http://wccftech.com/amd-a...ics-4-ghz-boost-clock/
 

 

 

 
 
 

 
 
 

 
 
 
 
 
 
 

Wait, are they stating that the desktop version of the 7850K and 7700K are going to have TDP's between 35W and 15W? I have a hard time believing that if true...

LOL @ AMD...and a stock 2500k will smoke it at 5.0ghz......what a joke

SeanEJohan
LOL @ AMD...and a stock 2500k will smoke it at 5.0ghz......what a joke

LOL @ Intel...and a stock iGCN will smoke any Intel iGPU at any frequency...what a joke

Doing a Kaveri build for my parents. They couldn't understand why the case is so small for mITX. At the moment they have a Coolermaster HAF 930. The new case is a Bitfenix Prodigy. They're going from a 2007 Phenom 9500 build to a Kaveri build. Should be a pretty drastic change.

Will we see a 35watt quadcore?

knightsilver
Will we see a 35watt quadcore?

Not sure how legit this source is, but it's showing the desktop versions at 65w and 100w TDP
 
http://www.cpu-world.com/news_2013/2013120401_First_AMD_Kaveri_APUs_detailed_A10-7700K_and_A10-7850K.html

Still waiting on benchmarks.
It's nice hearing "IGPU is xxx% better".
But I wonder how much, or how little the actual cpu performance will be compared to the 6800k.
I was tempted to wait for it instead of the 8350 I was just given, but I just couldn't see the chip comparing as well in general CPU performance.

Baltothewolf
Tbh this is the direction we need to be stepping. Why the hell we still have dedicated gpu's are beyond me. The technology is obviously there to combine both in one, so why not do it? I would love to see rigs the size of mitx able to max the future games with no problem...

Well your going to have to wait a while for that to happen brotha. I would always rather have a dedicated GPU than it combined in one little chip with the CPU. Just think about it, if you took the GPU out of the CPU, the CPU typically should run a lot cooler as well as being able to devote its processing power towards other things besides displaying imagery. Basically less load on the CPU. Catch my drift?

Baltothewolf
Tbh this is the direction we need to be stepping. Why the hell we still have dedicated gpu's are beyond me. The technology is obviously there to combine both in one, so why not do it? I would love to see rigs the size of mitx able to max the future games with no problem...

Getting exponentially more powerful IGPU's onboard the CPU is excellent IMO, but I think it would be the death of the enthusiast PC market if they ever got rid of dedicated graphics. Think about it, that would mean that every time you wanted to upgrade your graphics solution, you'd also have to upgrade your CPU. Considering most enthusiasts change GPU's as much as they change their underwear (i.e. at least once a month), it would be an absolutely terrible idea.
 
Not to mention you'd also now lose NVidia as a viable alternative to AMD, unless of course they developed their own version of an x86 processor.
 
I think if we were to migrate to graphics with a smaller footprint, I'd rather see a dedicated socket/port used exclusively by the graphics card and similar to the CPU. Then you could develop cooling solutions that work in tandem with the CPU and the GPU, while saving on the space that most huge graphics cards take up these days.

Inspiration and exploitation.

srtie4k
Getting exponentially more powerful IGPU's onboard the CPU is excellent IMO, but I think it would be the death of the enthusiast PC market if they ever got rid of dedicated graphics.

I think the enthusiast PC market is dying anyway seeing how the cost to keep up is going up along with marginal IPC improvements by both AMD and Intel.  But it does lend to your idea of upgrading graphics more often than CPU-- until the IPC of graphics begins to level out (which has probably already happened since many of us rely on core count and frequency to give us an idea of improvement).  Think about it, we have yet to see a graphics card come with the same specs of a previous and somehow being much faster due to an architectural IPC change in the GPU, or phenomenally improved drivers defeating the purpose of changing the graphics card (which is, BTW, how it is done with consoles).  
 
 
Most PC enthusiasts like the freedom of choosing whatever they like, but they will still gravitate towards whatever is popular and ignore the rest, thereby defeating the purpose of variety to begin with.  If the best/popular was embedded every time, no one would complain; leaving only those that can't afford best/popular that complain.
 
 
Except, per starsmine's response above, die size prevents the best/popular from being embedded with the CPU, thus why we still have dedicated cards.   It would be nice if the Steamroller quad-core was married with the Hawaii GPU, but the total die area would be around 560mm2, or slightly bigger than the Intel 10-core Westmere Xeon E7's.  But to replicate 512-bit interface, there needs to be an 8-channel controller w/DDR4-5000 just to match R9-290X.  
 
 
The CPU+GPU is physically possible yes, memory bandwidth doesn't exist in modular form, unless we embed GDDR5 onto the board for reduced latency.  But who the hell is going to buy what might as well become a high-end console that would easily shoot north of $2000, not just in components, but due to convenience?
 
 
People seem to make the mistake that anything is possible for cheap, that isn't true.  The only way any product is made at all is if we put our money where our mouth is.  The whole reason something doesn't exist in mass production is because the cost to develop it exceeds what people are willing to spend.
 
