wccftechAMD Pirate Islands : R9 300 Series Alleged Specifications Detailed

[Silverforce11]

at least when you're considering multicard setups. Nvidia is still faster overall for single cards.

Was faster, latest drivers, R290X caught up:
http://forums.anandtech.com/showpost.php?p=36251663&postcount=63

 

[RussianSensation]

So your comparison is flawed from the very beggining.

I think you guys are not understanding me. I'll try it again. For simplicity's sake, let's even assume that R9 290X = GTX780Ti.

561mm2 780Ti -> Kepler = 100%
438mm2 R9 290X -> GCN 1.1 = 100%

Now, if NV were to simply produce a 561mm2 Maxwell, they'd get a 35% increase in IPC courtesy of the new architecture. Let's say that chip is 45% faster than 780Ti from a node shrink, add another 35% for IPC. NV would then end up with a chip that's 100% x 1.45 x 1.35 = 96% faster than 780Ti!

If AMD only makes a 438mm2 die, with a die shrink, let's assume they also get a 45% faster chip than R9 290X. How are they going to make up the 35% IPC deficit Maxwell's architecture brings? They either have to make a much larger chip than 438mm2 and/or GCN 2.0 will need to have a substantial increase in IPC to just match Maxwell's 35%.

There are 2 things that are still in favour of NV next round even if AMD gets GCN 2.0 to somehow match Maxwell's 35% IPC increase:

1) 780Ti OC is 20% faster than R9 290X OC. (HardOCP) Based on how much more dense the transistors are inside AMD chips, they produce a lot more heat which is why AMD's 438mm2 chip consumes similar amounts of power to NV's 561mm2. Therefore, if AMD makes a chip larger than 438mm2 and clocks it high, it should have less overclocking headroom than NV's.

All of NV's flagship large monolith chips overclocked better than AMD's equivalent ---> NV's 8800GTX, 285, 480, 580, 780 all overclocked far better on air than 3870, 4870/4890, 5870, 6970 and R9 290X. That's a track record if you ask me. I bet Maxwell's flagship chip will also overclock better than AMD's 20nm.

2) Maxwell has a 2x performance/watt increase on 28nm. That implies on 20nm, it could exceed that.

I don't see how AMD can compete next round unless they have 20-30 Mantle games, seriously improve their drivers, GCN 2.0 brings a 20-25% increase in IPC and they also make a 480-500mm2 die....or unless NV completely screws up Maxwell and delivers a 480mm2 Maxwell as flagship.

 

[SlowSpyder]

Without knowing what GCN 2.0 brings in terms of IPC, we're really just guessing, no?

 

[Enigmoid]

With regards to memory buses.

The 850m DDR3 (32 GB/s) gets around 3100 on firestrike and 4400 on 3dmark 11 GPU. The 850m GDDR5 (80 GB/s) gets around 3600 on firestrike and 4600 on 3dmark 11 gpu.

http://www.notebookcheck.net/NVIDIA-GeForce-GTX-850M.107795.0.html

This is way faster than the 7750 DDR3. It really looks like Nvidia only needs about half the BW of GCN 1.0.

Even an AMD 384 core DDR3 chip has trouble.

The perf of Maxwell with DDR3 is absolutely insane. AMD needs to reduce BW needs or biuld wider busses.

 

[Abwx]

Based on how much more dense the transistors are inside AMD chips, they produce a lot more heat which is why AMD's 438mm2 chip consumes similar amounts to NV's 561mm2.

Beware of theses kinds of "logics" unless you are aware that they are the ones that make sand houses suddenly collapse...

 

[raghu78]

Without knowing what GCN 2.0 brings in terms of IPC, we're really just guessing, no?

yeah we are. RS has already decided the winner of the next gen GPUs. its the GTX 880. hurray to the new king. :biggrin:

Infraction issued for member callout.
-- stahlhart

 

[PPB]

Wait, so in his fairy tale world only nv is capable of improving their uarchs? Who would have guessed...

Infraction issued for personal attack.
-- stahlhart

 

[sushiwarrior]

.

 

[stahlhart]

Observe forum posting rules when participating in discussions here, or the consequences will continue.
-- stahlhart

 

[Gloomy]

Fiji will be something new and different, with respects to memory config. Big die.

...is that your professional opinion? wink wink

 

[raghu78]

1) 780Ti OC is 20% faster than R9 290X OC. (HardOCP) Based on how much more dense the transistors are inside AMD chips, they produce a lot more heat which is why AMD's 438mm2 chip consumes similar amounts to NV's 561mm2. Therefore, if AMD makes a chip larger than 438mm2 and clocks it high, it should have less overclocking headroom than NV's.

that review was questioned in their own forums for having lower perf than previous reviews.

http://www.hardocp.com/article/2014/02/10/msi_geforce_gtx_780_ti_gaming_3g_video_card_review/4

looks like throttling issues due to 14.3 beta.

