AMD CPU speculation... and expert conjecture

[ohyouknow]

Again once Nintendo gets their titles out the door The system will sale better. Its already doing better then the 360+PS3 at launch time, just wont be as big of a hit as the Wii.

I agree. The original Wii was an extraordinary success since it changed the concept of controller and console interaction. Once games like Zelda and other first party titles hit the Wii U Nintendo will hit its' stride. Get monster hunter released, and the Japanese consumers will jump all over it.

 

[cgner]

Your reasoning is lacking logic. Having more threads does not make the threads that are used slower. If a game uses 2 cores and only 2 cores, it will perform almost exactly the same on a dual-core i3, quad-core i5, and hexa-core i7. AMD does not do as well in *Edit: SOME *games due to how powerful each thread is. Also, I imagine there are very few people who would play a game at 30 fps (minimum, mind you) without seeing that it is 30 fps and call the game "unplayable" because of frame rates.

Pardon me good sir but I meant cores, not threads. The game will obviously perform the same on dual/quad/hexa cores that use same architecture and clocks because it will use equivalent two cores and leave extra ones alone. This is why everyone complains that more games should use 4 cores if they are heavy on CPU usage. Going back do DoW 2 example, it requires heavy CPU processing but only uses one single core when CPU offers 6.... If it used 2 cores, FPS drops would be far less dramatic.

This is my point, there is no need for 6-8 cores in a gaming CPU when games only use 2 of those cores. Channel that processing power into 2-4 cores instead of splitting it. (NOT saying its easy)

 

[sarinaide]

um.... no. Like people mentioned here many times, AMD chips are inferior because of many threads. Vast majority of games does not use more than 2 cores. Even not so old titles like Dawn of War 2 show major FPS drop to 30-35 on 1100T @ 4ghz, when there is a battle on the screen because they are single threaded.

I don't know so much, I do agree about the pointlessness of 8 threads, but then again AMD are not going to make just a CPU for gaming, when they designed it it was designed for multi faceted computing so in instances the 8 threads are better than say just having 2 threads. If thread count was the issue then a i5 will be slower than a i3, but the i7 will be the slowest of the lot and this is most certainly on every level proven wrong.

In many testings we did between Deneb, Thuban, Zambezi and Vishera what we found was the limiting factor was memory latencies. Deneb and Thuban while having weaker IMC's had lower latency relative to Zambezi and that is why at times both the older Phenoms gamed better than the Zambezi's due to irratic stuttering even when the Phenoms used DDR3 12800 and the Zambezi DDR3 17000. Roll on Vishera and AMD made massive strides here, I am able to get the exact same memory scores testing read/write and latency as I do on a similar i7 950 setup. This is the reason why in general gaming notably on demanding titles like BF3, COD BO2, Crysis 3, Far Cry 3. In all of these titles the Vishera is smooth and fluid. BF3 alone the the FX8350 was shown to be the best CPU in the Tech Report FPS bench, lowest frame latencies and in general comparing all these processors the Vishera parts are perfect gaming alternatives to Intel.

Roll onto Steamroller and AMD's focus is heavily on reducing Cache and Memory latencies targeting 40% reductions in both area's this should affect the overall performance substantially, that with wider and deeper instruction execution's we could say it will be the biggest leap AMD have made yet but where it will sit is speculative, I like to think somewhere near to or around or equal to intel's 1155 offerings, if that happens AMD is very much back in the game as a serious threat to the de facto Intel position.

I am also going to agree a bit with De5 Roy here and say that AMD's intent is also to improve mobility segment parts, they were not so prioritized in that segment hence why Trinity Mobile is underwhelming relative to desktop Trinity, I can see AMD focusing a bit here as they need that market and to be fair a APU is better suited to that market than budget desktop. Whatever it may be the A-series remains a very impressive product that will improve through each generation.

I replaced my Intel iTX build with a new ASRock A85E mobo which will release soon and a A10 5800k all in a SUGO7, the absence of the need to use a discrete card lowered, heat, noise and power considerably and Runs on the 72" television we tested it on just fine and dandy. Another testament to AMD focusing outside the x86 box.

