AMD RX 400 series (Polaris) MegaThread! FAQ & Resources

[Martell1977]

Hello, I need some clarity, the RX 400 series is Polaris so does that mean that the RX 490 is a Polaris card or?

There are 2 flavors of Polaris: Polaris 10 & 11. Polaris 10 is the 480 (full chip) and the 470 (cut down chip), Polaris 11 is the 460 (cut down chip). Which is what was discussed before your post, will AMD release the full chip to compete with the 1050(ti).

The 490 will likely be Vega 11 and the models above it likely the Vega 10. Vega is and has always been officially expected in Q1 2017. It's not delayed, but since it uses HBM2, the launch date was probably scheduled for when mass production of HBM2 was online and available. Plus, it falls in line or at least very close to the launch of ZEN.

 

[littleleo]

Well there was only going to be two Polaris chips the 10 & 11.The RX 460 & RX 470 use the slower Polaris 11 and the RX 480 uses the Polaris 10. SO the question everyone is asking is what does the RX 490 use?

 

[Martell1977]

Well there was only going to be two Polaris chips the 10 & 11.The RX 460 & RX 470 use the slower Polaris 11 and the RX 480 uses the Polaris 10. SO the question everyone is asking is what does the RX 490 use?

Close, the 470 is Polaris 10, but with 32 CU's instead of 36. The 480 has all 36 CU's running.
http://www.tomshardware.com/reviews/amd-radeon-rx-470,4703.html

It's safe to assume the 490 would be Vega 11 as it should be the "low end" of the upcoming GPU's. It's entirely possible that AMD could release a cut down Vega 11 as the 485 or 480X, but for now I think they are sticking with rounded numbers. Yes, there is a lot of speculation there, but it's a trend that's easy to see. I'd expect a 490 Vega 11 cut down, 490X Vega 11 full, and maybe the Fury successor from Vega 10 cut down and Fury X successor from Vega 10 full.

We will all know come Q1 2017.

 

[Martell1977]

Anyone know if a 470 can be modded into a 480 or are the disabled CU's actually defective?

I'm thinking along the lines of people that were able to flash the BIOS of their 7950's into a 7970.

 

[Robert Cook]

Anyone know if a 470 can be modded into a 480 or are the disabled CU's actually defective?

I'm thinking along the lines of people that were able to flash the BIOS of their 7950's into a 7970.

based on the 7950 and I think even something before that and the fury, I would not be surprised if you could unlock part of it I guess there is not much of a push for it. although I guess even say two extra CUs could help. I am sure someone somewhere is working on it (no I do not mean the mod on here)

 

[jaymc]

Well not arcoding to AMD's own roadmap...

Vega looks to be due around the end of 2016..(not Q1 2017)

It can be seen here:
http://wccftech.com/amd-vega-10-gpu-hbm2-q4-2016-professional/

 

[Martell1977]

Well not arcoding to AMD's own roadmap...

Vega looks to be due around the end of 2016..(not Q1 2017)

It can be seen here:
http://wccftech.com/amd-vega-10-gpu-hbm2-q4-2016-professional/

The roadmap has Vega right above the 2017, making the window for launch the beginning of 2017, maybe an option for the end of 2016. As for the rumor of an October release, AMD squashed it and October is nearly gone. AMD confirmed the Q1 2017 launch time frame.

It would have been great for them to get Vega out before Christmas, but will settle for just a bit after. They can still get the people looking to burn up some gift cards, as well as the regular customers, LOL.

 

[jaymc]

That's arguable, it looks to be just before 2017, end of 2016...

It's very close were splitting hairs here... But it does looks like December. Remember Q1 is anything up to April, it definitely does not look like that anyway.

Sorry when did they confirm a Q1 launch timeframe ?? I wasn't aware that they did that.

Jay

 

[Martell1977]

That's arguable, it looks to be just before 2017, end of 2016...