All of NV's flagship large monolith chips overclocked better than AMD's equivalent ---> NV's 8800GTX, 285, 480, 580, 780 all overclocked far better on air than 3870, 4870/4890, 5870, 6970 and R9 290X. That's a track record if you ask me. I bet Maxwell's flagship chip will also overclock better than AMD's 20nm.

performed better. yeah. clocked better. not really. HD 4890 clocked at 850 mhz stock. sapphire toxic was 960 mhz stock and easily hit 1 Ghz.

http://www.guru3d.com/articles_pages/sapphire_toxic_hd_4890_vapor_x_review,18.html

I don't remember any GTX 285 hitting 1 Ghz.

Also the HD 7970 was as good if not a better overclocker than GTX 680. Tahiti was known for its overclocking potential with watercooling. yeah it did require more power than GTX 680.

I don't see how AMD can compete next round unless they have 20-30 Mantle games, seriously improve their drivers, GCN 2.0 brings a 20-25% increase in IPC and they also make a 480-500mm2 die....or unless NV completely screws up Maxwell and delivers a 480mm2 Maxwell as flagship.

1. We don't know what GCN 2.0 brings in terms of architectural improvements
2. AMD looks likely to bring HBM to their next gen GPUs. They would be a generation ahead just as they were with GDDR5 on HD 4870.
3. We are already seeing Mantle show better frametimes in BF4 due to it being a better API than DX.

BF4 Ultra 4X MSAA at 60 fps . easily above 60 fps once you push the card to 1100 Mhz which seems to be easy on R9 295X2 .

http://hardocp.com/article/2014/04/08/amd_radeon_r9_295x2_video_card_review/3#.U0gX76KfZ8E

 

[NTMBK]

These rumours both look like nonsense. However, if I were to speculate about the next generation of GPUs, I would remember this- Intel is bringing out Knight's Landing with stacked memory in 2015. I wouldn't be surprised if the GPU makers followed suit.

 

[blackened23]

These rumours both look like nonsense.

Pretty much. Rumors start becoming more reliable when products are 2-3 months out; this has been the case with every prior GPU release. Chiphell picked up on the 7970 specs 3 months out. GTX 680? 3 months out. GTX 780/titan, etc etc etc 3 months out at chiphell. [FYI: chiphell is a TW/China based website with tech guys who work with Foxconn AND AIBs in China/Taiwan. So they know their stuff] Pay no attention to any rumors until a product is imminent. Chiphell has nothing on either 20nm product from NV or AMD. What does this tell us? Since Maxwell 2nd gen specifically seems to be a product for late this year perhaps, I don't buy into either rumor. Aside from this, Hawaii is still pretty fresh for AMD. There's just no way they have a product which is imminent in the next couple of months. I'm pretty sure they won't have anything until well into next year if it's a new uarch. Hawaii is fairly new still.

When chiphell starts getting reliable information, that's when everyone should start perking their ears and eyes at the rumors. Because those guys are close to the supply chain and are not constrained by NDA. Well, technically they are, but nobody over there cares. Because all cards are produced in China/TW first and foremost. And that's where the reliable rumors start and begin, this was the case for the past 5 years. And those rumors become reliable and reality based 2-3 months out.

Since we're not 2-3 months out from a new product. At least I certainly don't think so. Chances are , both of these rumors are absolutely fake. Good for a laugh though.

 

[realibrad]

Without knowing what GCN 2.0 brings in terms of IPC, we're really just guessing, no?

It seems like you are saying that the increase in IPC is due to the node shrinking and thus being able to add more to the same size chip. If Nvidia gets an increase of 35% because the number of transistors goes up per size unit, why would AMD not also get the same increase when they shrink down to 20nm?

Both AMD and Nvidia should get equal benefits of shrinking in terms of adding more transistors right?

Check out the review from Anand on the Sapphire Radeon R9 290 Tri-X OC. Look at the difference in clock rate, fps and temp. You will find that when you look at the cards you see that clock per clock the fps are within 10% or less diff. Then you look at the power draw not a huge diff. The efficiency difference is not all that great when you properly cool a 290. The main diff is that Maxwell can OC better because the larger chip means lower heat density. That comes at the cost of being more expensive to produce.

 

[raghu78]

Fiji will be something new and different, with respects to memory config. Big die.

HBM (High bandwidth memory). I think its pretty obvious by now that AMD's flagship 20nm GPUs. I am guessing the flagship will sport a 4 stack HBM for a massive 512 Gb/s bandwidth.

http://electroiq.com/blog/2013/12/amd-and-hynix-announce-joint-development-of-hbm-memory-stacks/

http://semiaccurate.com/2013/05/20/nvidias-volta-gpu-raises-serious-red-flags-for-the-company/

btw AMD had prototypes of stacked DRAM working way back in 2011.

http://semiaccurate.com/2011/10/27/amd-far-future-prototype-gpu-pictured/

http://sites.amd.com/us/Documents/TFE2011_006HYN.pdf

HBM will be in volume production in H2 2014.

http://www.simmtester.com/page/news/shownews.asp?num=15816

SK Hynix and AMD joined in development of HBM

Monday, December 23, 2013

An executive of SK hynix said Thursday the firm will cooperate with Advanced Micro Devices (AMD) of the United States to supply highly-advanced chips for data-intensive devices.