I am a little distressed at the Intel comments that DT will die soon, I don't think that is correct. I liken Intel to Man City of the CPU world, they have the money and backing to buy the best but at grass roots they lack the development to ensure internal growth, Intel have the money and resources to plunder x86 into the ground but beyond x86 Intel have stark nothing and trying to force x86 onto smartphones aggressively running down the die's while abandoning its bread and butter market is more indicative of desperation when they don't need to be. The desktop market or PC market will die if AMD disappears, they are the more expansive thinkers in the PC market every good idea is coming from AMD, they just need the money to help drive that idea. While AMD only offer good enough x86 the focus is not heavily on that but adding other paradigms to computing using the x64 instruction set to drive HSA this in not to long will pay dividends as Intel have shown no interest in this approach, as we have seen in the limited HSA scenarios AMD delivers 3-4x faster performance than intels fastest x86 processor, this should be very alarming to Intel as more developers and HSA partners are being signed up, non more important than Microsoft.

 

[JAYDEEJOHN]

This is a bit off track here, but margins run companies, and the PC market isnt keeping up, unless its highend, big dollars.
AMD has and can do more with less, as they dont have the Intel business model like Intels.
Can AMD carry the PC market to a greater degree than they are currently doing is another question, decent margins, still a decent market size etc.
Gaining marketshare on DT may not be so lucrative unless highend, and AMD has never shown the ability to mass produce enough, even for a shinking DT market, besides their efforts towards SoCs and mobile etc

 

[truegenius]

according to my rough calculations
as each k10 core is 17.7mm² including the L2 cache and power gating in 32nm llano. so tweaked (no l3 and increased l2) 4x k10 cores will take 100-150mm² die , which is 1.5-2 times less than that of a8-3870k's original die and if it can be clocked in 3.5-4.5ghz range woth improved imc then then it can clearly compete with i3 and locked i5
but not to forget that this much less die means more good chips and less production cost per die so they can sell them in under $100 department so none of the intel offerings can offer that much value for money

and performance
eg a x4 k10 @3.8ghz vs a10-5800k (without turbo at stock)
x4 k10 will perform 40% better in multithreaded tasks and they will perform same in single threaded tasks

power consumption
tweked x4 k10 may take around 65w at 3.8ghz according to my rough calculations

 

[jdwii]

according to my rough calculations
as each k10 core is 17.7mm² including the L2 cache and power gating in 32nm llano. so tweaked (no l3 and increased l2) 4x k10 cores will take 100-150mm² die , which is 1.5-2 times less than that of a8-3870k's original die and if it can be clocked in 3.5-4.5ghz range woth improved imc then then it can clearly compete with i3 and locked i5
but not to forget that this much less die means more good chips and less production cost per die so they can sell them in under $100 department so none of the intel offerings can offer that much value for money

and performance
eg a x4 k10 @3.8ghz vs a10-5800k (without turbo at stock)
x4 k10 will perform 40% better in multithreaded tasks and they will perform same in single threaded tasks

power consumption
tweked x4 k10 may take around 65w at 3.8ghz according to my rough calculations

Hold on their are you also getting rid of IGPU?

 

[jdwii]

This is a bit off track here, but margins run companies, and the PC market isnt keeping up, unless its highend, big dollars.
AMD has and can do more with less, as they dont have the Intel business model like Intels.
Can AMD carry the PC market to a greater degree than they are currently doing is another question, decent margins, still a decent market size etc.
Gaining marketshare on DT may not be so lucrative unless highend, and AMD has never shown the ability to mass produce enough, even for a shinking DT market, besides their efforts towards SoCs and mobile etc

20% of Amd's money next year will come from consoles. And i think 5 % will come from the GPU market. So what about 75% well that will be from client +server CPU's/APU's.

Now with haswell coming out and Amd's small update in piledriver and APU i doubt Amd will gain anything from the Server market, Probably even decline in the APU market until Amd can get 28nm and steamroller out.


In other words expect nothing huge in their market cap until steamroller and lets hope that gets them somewhere in the server market(Module design was never for gaming).

