AMD CPU speculation... and expert conjecture

[JAYDEEJOHN]

OK
Many of the cpu improvements are being held in check by gpu power usage, mainly on the Intel side.
Their cpu is smaller, their igpu just got alot bigger, and draws somewhat more current , despite process advantage.
Now, AMD has better solutions coming soon for their gpus, as well as power management, and, it isnt only BW holding back the gpus, its also cpu as well.
Strap a AMD igpu into a Intel HSW sans their igpu solution, making AMD that solution, and your fps would most likely go up some.
Intels cpu IPC is crawling forwards, yet AMDs has more room to grow/gain.
Combine this with having to increase power and die space to Intels solution, where they already have power management down, and supply their gpu with plenty, and is why we see such good results at lower resolutions, and the only way forwards for Intel is, either sacrifice on cpu (wont happen) or make huge igpu improvements, which is yet to be seen in HSW.
Its a completely different story with AMD, where we know their current solutions are hindered in several ways, cpu to small extent, power management, and BW, which is an easier fix of all three.

 

[skitz9417]

i jutst hope that amd an get into the cpu game and i also hope steamroller is coming at the end of the year

 

[skitz9417]

and amd is on track with steamrolller http://api.viglink.com/api/click?format=go&key=7777bc3c17029328d03146e0ed767841&loc=http%3A%2F%2Fwww.overclock.net%2Ft%2F1377248%2Fxbitlabs-amd-we-are-on-track-with-steamroller-micro-architecture-in-2013%2F0_100&v=1&libid=1364939840220&out=http%3A%2F%2Fwww.xbitlabs.com%2Fnews%2Fcpu%2Fdisplay%2F20130331080217_AMD_We_Are_On_Track_With_Steamroller_Micro_Architecture_in_2013.html&ref=http%3A%2F%2Fwww.overclock.net%2Ff%2F225%2Fhardware-news&title=%5Bxbitlabs%5D%20AMD%3A%20We%20Are%20On%20Track%20With%20Steamroller%20Micro-Architecture%20in%202013.&txt=Source&jsonp=vglnk_jsonp_13649398632862

 

[palladin9479]

Spoiler

not without access to confidential engineering information

there are many books regarding 8085/8086.......etc cpu (old) arch

Anyhow the take away from all this is to stop using the word "IPC". You can simply say processing efficiency, or single core capacity.

Performance Per Clock sounds better and sounds same as instruction per clock

Also do not use clock as a measurement of anything involving two different CPU designs.

PPC per ghz i.e, PPCPG would be better in compairing performance of different arch
(PPCPG = PPC/clock speed) (a term coined by me no need to thank me )
it will be like dimps in arm's arch comparision to give a general idea of what kind of performance difference we can expect at same speed and core count
example : 3.5 for arm cortex-a15, 3.3 for qualcomm krait, 2.5 for cortex-a9 etc
then we only need to multiply this PPCPG with clock speed and core counts to have full capacity of a cpu

The two best metrics are per cost and per energy used as they describe what the user / integrator is paying for that capacity.

PPC per watt
i.e, PPCPW (another term coined by truegenius no need to thank me again )

Another thing is if they are able to create
a APU that has cores on par with an i5 2500K and strap on decent
graphics at a better price point, They would have a hot seller.

it is too much of an expactation
there is no way that steamy can match sb performance
if steamy can match nehalem in PPCPG (performance per ghz) for equal number of threads then steamy will rock

i expect performance (PPCPG) in range of wolfdale to nehalem

Again clock rates between different uArch's are nearly meaningless. It's only performance per cost or per unit of energy used. That's all there is and all there ever has been. As for the confidential information, that went way over your head. The internal micro-instructions and how their processed is confidential. The 8086 didn't use micro-instructions, none of the x86 CPU's did until the Pentium and none did ILP until the Pentium Pro.

 

[8350rocks]

Spoiler

not without access to confidential engineering information

there are many books regarding 8085/8086.......etc cpu (old) arch

Anyhow the take away from all this is to stop using the word "IPC". You can simply say processing efficiency, or single core capacity.

Performance Per Clock sounds better and sounds same as instruction per clock

Also do not use clock as a measurement of anything involving two different CPU designs.

PPC per ghz i.e, PPCPG would be better in compairing performance of different arch
(PPCPG = PPC/clock speed) (a term coined by me no need to thank me )
it will be like dimps in arm's arch comparision to give a general idea of what kind of performance difference we can expect at same speed and core count
example : 3.5 for arm cortex-a15, 3.3 for qualcomm krait, 2.5 for cortex-a9 etc
then we only need to multiply this PPCPG with clock speed and core counts to have full capacity of a cpu

The two best metrics are per cost and per energy used as they describe what the user / integrator is paying for that capacity.

