RUMOR: Radeon 480 to be priced to replace 380, similar output as 390.

I don't even know why he's hung up on dual-gpu Polaris 10. I meant two Polaris 10 cards xfire.

With 1000 euros saved vs buying Radeon Duo, you can buy a whole new system, make up the extra pci-express slot you used xD
 
Why is this about nVidia vs AMD?

This is about AMD and what performance levels they will hit.

Some people here are bashing AMD saying they wont even hit Fury levels, and I think they is ZERO evidence to back that up. Its nonsense that dismisses that AMD will have significant performance gains along with their power gains... And then to follow it up by adding that nVidia is making performane gains and is better just makes it seem like fanboy nonsense.

Ieldra said:
I don't even know why he's hung up on dual-gpu Polaris 10. I meant two Polaris 10 cards xfire.

With 1000 euros saved vs buying Radeon Duo, you can buy a whole new system, make up the extra pci-express slot you used xD
Click to expand...

295x2 tanked in price pretty quickly. They are price grabbing while they know Dual Fiji is the #1 single slot card, and they will obviously lower price with time once Polaris 10 is out.

Are you using this logic to make conclusions?
 
jamesgalb said:
Why is this about nVidia vs AMD?

This is about AMD and what performance levels they will hit.

Some people here are bashing AMD saying they wont even hit Fury levels, and I think they is ZERO evidence to back that up. Its nonsense that dismisses that AMD will have significant performance gains along with their power gains... And then to follow it up by adding that nVidia is making performane gains and is better just makes it seem like fanboy nonsense.
Click to expand...
Hahaha who said anything about nvidia? You have some kind of inferiority complex man.

Polaris 10 will compete with gp106

Vega 11 with gp104

Vega 10 with gp 102

You're literally putting words in our mouths
 
jamesgalb said:
Why is this about nVidia vs AMD?

This is about AMD and what performance levels they will hit.

Some people here are bashing AMD saying they wont even hit Fury levels, and I think they is ZERO evidence to back that up. Its nonsense that dismisses that AMD will have significant performance gains along with their power gains... And then to follow it up by adding that nVidia is making performane gains and is better just makes it seem like fanboy nonsense.
Click to expand...


nV isn't dropping their power usage to make those gains (you can have one or the other power reduction or performance increase or mixture of both) for course some added benefit of redesign to help through put..... And you were the one that brought up all that
 
Best of it all is, remember people were saying AMD is going to have months of time to take advantage of nV's absence in the market because AMD showed off Polaris and nV was quiet? Guess what nV was doing its thing and was focused on what they had to do.

And it ain't bashing if AMD isn't marketing the cards to go up against Pascal's initial release, shit I even stated its a good move if they have the time to capitalize on it! And now it looks like a risky move because gp106 is around the corner.
 
jamesgalb said:
AMD briefly dropped the word 'mainstream' in a conference.. They never made an official slide or Polaris roadmap that shows it will 'Only be a mainstream card', they didnt even say 'only mainstream' when they said it briefly during the conference. Even if they did there is no saying they are using some consistent AMD definintion of what 'mainstream' applies to in all contexts...

People are REALLY reaching here in an attempt to show that AMD has implied they wont be making a high end card, and I dont know why.



I dont think there will be a Dual Polaris. Where has there been intel on a dual Polaris coming out?

Dual Fiji will beat Polaris and no Dual Polaris will come out, ensuring the selling of Fiji and HBM chips until Vega,
Click to expand...
All we are saying is that we do not think AMD is releasing a replacement for the 290x/390x models yet (so no 490x) and they are focusing on below enthusiast.
I just went back through the interview with Raja on PCPerspective and around 8min 20sec he outlines that their priority is the millions of potential customers using Polaris and then follow that up later with enthusiast - his words bring it to millions initially (actually said priority) and enthusiast later.
This fits the mainstream model all of us have been talking about and seen in slides.
In the same interview Raja also mentions they are not looking to do dual GPUs just yet, the dual Fiji is to try and get developers to buy into the concept that then opens it up to lower cards, but again this looks to be the enthusiast line.
I think we will see the current top Polaris possibly with similar performance to the 1070, but nothing comparable to the 1080 (and this sucks IMO as I would love AMD to be in a position to do so as it would help with pricing and approach NVIDIA would have).
Cheers
 
Last edited:
I just noticed your 'you think they're only making two Polaris gpus?' point

It seems like there's two SKUs for each Polaris 10 and 11, by my math that makes four, not two.