“SK hynix will mass-produce high bandwidth memory (HBM) chips by using state-of-art through-silicon via (TSV) technology from the latter half of next year at our DRAM plant in Icheon, Gyeonggi Province,” the executive told The Korea Times by telephone.

“We will partner with AMD to meet the explosive worldwide demand for HBM chips.” SK hynix is the world’s second-biggest computer chipmaker, and AMD is the world’s No. 2 supplier of microprocessors based on x86 architecture and also one of the largest suppliers of graphic processors."

 

[Mondozei]

Ignoring what's best for either AMD or Nvidia, at this stage for all people who buy GPUs, it'd be best if AMD came out as early as possible with 390X and began beating the big drum

This would force Nvidia to scrap their plan to revisit their launch of GTX 680 but only this time for GTX 880 and be forced to counter in full force.

Finally, even if you're an Nvidia fanboy(not calling anyone out here), you have to admit that if Nvidia didn't have any serious competition what we'd have instead is a gimped rolled out with underpowered cards and overpriced pricetags.

The only thing forcing Nvidia to keep on their toes is AMD, so it is good for everyone if AMD is strong - however not too strong, because they also need a strong competitor in Nvidia.

I personally wouldn't mind a third entry to the dGPU market. Intel is pushing ahead in the iGPU market and it'd be pretty awesome if they want for dGPU market even if I think it is beyond unlikely, it'd be pretty cool.

 

[Gloomy]

I love how the "no competition" narrative as an excuse for the 680 hasn't died yet.

 

[Keysplayr]

Ignoring what's best for either AMD or Nvidia, at this stage for all people who buy GPUs, it'd be best if AMD came out as early as possible with 390X and began beating the big drum

This would force Nvidia to scrap their plan to revisit their launch of GTX 680 but only this time for GTX 880 and be forced to counter in full force.

Finally, even if you're an Nvidia fanboy(not calling anyone out here), you have to admit that if Nvidia didn't have any serious competition what we'd have instead is a gimped rolled out with underpowered cards and overpriced pricetags.

The only thing forcing Nvidia to keep on their toes is AMD, so it is good for everyone if AMD is strong - however not too strong, because they also need a strong competitor in Nvidia.

I personally wouldn't mind a third entry to the dGPU market. Intel is pushing ahead in the iGPU market and it'd be pretty awesome if they want for dGPU market even if I think it is beyond unlikely, it'd be pretty cool.

What plan to revisit GTX680 launch? Do you mean repeating history where Nvidia launched a mainstream card (GTX680) to compete with AMD's flagship card (7970) ?
So like Nvidia will launch a mainstream card (GTX880) to compete with AMD's 390X flagship?

Of course this is a possibility.

 

[TreVader]

Nvidia won't stop gouging people until they stop making a ton of money doing it, so I really hope AMD comes out on top.


GK110 is great but you shouldn't have to fork over 999 just to get a single fully functional card.

 

[Keysplayr]

Nvidia won't stop gouging people until they stop making a ton of money doing it, so I really hope AMD comes out on top.


GK110 is great but you shouldn't have to fork over 999 just to get a single fully functional card.

GTX780Ti is $999.00?

Is that sorta like a GTX750Ti boosts to 1350?

 

[TreVader]

What plan to revisit GTX680 launch? Do you mean repeating history where Nvidia launched a mainstream card (GTX680) to compete with AMD's flagship card (7970) ?
So like Nvidia will launch a mainstream card (GTX880) to compete with AMD's 390X flagship?

Of course this is a possibility.

The 680 was $700, and the best thing in their arsenal for months. How is that midrange?



A midrange card by definition sits in the middle of a range of prices. How exactly did the 680 sit in the middle of anything?

 

[caswow]

What plan to revisit GTX680 launch? Do you mean repeating history where Nvidia launched a mainstream card (GTX680) to compete with AMD's flagship card (7970) ?
So like Nvidia will launch a mainstream card (GTX880) to compete with AMD's 390X flagship?

Of course this is a possibility.

or like now where one mid range 435mm² is beating its highend counterpart 550mm²?

both were highend at that time and both are mid now.

 

[TreVader]

GTX780Ti is $999.00?

Is that sorta like a GTX750Ti boosts to 1350?

That card can't even run titan fall on ultra detail, hardly "top of the line".


If a basic source-engined game is maxing out 3GB, you don't have enough.


The card is called Titan Black.

 

[Keysplayr]

or like now where one mid range 435mm² is beating its highend counterpart 550mm²?

both were highend at that time and both are mid now.

No idea what you're talking about, and we were talking about 680 I believe. Unless you've declared Hawaii a midrange GPU.

 

[Keysplayr]

That card can't even run titan fall on ultra detail, hardly "top of the line".


If a basic source-engined game is maxing out 3GB, you don't have enough.


The card is called Titan Black.

Pffft....