD, and GF stated 14nm would give. 60% drop in power usage would be..... 2.0 times increase in perf/watt.
Click to expand...

I am not sure what this particular thing has to do with the topic, but it's a scale of power consumption, clockspeed, and die size that needs to be balanced.


razor1 said:
check as the slide is switching, Raja states he wasn't talking about gaming. So up till that point he was only talking about VR.
Click to expand...

No he doesn't. He says, "But wait, I talked about VR, what about gaming?" He talked about VR starting at 8:45, and then he reveals the card at 13:15, specifications at 14:10, power efficiency at 15:04. Now if 2.8x is power efficiency is automatically VR because of his last line, under that assumption the card reveal, specifications (including 1.4 DP support), and the 14nm FinFet, should only be about VR as well.
 
Trust me its only VR, there is a reason for that, there is also a superscripted 1 lol, you don't know what it points to do you? I might.......

Yeah I asked someone that would have known lol.

Variant I stated this before, its great to have Hope for better than what seems reality, but don't fool yourself.

I'm still hopeful its 2.0 perf/watt over tonga, but seriously after that presentation, I don't think that is average at ALL.
 
razor1 said:
check as the slide is switching, Raja states he wasn't talking about gaming. So up till that point he was only talking about VR.

I think he specifically stated this "wait, we haven't talked about gaming yet"



Ok this is what he stated

"So wait I spoke about VR, but what about gaming, Redeon RX series also delivers power gaming"

So all the slides they had prior to that was about VR, which makes sense with the ATI technologies moniker.
Click to expand...


I have to disagree there. He was talking about VR but during power efficiency he never once mentioned that power efficiency was only for VR. Correct me if I am wrong there, because I watched that entire portion. I think its unfair to label it for VR when its not. Little unfair assumption when he was talking about the slide he never mentioned VR once but he was talking about general efficiency.

Could it be VR? Yea possibly but it was not something they directly mentioned.
 
Last edited:
razor1 said:
well the perf/watt is not as good as pascal, otherwise they would have shown that or talked about perf/watt in current games, which they were talking about before, now they have mysteriously omitted that from their presentation and talked about pref/watt with VR......

Marketing only changes strategies when something is better or worse than originally expected (in this case Pascal came out), they prefer not to change if at all possible because they want to keep a unified message. Confusing end consumers is never good a thing.
Click to expand...

Impossible to make such a claim/assumption when don't know the Wattage and the performance in an equal comparison.
The ONLY measure we have seen of performance is AOTS, however no power (wattage) numbers were given. On purely GFLOPS/W, again impossible to say without knowing W, however AMD made a larger generational leap forward in IPC, from last gen to this gen than Nv did, given that AMD was further behind to begin with, AMD catching up to NV.

IPC
SP X Rate X (Ops/Cycle) = GFlops / (SPxRate)
Using the 1266Core clock from TPU

RX 480 = 2304SP x 1266Mhz x 2 (2 floating point operations per cycle) = 5834 GFlops = 2.000093
GTX 1070= 1920 x 1683 = 6462 GFlops = 1.999777
GTX 1080= 2560 x 1733= 8872 GFlops = 1.999783

Assuming rumors are right..
Polaris @ 96W = 60GFlops/W, @ 130W = 44.8GFlops/W
Pascal @180W (GTX1080) = 49.8GFlops/W, @ 150 = 43.8Gflops/W

Using AOTS:
480 = 40FPS / 96 = .41fps/W, @130W = 30.1fps/W
GTX 1080 = 52.2FPS / 180 = .29fps/w

Now as I mentioned, since AMD was further behind to begin with it had much more to make up. When you compare AMD and NV parts to their previous gen parts, AMD again comes out ahead.
ie the RX 480 is to the R9 380, what the GTX 1070 is to the 970

1070 @ $449/379, vs the 970@ $329.
The 1070 on average is about 60% greater performance average ot +25% more $$ and 10% less power (160W vs 145W)
480 @ $199, vs R9 380 @ $199
The 480 early estimates, falls between the 390-390X
http://www.anandtech.com/bench/product/1591?vs=1559

Avg difference between 390 vs 390X about 15% so split that.. about 7.5% so compare the 480 vs 380..
390X = on avg 75-80% over 380 - 7.5% .. places it about 70% FASTER than 380, being the SAME costs, and assuming rumored 130W usage (vs 180W of 380).. using less then 3/4th the power
Even comparing to the $229 8GB.. and lets say OC'd to sh!t.. using 150W.. is still at a minimum 70% faster for +10% more $ and %21 less W.
Using the aforementioned 96W (questionable), it's almost reaches the 2x Perf/W claim AMD made, nearly 50% less (47.7)

So to wrap up, again until hard numbers come out on Polaris, its hard to say conclusively one way or another if AMD is caught up or suprassed Pascal, however given the estimates we have seen so far, it would appear any advantage Pascal held has been erased.
 
I thought we couldn't compare GFLOPs between the two? If we could the Fury X (8.3 TFLOPs) would completely destroy the 980ti (5.8 TFLOPs) which we know it's actually worse.
 
You must log in or register to reply here.
Back
Top