Still hoping to see OpenCL get used more.

 

[de5_Roy]

More Richland details leak, six parts confirmed
http://www.fudzilla.com/home/item/30264-more-richland-details-leak-six-parts-confirmed

 

[truegenius]

Hold on their are you also getting rid of IGPU?

yes
because am3+ boards are familiar with k10 cores so it will not take much effort to do that
it will be a die shrink only with tweaks

-> they can conitnue their apu line for complete (cpu+gpu) solution at cheap price
-> x4 k10 32nm will take care of i3 and locked i5
-> and fx8 will tackle i5 and i7:
-> and leave $250+ dt market to intel's i7 extreme

so in this way they can remain competitive in many departments with good profit margins

x4 980 is still competing with fx6300
http://www.anandtech.com/bench/Product/362?vs=699
http://www.anandtech.com/bench/Product/362?vs=700

so die shrink of 980 will cost less production , better performance per watt, more profit

they know k10 very well so very less R&D cost and they can produce/ship 32nm k10 for am3+ in less than 6months :/

 

[cgner]

I don't know so much, I do agree about the pointlessness of 8 threads, but then again AMD are not going to make just a CPU for gaming, when they designed it it was designed for multi faceted computing so in instances the 8 threads are better than say just having 2 threads. If thread count was the issue then a i5 will be slower than a i3, but the i7 will be the slowest of the lot and this is most certainly on every level proven wrong.

In many testings we did between Deneb, Thuban, Zambezi and Vishera what we found was the limiting factor was memory latencies. Deneb and Thuban while having weaker IMC's had lower latency relative to Zambezi and that is why at times both the older Phenoms gamed better than the Zambezi's due to irratic stuttering even when the Phenoms used DDR3 12800 and the Zambezi DDR3 17000. Roll on Vishera and AMD made massive strides here, I am able to get the exact same memory scores testing read/write and latency as I do on a similar i7 950 setup. This is the reason why in general gaming notably on demanding titles like BF3, COD BO2, Crysis 3, Far Cry 3. In all of these titles the Vishera is smooth and fluid. BF3 alone the the FX8350 was shown to be the best CPU in the Tech Report FPS bench, lowest frame latencies and in general comparing all these processors the Vishera parts are perfect gaming alternatives to Intel.

Roll onto Steamroller and AMD's focus is heavily on reducing Cache and Memory latencies targeting 40% reductions in both area's this should affect the overall performance substantially, that with wider and deeper instruction execution's we could say it will be the biggest leap AMD have made yet but where it will sit is speculative, I like to think somewhere near to or around or equal to intel's 1155 offerings, if that happens AMD is very much back in the game as a serious threat to the de facto Intel position.

I am also going to agree a bit with De5 Roy here and say that AMD's intent is also to improve mobility segment parts, they were not so prioritized in that segment hence why Trinity Mobile is underwhelming relative to desktop Trinity, I can see AMD focusing a bit here as they need that market and to be fair a APU is better suited to that market than budget desktop. Whatever it may be the A-series remains a very impressive product that will improve through each generation.

I replaced my Intel iTX build with a new ASRock A85E mobo which will release soon and a A10 5800k all in a SUGO7, the absence of the need to use a discrete card lowered, heat, noise and power considerably and Runs on the 72" television we tested it on just fine and dandy. Another testament to AMD focusing outside the x86 box.

I am a little distressed at the Intel comments that DT will die soon, I don't think that is correct. I liken Intel to Man City of the CPU world, they have the money and backing to buy the best but at grass roots they lack the development to ensure internal growth, Intel have the money and resources to plunder x86 into the ground but beyond x86 Intel have stark nothing and trying to force x86 onto smartphones aggressively running down the die's while abandoning its bread and butter market is more indicative of desperation when they don't need to be. The desktop market or PC market will die if AMD disappears, they are the more expansive thinkers in the PC market every good idea is coming from AMD, they just need the money to help drive that idea. While AMD only offer good enough x86 the focus is not heavily on that but adding other paradigms to computing using the x64 instruction set to drive HSA this in not to long will pay dividends as Intel have shown no interest in this approach, as we have seen in the limited HSA scenarios AMD delivers 3-4x faster performance than intels fastest x86 processor, this should be very alarming to Intel as more developers and HSA partners are being signed up, non more important than Microsoft.