It's very close were splitting hairs here... But it does looks like December. Remember Q1 is anything up to April, it definitely does not look like that anyway.

Sorry when did they confirm a Q1 launch timeframe ?? I wasn't aware that they did that.

Jay

It's likely vague to spark discussions like this, LOL.

This isn't the article I read about it, but it's similar info:
http://www.asiastarz.com/articles/10100/20160915/amd-vega-graphics-card-release-date.htm
The article I'm referring to was talking about Zen and Vega both launching in Q1 2017, but maybe not exactly at the same time. AMD was supposedly trying to launch them as closely together as they could. I didn't bookmark the article though...

 

[Math Geek]

yah read the same info myself. also can't link it but i'm pretty sure q1 was what amd said for both vega and zen releases.

 

[jaymc]

Think I seen the same story alright...

Did I read here somewhere that they are waiting on HBM2 memory to ramp up production for Vega ?

 

[Sakkura]

If you didn't read it here you probably read it somewhere else. They kinda have to wait for HBM2 production to ramp up, obviously.

And I suspect their internal roadmap probably just said Winter 2016-17 or something for both Vega and Zen. When exactly it launches isn't fully under their control anyway, what with HBM2 supply and fabs being out of their hands.

 

[-Fran-]

In any case, remember "launch" is not the same as "mass availability".

They can totally do a paper launch, even sending samples, but actual card selling could start later on.

And if HBM2 is not ramping up in big numbers, you can bet your furry behinds we won't see any cards in a good 4 months time after official launch because they'd be sold out everywhere.

Cheers!

 

[jaymc]

Can anyone shed any light on HBM2 production.. ?

I tried google but I can't seem to find anywhere how they are doing with production..

Say's here 16GB HBM2 production in Q3 2016... for Vega and Nvidia's "big pascal"..
http://www.gamersnexus.net/news-pc/2517-hbm2-specs-and-launch-in-3q16-under-sk-hynix

 

[Martell1977]

"Nvidia's "big pascal"

Could be reference to the Titan XP, which doesn't use it. It could hint to a 1080ti using it, but nVidia has been closed lipped about even the existence of one. However, IIRC, Titan XP is a cut down chip, so unless they are only using the full chips for data centers, it could be referring to a release of that. But the price would be huge as the XP is $1200 and HBM2 costs more than GDDR5X and the cost of redesigning the memory interface on the chip...

That being said, I think nVidia is waiting for Tesla for HBM2.

I hadn't really paid much attention before, but it looks like AMD's Navi will be HBM3?

 

[con635]

Will be interesting to see if AMD tosses out a '465' to fill the gap..

Since the 460 is a fully enabled Polaris 11 chip, the only option they would have is a further cut down Polaris 10. I don't think they would do that unless there are a lot of partially defective chips that could be used for it and disabling good sections wouldn't be cost effective.

AFAIK 460 is not fully enable polaris 11. that's why i wonder why AMD did not counter 1050ti with fully enable polaris 11 and drop 470 price instead.

I think a zen apu would be better for that end of the market, the igp would be just need to be sort of close if the cpu was around i5 level, lets face it the igp would be technically 'free' vs the 1050 if the cpu portion was up to scratch vs intel.

doubt that even the fastest APU based on Zen the gpu portion can touch 1050/RX460 performance.

What about this? Just rumor of course but I'd say priced right will go a long way to kill low end dgpu. More cu and hbm.
http://www.bitsandchips.it/52-english-news/7622-rumor-two-versions-of-raven-ridge-under-development

 

[renz496]

"Nvidia's "big pascal"

Could be reference to the Titan XP, which doesn't use it. It could hint to a 1080ti using it, but nVidia has been closed lipped about even the existence of one. However, IIRC, Titan XP is a cut down chip, so unless they are only using the full chips for data centers, it could be referring to a release of that. But the price would be huge as the XP is $1200 and HBM2 costs more than GDDR5X and the cost of redesigning the memory interface on the chip...