 
If AMD were to create a super APU for x86, the fact is none of us will afford it and it will be slower than any CFX setup.
 

srtie4kThink about it, that would mean that every time you wanted to upgrade your graphics solution, you'd also have to upgrade your CPU.

Don't we do this anyway?With Intel's new sockets every few years, plus controlling which CPU/chipset combinations have the ability to overclock, we don't have much choice.  If everything was put into one system, what difference would it really make other than come off as a high-end console?
 

srtie4kConsidering most enthusiasts change GPU's as much as they change their underwear (i.e. at least once a month)...

I'm going to have to disagree on the bold.  There is an issue of demographics and perception; not all enthusiasts are vocal, participate in forums like these, let alone are they all wealthy enough to change their systems often.  I think it is a common misconception that people here treat overclockers, DIY'ers, gamers and enthusiasts as synonyms; there are plenty that qualify for some but not all.   I rather think most of us change graphics only as often as every new CPU.
 

srtie4k
Not to mention you'd also now lose NVidia as a viable alternative to AMD, unless of course they developed their own version of an x86 processor.

Unless ARM took off then suddenly the successors to nVidia's "Project Denver" have a future, and I'm sure that is their gamble.  It isn't like whatever technology made is limited to one use; nVidia uses their GPUs for more than just games, so a future ARM APU won't be limited to HTPC and mobile. I do see nVidia trying to push a "Tesla APU" at some point against ARM versions by both AMD and Intel.
 

srtie4kI think if we were to migrate to graphics with a smaller footprint, I'd rather see a dedicated socket/port used exclusively by the graphics card and similar to the CPU.

Same logic could have kept the northbridge dedicated as well, but that has long since been integrated and no one cares about loosing it.
 

srtie4kThen you could develop cooling solutions that work in tandem with the CPU and the GPU, while saving on the space that most huge graphics cards take up these days.

Save even more space and the cost of cooling if we integrate everything together?


I know what you're trying to say, but companies think about their bottom line first,  not consumer preference.  Otherwise Intel would have allowed us to choose any CPU and chipset we like to overclock like Nehalem on back. But despite restriction to a handful of unlocked processors with just two chipsets, no one cares.
 
Freedom is the illusion between those with power and those without; you paint the picture that reduced variety is bad for enthusiasts, but as history has shown thus far, as long as we get the same end-result experience, no one will care how restricted things will get.  It is analogous to letting go of liberties for security.  The future "enthusiast" will believe they have freedom of choice against everyone else, but in fact will have lost freedoms and choices compared to enthusiasts of years ago.
 
We could have one processor with one chipset, as long as it overclocks better than everything else, no one  will care.  Imagine soldering them together, few will complain.  Intel knows this, and eventually both AMD and nVidia will exploit this.
 
I'm sure the future PC will be a console where the one marketed to enthusiasts will have unlocked specs; of course at that point, drivers will improve every game/app/benchmark, there is no reason to change hardware.

The A10-7850K part is not the full die.
 
AMD Kaveri's highest specifications are not;
4 Cores with 1 x 256-bit Flex FPU and 1 x 128-bit MMX/Shuffle.
8 CUs with 512 ALUs, 32 TMUs, 8 ROPs.
128-bit DDR3.

AMD's Kaveri's highest specifications are;
4 Cores with 2 x 256-bit Flex FPU and 1 x 256-bit MMX/Shuffle.
16 CUs with 1024 ALUs, 64 TMUs, 16 ROPs.
128-bit DDR3 + 128-bit DDR3/GDDR5 or 256-bit DDR3/GDDR5. (Wording for the combo is weird)
---
https://i.imgur.com/BLQ9Nuk.jpg
https://i.imgur.com/sXNYo0K.jpg
 
Kaveri's CU being 7.1 mm² aroused no suspicions instead everyone thought it was the norm. 
Tahiti: ~5.5 mm²
Hawaii : ~4.9 mm²
Pitcairn: ~4.4 mm²
---
https://i.imgur.com/fO8I6GO.png
The leak was accurate and it is what is on the Kaveri die.
 
--
28nm SHP is BUILT for density!  It is based on the 22nm IBM SOI node but without the SOI.


So it has the same exact specifications as it;
~100nm for CPP(source to drain length (divide by 4 to get Gate Length))
~80nm for 1x metal layers (double patterning)
 
AMD's custom 22nm node uses only 12 metal layers because of HDL.  IBM SOI is built for high speed so it uses all 15 metal layers.

wolfpack122Seronx, when can we expect to see the mobile Kaveri chips?

With the Kaveri 2.0, full enabled parts?


July to September, general retail availability.  Iceland and Tonga are already shipping so it would make sense.  Iceland and Tonga are Topaz and Amethyst which are the GPUs Kaveri 2.0 can crossfire with.
 
AMD wants to destroy the back to school market with Kaveri 2.0 Mobile, Beema Mobile, Mullins Mobile.