How many gcn1. 2 SKUs were there hm?

Fiji, Fiji xt

Tonga, tonga (xt?)

Also seems like four to me
 
Last edited:
Yeah that is what we are kinda getting at before, a full chip and one with 4 cu's cut out for Polaris 10.

Polaris 11 same thing full chip and one with some cu's cut.
 
razor1 said:
Yeah that is what we are kinda getting at before, a full chip and one with 4 cu's cut out for Polaris 10.

Polaris 11 same thing full chip and one with some cu's cut.
Click to expand...
It's almost like you can predict, with reasonable accuracy, future releases based on release history. Madness!
 
Well its been working for them why not keep going, any more versions things just get crazy. Cost of creating a full die chip and then cutting out parts for lower bins, isn't cost effective. So if yields are so poor that you need to cut more parts to create a third lower end sku, that is just bad news. Better to wait for yields to get good enough to avoid that.
 
razor1 said:
Well its been working for them why not keep going, any more versions things just get crazy. Cost of creating a full die chip and then cutting out parts for lower bins, isn't cost effective. So if yields are so poor that you need to cut more parts to create a third lower end sku, that is just bad news. Better to wait for yields to get good enough to avoid that.
Click to expand...
Yes, this is why I'm very curious about reported 3 SKUs for gp104
 
I'm thinking the lower end sku is for mobile...... If I remember correctly the 970m was released right after the gtx 980 and 970 were released.
 
Yes that is my guess as well, part number very close to x70, in line with maxwell launch

Perhaps they will launch... Asynchronously
 
serial only lol. asynchronously will absolutely screw up their financials
 
  • Like
Reactions: N4CR
like this
jamesgalb said:
Why is this about nVidia vs AMD?

This is about AMD and what performance levels they will hit.

Some people here are bashing AMD saying they wont even hit Fury levels, and I think they is ZERO evidence to back that up. Its nonsense that dismisses that AMD will have significant performance gains along with their power gains... And then to follow it up by adding that nVidia is making performane gains and is better just makes it seem like fanboy nonsense.



295x2 tanked in price pretty quickly. They are price grabbing while they know Dual Fiji is the #1 single slot card, and they will obviously lower price with time once Polaris 10 is out.

Are you using this logic to make conclusions?
Click to expand...

Hey James, I am not a fanboy of either camp. I haven't had a gaming rig for a bit I have an i5 gigabyte brix pc, but I do stay up to date about everything and follow discussion as I will probably build a little more powerful pc when Zen comes out and see how it compares to intel and taking price point in to account.

To the point, I really don't think anyone here is bashing AMD. Just imagine a product that might perform similar to 390x or may be 1070 and costing less than 300 bucks. That is a killer product regardless of the fact that its enthusiast card or not. I know I would recommend it to a lot of people.

Now here is the thing. AMD I believe had plans to release Vega right along with these cards. Vega seems to be much much more powerful chip. I believe they had every plan to launch it but considering both sides decided to delay HBM2, probably because it isn't being mass produced yet.

I bet vega 10 is ready to go but they won't release it with gddr5 if it was always designed to leverage HBM2.

Polaris isn't a bad chip, its a small but powerful chip. They could have easily made it bigger to compete with Pascal but they weren't expecting the HBM2 delays. So for now they are going for market share and selling a powerful card for cheap price.