Yep. The problem was that most people with workstations still go with quad or even hexa core i7's since they tend to deliver better performance overall in workstations.

 

[MU_Engineer]

This threw me off a bit; http://cpuboss.com/cpus/Intel-Core-i7-3770K-vs-AMD-FX-8350#differences

FX-8350 apparently has lower overall power consumption than the i7 3770K? I really would have thought otherwise going by all the reviews that say Piledriver is such a power hog.

That website is THE worst computer-related website I have ever seen. They at best had one benchmark listed, which was Passmark. They then continued to "rate" how good a CPU was by assessing things like amount of cache it has, clock speed, and how long it has been since it was released. Oh, and a lot of the information was incorrect or missing as well. There were some real doozys in there like the FX-8150 with 1 MB of L3 cache and the single-core Opteron 6300s.

In many testings we did between Deneb, Thuban, Zambezi and Vishera what we found was the limiting factor was memory latencies. Deneb and Thuban while having weaker IMC's had lower latency relative to Zambezi and that is why at times both the older Phenoms gamed better than the Zambezi's due to irratic stuttering even when the Phenoms used DDR3 12800 and the Zambezi DDR3 17000. Roll on Vishera and AMD made massive strides here, I am able to get the exact same memory scores testing read/write and latency as I do on a similar i7 950 setup. This is the reason why in general gaming notably on demanding titles like BF3, COD BO2, Crysis 3, Far Cry 3. In all of these titles the Vishera is smooth and fluid. BF3 alone the the FX8350 was shown to be the best CPU in the Tech Report FPS bench, lowest frame latencies and in general comparing all these processors the Vishera parts are perfect gaming alternatives to Intel.

Roll onto Steamroller and AMD's focus is heavily on reducing Cache and Memory latencies targeting 40% reductions in both area's this should affect the overall performance substantially, that with wider and deeper instruction execution's we could say it will be the biggest leap AMD have made yet but where it will sit is speculative, I like to think somewhere near to or around or equal to intel's 1155 offerings, if that happens AMD is very much back in the game as a serious threat to the de facto Intel position.

I wonder if AMD will finally bump the L3 speeds in the Steamroller parts. They have been in that 1.8-2.2 GHz range ever since the first L3-equipped AMD chips came out in 2007. The problem is that the rest of the chip has sped up while the IMC/L3 has not in many cases. The original parts were 1.8-2.2 GHz L3 feeding four cores in the low 2 GHz range. Now that same speed L3 is feeding 8 cores at close to 4 GHz. Many of us have known that the fairly low speeds for the L3 has been holding back the chips somewhat (as evidenced by performance increases when bumping IMC/L3 speeds up to say 2.6 GHz.)

I am a little distressed at the Intel comments that DT will die soon, I don't think that is correct. I liken Intel to Man City of the CPU world, they have the money and backing to buy the best but at grass roots they lack the development to ensure internal growth, Intel have the money and resources to plunder x86 into the ground but beyond x86 Intel have stark nothing and trying to force x86 onto smartphones aggressively running down the die's while abandoning its bread and butter market is more indicative of desperation when they don't need to be. The desktop market or PC market will die if AMD disappears, they are the more expansive thinkers in the PC market every good idea is coming from AMD, they just need the money to help drive that idea. While AMD only offer good enough x86 the focus is not heavily on that but adding other paradigms to computing using the x64 instruction set to drive HSA this in not to long will pay dividends as Intel have shown no interest in this approach, as we have seen in the limited HSA scenarios AMD delivers 3-4x faster performance than intels fastest x86 processor, this should be very alarming to Intel as more developers and HSA partners are being signed up, non more important than Microsoft.