That being said, I think nVidia is waiting for Tesla for HBM2.

I hadn't really paid much attention before, but it looks like AMD's Navi will be HBM3?

they said new memory architecture for Navi. with Vega they specifically said it will be HBM. so it might be something completely new. some suggesting that AMD moving to new memory architecture with Navi because of HBM complexity.

 

[renz496]

Will be interesting to see if AMD tosses out a '465' to fill the gap..

Since the 460 is a fully enabled Polaris 11 chip, the only option they would have is a further cut down Polaris 10. I don't think they would do that unless there are a lot of partially defective chips that could be used for it and disabling good sections wouldn't be cost effective.

AFAIK 460 is not fully enable polaris 11. that's why i wonder why AMD did not counter 1050ti with fully enable polaris 11 and drop 470 price instead.

I think a zen apu would be better for that end of the market, the igp would be just need to be sort of close if the cpu was around i5 level, lets face it the igp would be technically 'free' vs the 1050 if the cpu portion was up to scratch vs intel.

doubt that even the fastest APU based on Zen the gpu portion can touch 1050/RX460 performance.

What about this? Just rumor of course but I'd say priced right will go a long way to kill low end dgpu. More cu and hbm.
http://www.bitsandchips.it/52-english-news/7622-rumor-two-versions-of-raven-ridge-under-development

we will need to see actual chip first with complete performance figure and power consumption. RX460 dedicated 75w+ alone for gpu processing. and those APU were rated for 95w for both CPU and GPU. and we probably will not going to see APU with HBM soon. because those HBM will be expensive.

 

[Sakkura]

"Nvidia's "big pascal"

Could be reference to the Titan XP, which doesn't use it. It could hint to a 1080ti using it, but nVidia has been closed lipped about even the existence of one. However, IIRC, Titan XP is a cut down chip, so unless they are only using the full chips for data centers, it could be referring to a release of that. But the price would be huge as the XP is $1200 and HBM2 costs more than GDDR5X and the cost of redesigning the memory interface on the chip...

That being said, I think nVidia is waiting for Tesla for HBM2.

I hadn't really paid much attention before, but it looks like AMD's Navi will be HBM3?

GP100 uses HBM2. That goes into the Tesla P100. The full launch of the Tesla P100 is scheduled for, well, Q4 2016, so any day now. But I think it may have been shipping in limited numbers for a little while already. Not entirely clear as supercomputer specs aren't always updated promptly.

 

[littleleo]

That Tesla P100 card is around $5700 ea. ouch!

 

[Math Geek]

but if it gives me an extra 10 fps or so, then it's worth it. 😛 no such thing as overkill!!!! (at least that's what folks keep saying to me when i use the word to describe 1080p on a gtx 1080)

 

[littleleo]

but if it gives me an extra 10 fps or so, then it's worth it. 😛 no such thing as overkill!!!! (at least that's what folks keep saying to me when i use the word to describe 1080p on a gtx 1080)

I love the angry baby; God Bless you! Can I send you two pcs?

 

[Math Geek]

absolutely 😀

 

[TechyInAZ]

but if it gives me an extra 10 fps or so, then it's worth it. 😛 no such thing as overkill!!!! (at least that's what folks keep saying to me when i use the word to describe 1080p on a gtx 1080)

I love the angry baby; God Bless you! Can I send you two pcs?

Send him an NVidia packed Tesla while your at it. 😛

 

[littleleo]

but if it gives me an extra 10 fps or so, then it's worth it. 😛 no such thing as overkill!!!! (at least that's what folks keep saying to me when i use the word to describe 1080p on a gtx 1080)

I love the angry baby; God Bless you! Can I send you two pcs?

Send him an NVidia packed Tesla while your at it. 😛

Shipping him 2pcs COD for $10K+ plus; man I love it! Thanks I'll have an early XMAS, lol.

I'm writing in Math Geek (Aka the Mad Baby) for President!