There is absolutely nothing wrong with polaris as to where it stands. Less then 150w and performance close to 390x? That deserves a clap. Vega would be enthusiast chip but due the delay it seems like amd didn't plan for enthusiast chip. They might release a 490x with gddr5x that might get you 980ti performance? But we don't know about that yet. All we have seen is chip that has little less shaders than full polaris 10 chip.

I think you will see 480, 490 and quite possibly 490x and then Vega with HBM 2.

Cheers.
 
Raja definitely said AMD would be focusing on multi GPU cards in 2017-2018 timeframe, and that NV would be forced to do the same. Polaris is 2016, obviously, but that doesn't actually rule it out, particularly as all evidence seems to say that Polaris 10 will offer 390-like performance and thus will not answer to GP104.

He also said they had a huge investment in HBM, so it is plausible that top-end Polaris 10 cards, possibly multi-GPU varients, would come armed with HBM1. The 4GB limit is per-GPU, and memory is pooled in explicit multi-adapter.

He also implied that Polaris would hit every segment including enthusiast, but I personally feel that's just AMD blowin' smoke like they always do.
 
razor1 said:
I don't know looks like he is looking for the holy trinity or something.

Jasmesgalb

We know more about pascal's inner workings than we know about Polaris (GTC pascal white papers),

We know Pascal will have better through put than Maxwell 2 did close to 20% more through put and in compute based situations its going to be even higher than that

We know they will have higher clocks too. We also know, 16nm tends to give a bit more advantage over 14nm with power consumption, look at Iphone's A9 chip it was built on both 14nm Samsung and 16nm TSMC.

We know nV has more experience in extracting performance on nodes by using custom libraries. Should I go on?
Click to expand...

I wouldn't say using the A9 chips as an example is indicative of node performance. The "advantage" was 1-3% better battery life, but again it's on an ARM SOC. Then you make some absurd statement about how Nvidia has some magic voodoo and is always better at extracting performance. The only statement I can agree with is that the finfets will likely have higher clock speeds than we see now. We know about the same from both AMD and Nvidia, currently regarding GPUs. I just hate how everyone loses their mind believing all these "leaks" which are usually nothing more than hype to get a million suckers visiting their site for ad revenue.
 
razor1 said:
serial only lol. asynchronously will absolutely screw up their financials
Click to expand...
Yes but is 10% financial screw-up the same at $2bn revenue as at $10bn revenue?
ebduncan said:
I wouldn't say using the A9 chips as an example is indicative of node performance. The "advantage" was 1-3% better battery life, but again it's on an ARM SOC. Then you make some absurd statement about how Nvidia has some magic voodoo and is always better at extracting performance. The only statement I can agree with is that the finfets will likely have higher clock speeds than we see now. We know about the same from both AMD and Nvidia, currently regarding GPUs. I just hate how everyone loses their mind believing all these "leaks" which are usually nothing more than hype to get a million suckers visiting their site for ad revenue.
Click to expand...


These leaks are pretty accurate, and the 'magic voodoo' are custom libs... Not magic at all... Just custom libs.

Custom libs can very fucking expensive to license and/or produce in-house, the advantages are tighter timing, among others, and this usually translates into higher clocks, aka extracting more performance
 
Last edited:
ebduncan said:
I wouldn't say using the A9 chips as an example is indicative of node performance. The "advantage" was 1-3% better battery life, but again it's on an ARM SOC. Then you make some absurd statement about how Nvidia has some magic voodoo and is always better at extracting performance. The only statement I can agree with is that the finfets will likely have higher clock speeds than we see now. We know about the same from both AMD and Nvidia, currently regarding GPUs. I just hate how everyone loses their mind believing all these "leaks" which are usually nothing more than hype to get a million suckers visiting their site for ad revenue.
Click to expand...


ah no nV doesn't have a magic wand, its called actually tuning the fabrication cell libraries based on their architecture. Yes this can be done, and nV has been doing it since the g80, this is partially why they have been able to push clocks higher. AMD/ATi has not done this before. I do expect them to do it with Vega as there has been hints. And these aren't from leaks I rarely get my info from leaks........