Intel depends on x86 being *the* dominant ISA in some platform that is a growth market able to support high ASPs and has continual revenue growth year after year. Intel's real place in the market is that they can produce the most technically-complex parts on the most technically-advanced manufacturing process in large quantities. The price of keeping on the cutting edge of fabs and semiconductor R&D keeps increasing dramatically with each process shrink. Thus they need to keep average selling prices and total revenues very high in order to finance the fabs and R&D. Anything that can reduce the ASPs or revenues is very bad for Intel.

Intel is also dependent on the x86 ISA being de facto required to do a specific workload because Intel controls all of the x86 licensing and most of the x86 patents. Intel needs x86 to be "required" in some platform that has no alternative in order to guarantee themselves a large market and prevent lower-cost competitors with much leaner cost structures from moving in and underselling Intel. This is why you have seen Intel try to enter many other markets other than desktop and notebook CPUs in the past few years- GPUs and tablets/phones being the big ones. Intel was trying to shoehorn in on emerging markets and get them locked in on x86 since the Windows/Intel monopoly on computing that guaranteed Intel a big market is now broken.

Intel as we know it will be done for if the pundits are correct and most people do their computing on cheap, low-powered phones/tablets/cheap laptops/dumb terminals running a myriad of different OSes on several different CPU architectures. People want "good enough and cheap" and somebody putting some ARM or MIPS or whatever cores and ASICs on an older and now cheap bulk silicon process at a low-cost foundry will be fast enough to get the job done and be able to undersell Intel handily. Intel won't go away, but they will shrink quite a bit and most likely end up like a combination of TI and TSMC. They would have fabs similar to the other foundry conglomerates still out there and sell a lot of low-margin miscellaneous ICs and SoCs and do some design work for themselves and others for modest margins.

 

[JAYDEEJOHN]

Sounds like IBM in a way.
They could change their business model, as markets bare.
They have the ability to do the most, but also have the most to lose and to change.

 

[MU_Engineer]

Sounds like IBM in a way.
They could change their business model, as markets bare.
They have the ability to do the most, but also have the most to lose and to change.

It does sort of sound like IBM. They essentially owned computing from the mainframe all the way down to the PC. They got undercut and don't even make any computers except for servers any more. Ditto with typewriters, clocks, and POS terminals (cash registers.) They have their niche with big iron servers and big iron software. It isn't too big of a stretch to see Intel going down that road.

Now AMD on the other hand is a different story. They are now a design-only firm and don't have the market inertia that a huge firm like Intel and IBM have. I can see them doing well as they have always been the ones pushing the value angle. They are also a design-only firm and don't have to pay the massive expenses Intel does to keep ever-more expensive fabs up to date. So they can tolerate much lower revenues and margins and still do okay. However they may also go bankrupt in the next few years as their financial situation has been pretty tenuous in the past half-dozen years or so. They mortgaged themselves to the hilt to buy ATi (which they needed to do) but the unusually severe and unusually long recession kicked them right in the nuts at about the worst time.

 

[JAYDEEJOHN]

Ive a feeling a one horse town isnt what the others want, and until/unless ARM comes into a larger scenario, AMD will be the other choice still, and possibly even more so

 

[sarinaide]

That website is THE worst computer-related website I have ever seen. They at best had one benchmark listed, which was Passmark. They then continued to "rate" how good a CPU was by assessing things like amount of cache it has, clock speed, and how long it has been since it was released. Oh, and a lot of the information was incorrect or missing as well. There were some real doozys in there like the FX-8150 with 1 MB of L3 cache and the single-core Opteron 6300s.



I wonder if AMD will finally bump the L3 speeds in the Steamroller parts. They have been in that 1.8-2.2 GHz range ever since the first L3-equipped AMD chips came out in 2007. The problem is that the rest of the chip has sped up while the IMC/L3 has not in many cases. The original parts were 1.8-2.2 GHz L3 feeding four cores in the low 2 GHz range. Now that same speed L3 is feeding 8 cores at close to 4 GHz. Many of us have known that the fairly low speeds for the L3 has been holding back the chips somewhat (as evidenced by performance increases when bumping IMC/L3 speeds up to say 2.6 GHz.)

I am a little distressed at the Intel comments that DT will die soon, I don't think that is correct. I liken Intel to Man City of the CPU world, they have the money and backing to buy the best but at grass roots they lack the development to ensure internal growth, Intel have the money and resources to plunder x86 into the ground but beyond x86 Intel have stark nothing and trying to force x86 onto smartphones aggressively running down the die's while abandoning its bread and butter market is more indicative of desperation when they don't need to be. The desktop market or PC market will die if AMD disappears, they are the more expansive thinkers in the PC market every good idea is coming from AMD, they just need the money to help drive that idea. While AMD only offer good enough x86 the focus is not heavily on that but adding other paradigms to computing using the x64 instruction set to drive HSA this in not to long will pay dividends as Intel have shown no interest in this approach, as we have seen in the limited HSA scenarios AMD delivers 3-4x faster performance than intels fastest x86 processor, this should be very alarming to Intel as more developers and HSA partners are being signed up, non more important than Microsoft.

Intel depends on x86 being *the* dominant ISA in some platform that is a growth market able to support high ASPs and has continual revenue growth year after year. Intel's real place in the market is that they can produce the most technically-complex parts on the most technically-advanced manufacturing process in large quantities. The price of keeping on the cutting edge of fabs and semiconductor R&D keeps increasing dramatically with each process shrink. Thus they need to keep average selling prices and total revenues very high in order to finance the fabs and R&D. Anything that can reduce the ASPs or revenues is very bad for Intel.

Intel is also dependent on the x86 ISA being de facto required to do a specific workload because Intel controls all of the x86 licensing and most of the x86 patents. Intel needs x86 to be "required" in some platform that has no alternative in order to guarantee themselves a large market and prevent lower-cost competitors with much leaner cost structures from moving in and underselling Intel. This is why you have seen Intel try to enter many other markets other than desktop and notebook CPUs in the past few years- GPUs and tablets/phones being the big ones. Intel was trying to shoehorn in on emerging markets and get them locked in on x86 since the Windows/Intel monopoly on computing that guaranteed Intel a big market is now broken.

Intel as we know it will be done for if the pundits are correct and most people do their computing on cheap, low-powered phones/tablets/cheap laptops/dumb terminals running a myriad of different OSes on several different CPU architectures. People want "good enough and cheap" and somebody putting some ARM or MIPS or whatever cores and ASICs on an older and now cheap bulk silicon process at a low-cost foundry will be fast enough to get the job done and be able to undersell Intel handily. Intel won't go away, but they will shrink quite a bit and most likely end up like a combination of TI and TSMC. They would have fabs similar to the other foundry conglomerates still out there and sell a lot of low-margin miscellaneous ICs and SoCs and do some design work for themselves and others for modest margins.

CACHE:

I heard that the Cache speed will be almost doubled but its all speculative.

X86:

Yes people use x64 based tabs, pads and phones but honestly working on them is still very inappropriate, so the DT and Notebook still remain a better source of work platforms.

DT will not die because it still garners a majority market, the majority market and the gaming and enthusiast market is expanding not decreasing, then you have other tech involved in x86 micro architecture it is to prominent to fall away.

 

[Chad Boga]

Intel as we know it will be done for if the pundits are correct and most people do their computing on cheap, low-powered phones/tablets/cheap laptops/dumb terminals running a myriad of different OSes on several different CPU architectures. People want "good enough and cheap" and somebody putting some ARM or MIPS or whatever cores and ASICs on an older and now cheap bulk silicon process at a low-cost foundry will be fast enough to get the job done and be able to undersell Intel handily.

This assumes that the pause we are currently seeing in software applications requiring ever more demanding processors, continues for years, if not decades.

Do you really think that will be the case?

 

[Chad Boga]

Ive a feeling a one horse town isnt what the others want, and until/unless ARM comes into a larger scenario, AMD will be the other choice still, and possibly even more so

Who are these others?

That kind of sentiment means nothing, and unless the second supplier can deliver reasonably good products, a one horse town could easily come about, even though not desirable for most people.