Polaris this is not possible, as the two fabs they are using, the processes are similar but not the same so the libraries will not translate over to each other. And if you think they will do custom libraries on both Fabs, the risks with validation will be too great to do such a thing.

The rest of the node for node, TSMC says they can get more performance and lower power usage out of their node form their 28nm process 70% decreased power or 45% increased performance at the same power, which is more than the 14nm GF process (what GF, Samsung, AMD stated about 14nm process)
 
Last edited:
razor1 said:
ah no nV doesn't have a magic wand, its called actually tuning the fabrication cell libraries based on their architecture. Yes this can be done, and nV has been doing it since the g80, this is partially why they have been able to push clocks higher. AMD/ATi has not done this before. I do expect them to do it with Vega as there has been hints. And these aren't from leaks I rarely get my info from leaks........

Polaris this is not possible, as the two fabs they are using, the processes are similar but not the same so the libraries will not translate over to each other. And if you think they will do custom libraries on both Fabs, the risks with validation will be too great to do such a thing.

The rest of the node for node, TSMC says they can get more performance and lower power usage out of their node form their 28nm process 70% decreased power or 45% increased performance at the same power, which is more than the 14nm GF process (what GF, Samsung, AMD stated about 14nm process)
Click to expand...
I didn't know Polaris was being fabbed at both, gf has a die area advantage if I'm not mistaken
 
I'm honestly really excited about this. I understand why some users are frustrated but this will be a good move by AMD for system builders and for me even.
 
schizo said:
He also said they had a huge investment in HBM, so it is plausible that top-end Polaris 10 cards, possibly multi-GPU varients, would come armed with HBM1. The 4GB limit is per-GPU, and memory is pooled in explicit multi-adapter.
Click to expand...
We do know AMD plans to integrate leading discrete GPU level capabilities into future APUs using HBM with at least 128GB/s based on a paper. So they likely have a Polaris 10/11 in Zen requiring at least 1 stack of HBM to get that 128GB/s for one of their products. Since that lines up perfectly with a single stack of HBM1 that seems likely. So they could move a lot of HBM1 by integrating it into their APUs. It also means they need 128GB/s for whatever chip they included. Seems appropriate for Polaris 11, but Polaris 10 should be double. AMD has been very careful not to release information that would allow someone to triangulate their performance. They just avoid answering any questions about HBM1 and giving relative numbers without a baseline. HBM2 will also be significantly more expensive than HBM1, so a lot of little HBM1 stacks may be cheaper than HBM2 using FINFET and a smaller process. Not to mention better yields since the dies are half the area. The bandwidth numbers they're showing for their bus line up almost perfectly what would be required for a theoretical dual Polaris 10 part.
 
jamesgalb said:
Why is this about nVidia vs AMD?

This is about AMD and what performance levels they will hit.

Some people here are bashing AMD saying they wont even hit Fury levels, and I think they is ZERO evidence to back that up. Its nonsense that dismisses that AMD will have significant performance gains along with their power gains... And then to follow it up by adding that nVidia is making performane gains and is better just makes it seem like fanboy nonsense.



295x2 tanked in price pretty quickly. They are price grabbing while they know Dual Fiji is the #1 single slot card, and they will obviously lower price with time once Polaris 10 is out.

Are you using this logic to make conclusions?
Click to expand...


Because some people can't help themselves but make this about AMD vs NVidia.
 
Now there's a rummor saying 299USD for the Polaris 10, with 980ti performance.

Which it's pretty nonsense. looking how AMD and Nvidia works, they only share the final pricing to their AIB partners a couple of weeks before the official announcement not 2 months before.
And, it's silly as they would show their hand to the competition. You could never trust Taiwanese companies to hold a secret..
 
lukart said:
Now there's a rummor saying 299USD for the Polaris 10, with 980ti performance.