 

[oGhastly]

Well it won't come out for a while so I'm going to guess AMD is taking their time. I'd rather wait then have another disappointment.

 

[Cazalan]

Some Richland clock speeds leaked.

4.1Ghz/4.4Ghz for top A10-6800K (+300/+200Mhz)

Decent bump. More than Intel has been doing lately.

http://www.cpu-world.com/news_2013/2013012601_Specifications_of_upcoming_Richland_APUs.html

 

[jdwii]

Some Richland clock speeds leaked.

4.1Ghz/4.4Ghz for top A10-6800K (+300/+200Mhz)

Decent bump. More than Intel has been doing lately.

http://www.cpu-world.com/news_2013/2013012601_Specifications_of_upcoming_Richland_APUs.html

Nothing major just like i thought less of improvement then Llano-Trinity and with haswell coming out, its not good steamroller got delayed.

Haswell is not like Ivy, haswell is going to be a tock not a tick.

Why are they still using 32nm to make the next APU seem better by having people wait on 28nm, even wonder if that's the reason why the A10 trinty was clocked at 3.8Ghz instead of 4.1Ghz to begin with.

But who knows i guess maybe just maybe i'll be surprised

 

[esrever]

Nothing major just like i thought less of improvement then Llano-Trinity and with haswell coming out, its not good steamroller got delayed.

Haswell is not like Ivy, haswell is going to be a tock not a tick.

Why are they still using 32nm to make the next APU seem better by having people wait on 28nm, even wonder if that's the reason why the A10 trinty was clocked at 3.8Ghz instead of 4.1Ghz to begin with.

But who knows i guess maybe just maybe i'll be surprised

Richland is a stopgap for the OEMs is what I'd imagine.

 

[anxiousinfusion]

If we keep seeing CPU clock jumps like this - Llano 3.0ghz, Trinity 3.8ghz, Richland 4.1ghz - I can only extrapolate that Steamroller APUs would be around ~ 4.4ghz.

Did AMD ever end up using that resonant clock mesh that they were talking about a few months back?

 

[de5_Roy]

^^ the second paragraph (ed. comment)
http://www.tomshardware.com/reviews/fx-8350-vishera-review,3328.html

 

[anxiousinfusion]

Ah thanks. I'd been following that but word on the technology became suspiciously quiet since the release of Piledriver.

 

[sarinaide]

Nothing major just like i thought less of improvement then Llano-Trinity and with haswell coming out, its not good steamroller got delayed.

Haswell is not like Ivy, haswell is going to be a tock not a tick.

Why are they still using 32nm to make the next APU seem better by having people wait on 28nm, even wonder if that's the reason why the A10 trinty was clocked at 3.8Ghz instead of 4.1Ghz to begin with.

But who knows i guess maybe just maybe i'll be surprised

What does clock bump have to do with overall performance, Richland will be on a more mature process by that point so lower power, along with metal level evolution of the Piledriver cores will probably see a greater improvement than Llano to Trinity and then we go to the iGPU component which will be vastly stronger.

Tickity Tockity, I am sure Haswell will be fantastic, but not nearly as fantastic as its buttered up to be. Some say around 10% faster x86 performance a very fast IMC and a iGPU about 20-25% faster than HD4000 on the GT3 part, that's about as fast as Llano's highest end part or there about's but will be the same old stutter fest. Certainly a long way behind Trinity and going to get monstrously further behind.

The 32nm node is still not perfected from AMD's front but lets just wait and see what a GCN component along with a maturing 32nm offers. The power issue is getting tedious, 8350 significantly lowered peak load power over the 8150 while being clocked well higher, if it were a 3.3ghz part the power probably would have been much lower and that was on the exact same process. Its a case of being overly pedantic about AMD not looking to drive power numbers down by 50% per release. I just think nobody is ever going to be happy until AMD delivers a faster CPU on like a 10th of the resources, maybe we should all just take a step back and think think, they are doing a pretty sound job under immense pressure. AMD CPU's are not compared against Intel CPU"s, they are compared against its previous generation, so from that aspect Vishera and Trinity were notably better than Zambezi and Llano.