Which it's pretty nonsense. looking how AMD and Nvidia works, they only share the final pricing to their AIB partners a couple of weeks before the official announcement not 2 months before.
And, it's silly as they would show their hand to the competition. You could never trust Taiwanese companies to hold a secret..
Click to expand...


Let's say AMD releases Polaris 10, and it is somewhere between a 980ti and a 390x for the top end. And let's say then that card is $299.

For budget system builders and people on the hunt for the biggest bang for their buck, they are going to flock to that card. It's a smart ploy. The game here for AMD is to grab marketshare. If AMD can capture the low and mid market with a new architecture priced significantly cheaper than their forebearers, then at that point they will be able to assume a larger share of the market, which will help them out in the long run.
 
lukart said:
Now there's a rummor saying 299USD for the Polaris 10, with 980ti performance.

Which it's pretty nonsense. looking how AMD and Nvidia works, they only share the final pricing to their AIB partners a couple of weeks before the official announcement not 2 months before.
And, it's silly as they would show their hand to the competition. You could never trust Taiwanese companies to hold a secret..
Click to expand...
Yeah and that rumour was for a 490x......
I find it hard to believe it tbh as it ties to the 490x, which should be comparable to GTX 1080 and so above 980ti.
Seems this rumour is merging previous leaks-rumours and speculating *shrugs*.
Cheers
 
Last edited:
CSI_PC said:
Yeah and that rumour was for a 490x......
Click to expand...
More likely they saw a Polaris 10 and assumed it was a 490X. Statements and rumors are suggesting a $200 and $300 version respectively of the same chip. That bottom SKU could be one hell of an overclocker, unless there's a significant memory bottleneck involved there. More likely a 380X that appears to be landing real close to where I was estimating with my math.
 
Anarchist4000 said:
More likely they saw a Polaris 10 and assumed it was a 490X. Statements and rumors are suggesting a $200 and $300 version respectively of the same chip. That bottom SKU could be one hell of an overclocker, unless there's a significant memory bottleneck involved there. More likely a 380X that appears to be landing real close to where I was estimating with my math.
Click to expand...
You really think a 480x or even 490 (more believable) is going to have the performance matching that of the 980ti and Fury X, and compounded by AMD also continue selling the Fury X?
It destroys any reason to sell Fiji top tier, which is the only rumour we have not heard yet being Fury Fiji line being ceased.

Cheers
 
My initial thought was price cut, but that's too much of a loss.

Kinda weird they would r&d furyx just to scrap it after less than a year. AMD isn't good at handling money so I guess we chalk this up as their latest blunder.

I said last year HBM was disastrous for the FuryX. But AMD can't release a 350W flagship GPU.
 
CSI_PC said:
You really think a 480x or even 490 (more believable) is going to have the performance matching that of the 980ti and Fury X, and compounded by AMD also continue selling the Fury X?
It destroys any reason to sell Fiji top tier, which is the only rumour we have not heard yet being Fury Fiji line being ceased.

Cheers
Click to expand...
I think it will be just short on performance, but it's also a substantially cheaper card targeted down a market tier. Fiji I think will be retired shortly. It was a cut down design to meet the constraints of 28nm because 20nm fell through. Why it has the same ROP output of a 390X. 490 series I can't help but think are either dual die or small vegas. Maybe they leave fiji in place to clear inventory for a while, but I can't see continuing to produce them with their cost.

TaintedSquirrel said:
My initial thought was price cut, but that's too much of a loss.

Kinda weird they would r&d furyx just to scrap it after less than a year. AMD isn't good at handling money so I guess we chalk this up as their latest blunder.

I said last year HBM was disastrous for the FuryX. But AMD can't release a 350W flagship GPU.
Click to expand...
FuryX was more of a stopgap to me. The R&D money wouldn't be a total loss either. There would be lessons learned with HBM and interposers that would translate to Polaris, Zen, and Vega.
 
AMD wont do the same mistake as they did with the 5800 series. They priced it very competitive that for the first time seen the GPU prices actually going up the following months to the announcement...
 
On the mobile front, notebookcheck has just added an entry for the R9 m490 and m480. Extremely limited information, but they're listing the 490m somewhat above the GTX 980m (just below a desktop 290x) in performance, and the m480 just above a desktop 7970/below a GTX 680. So we're probably getting cut-down and/or underclocked Polaris 10 and 11 in mobile. Nothing revelatory or surprising from current rumors. Notebookcheck seems to have good insider lines and they're initial listings are normally pretty accurate (680m, 780m, m7870/m8970/m290x, m295x were right on the money a couple months before their release). In any case, looks like excellent news for mobile gaming (pricing dependent).

AMD Radeon R9 M490

Not to go done the rabbit hole to much, but if you consider mobile parts normally perform 15-25% slower than desktop counterparts (from binning and/or underclocking), this would mean the m490 would perform somewhere around a 980 ti, and the m480 around a 290.
 
Last edited:
Err they stated 980m the gtx 980m is cut down from the gtx 980. So yeah its not going to get to the 980 ti if you are looking at m490 that is 20% less than a regular 490
 
The 980m is essentially a slightly cut-down desktop GTX 970 with only 1536 cores, 96 TUs, 64 ROPs and lowered clock speeds. It performs a shade better than a desktop GTX 960.

But notebookreview puts the m490 right beneath a desktop 290x (and above a 980m) which means if the desktop 490 is >20% faster it will be in the realm of the 980 ti. Still, conjecture though...
 
harmattan said:
The 980m is essentially a slightly cut-down desktop GTX 970 with only 1536 cores, 96 TUs, 64 ROPs and lowered clock speeds. It performs a shade better than a desktop GTX 960.

But notebookreview puts the m490 right beneath a desktop 290x (and above a 980m) which means if the desktop 490 is >20% faster it will be in the realm of the 980 ti. Still, conjecture though...
Click to expand...
Well they are basing that news from Videocardz, and there they state everything as ? for M480 and M490, including ? against Polaris.
So unfortunately nothing can be obtained about those two models and their spec.
Cheers
 
lukart said:
Now there's a rummor saying 299USD for the Polaris 10, with 980ti performance.

Which it's pretty nonsense. looking how AMD and Nvidia works, they only share the final pricing to their AIB partners a couple of weeks before the official announcement not 2 months before.
And, it's silly as they would show their hand to the competition. You could never trust Taiwanese companies to hold a secret..
Click to expand...

The problem is that the performance metrics can be so abused in the way that you can cherry pick benchmarks and then project the result on it. You would not be lying but is it the whole truth or does it give you the complete picture , some common sense also might apply here 980TI prolly not in the same price range AMD will not put such a card with that kind of performance in a price bracket that low.
 
More mixed signals (if you were expecting more than "mainstream" at initial release) from AMD and Roy this time in an interview released today:
Why a reformed AMD is going all in on VR, rather than badmouthing Nvidia

Roy interview said:
"If you look at the total install base of a Radeon 290, or a GTX 970 or above [the minimum specs required for VR], it's around 7.5 million units," explained Taylor. "But the issue is that if a publisher wants to sell a £40/$50 VR game, there's not a big enough market to justify that yet. We've got to prime the pumps, which means somebody has got to start writing cheques to big games publishers. Or we've got to increase the install TAM [total addressable market].

"The reason Polaris is a big deal," continued Taylor, "is because I believe we will be able to grow that TAM significantly.
I don't think Nvidia is going to do anything to increase the TAM, because according to everything we've seen around Pascal, it's a high-end part. I don't know what the price is gonna be, but let's say it's as low as £500/$600 and as high as £800/$1000. That price range is not going to expand the TAM for VR. We're going on the record right now to say Polaris will expand the TAM. Full stop."
Click to expand...

Also notice the focus was on 290 and 970.

Cheers
 
Last edited:
You must log in or register to reply here.
Back
Top