PPC per watt
i.e, PPCPW (another term coined by truegenius no need to thank me again )

Another thing is if they are able to create
a APU that has cores on par with an i5 2500K and strap on decent
graphics at a better price point, They would have a hot seller.

it is too much of an expactation
there is no way that steamy can match sb performance
if steamy can match nehalem in PPCPG (performance per ghz) for equal number of threads then steamy will rock

i expect performance (PPCPG) in range of wolfdale to nehalem

Again clock rates between different uArch's are nearly meaningless. It's only performance per cost or per unit of energy used. That's all there is and all there ever has been. As for the confidential information, that went way over your head. The internal micro-instructions and how their processed is confidential. The 8086 didn't use micro-instructions, none of the x86 CPU's did until the Pentium and none did ILP until the Pentium Pro.

Very true, and x64 instructions weren't out until 2003 when AMD launched the Athlon 64; the following year, in 2004, Intel released it's 64 bit pentium line up.

 

[drinvis]

AMD needs to get the performance/power improved to a much better extent.If they are able to get more performance in a limited power they would be able to compete better in notebook space.They should come out with at least some kind of reply to the mobile i5 processors in terms of processing so that they can compete on higher price points as well.GPU side of AMD is quite good.For sure they have to balance power for CPU and GPU on an mobile APU with such a limited power envelope.Trinity mobile processors were good with power consumption but they need more performance.I wish Kaveri mobile parts are able to take on haswell mobile i5s or better directly.They need to get a lot more design wins than they got with trinity mobile parts.Notebook sector has way more revenue than desktop sector,that is where AMD can make money on.Also from processor side of things majority of the revenues for AMD comes from APUs only.Server side for AMD might get a lot better with steamroller server parts.

 

[JAYDEEJOHN]

My speculation is, Richland is a gap filler, to compete with HSW, where AMD maintains its gfx lead, and maybe a nudge on their side to getting a tad closer in cpu perf.
The problem is, the longer HSW is delayed, it will have to then compete with Kaveri, which will really set up the old "its only Intel gfx" mentality once again.
Just my 2c

 

[palladin9479]

SR seems to be mostly an upgrade to the decoder and caching subsystem. It just happens that those are the two area's that crippled BD's performance and resulted in the processing units stalling out. I'm not expecting much from the die shrink. It'll be more energy efficient so you clock it up a bit (clock speed matters for CPUs within the same uArch) but otherwise I expect them to take another cycle to do any real improvements based on a smaller die. The better decoders and caching should create a sizable performance increase on their own.

 

[JAYDEEJOHN]

I also see process maturity helping here.
Having more free distributive power and a slightly better power management helps as well, especially for gpu, where BW isnt a problem, but for mobile, it will help, where BW isnt seen as much of a problem due to lower res, and not talking about higher end solutions on mobile, where youd find a discrete more than likely anyways, or want one.
I agree, we are a gen out before most here will be taking a hard look at some of those solutions

 

[jdwii]

I still don't see why everybody thinks Anandtech is so bias towards Intel.

Its not that far fetched just read their reviews and their so called conclusions.

Okay this is how its going to go down. AMD going to say that steamroller going to get 30% performance increased. Then when the product comes out you will see that they just up the clock speed and tweak some things. AMD lost all their credibility after bulldozer and they shouldn't even put a percentage on how much better steamroller is then piledriver. All i care about is ipc! I couldn't careless about a cpu hitting 4.5ghz stock. I can overclock the cpu myself, of course it would be sweet hitting 4.5ghz on 125 watts but still won't make the cpu worth it. Hopefully there focus is on make a huge jump in ipc rather then adding more mhz to clock speeds cause that will only get them so far.

Based on what not only that Amd still never said anything but a 15% improvement in performance per watt per gen.

 

[JAYDEEJOHN]

Not only that, but average Joe doesnt OC either, so yes, for some, stock doesnt matter, it end perf that counts, thats like maybe 2% of the market.
It matters to AMD to have good stock perf, and in doing so, ensures future capabilities

 

[mayankleoboy1]

I can already read the Haswell reviews :

"One year after, and the CPU perf is incremental at best. The iGPU, much improved, still has lots of room to grow. It still cant compete with Trinity APU's from AMD. If you have to build a new system, and are going to add adGPU, we wholeheartedly recommend the 4770K. If you want to build a "good enough" PC, we think that a Richland APU system would be better. "

 

[esrever]

I can already read the Haswell reviews :

"One year after, and the CPU perf is incremental at best. The iGPU, much improved, still has lots of room to grow. It still cant compete with Trinity APU's from AMD. If you have to build a new system, and are going to add adGPU, we wholeheartedly recommend the 4770K. If you want to build a "good enough" PC, we think that a Richland APU system would be better. "

You are giving reviewers too much unbias. They are going to read. OMG AMAZING INTEL CPU, LOOK AT THAT GPU PERFORMANCE COMPARED TO BEFORE AND IF YOU WANT TO DO WORK, CPU PERFORMANCE IS SO MUCH HIGHER THAN AMD ON ITUNES.

And Richland review will be: 6/10 "its ok" - Anandtech.

 

[griptwister]

I can already read the Haswell reviews :

"One year after, and the CPU perf is incremental at best. The iGPU, much improved, still has lots of room to grow. It still cant compete with Trinity APU's from AMD. If you have to build a new system, and are going to add adGPU, we wholeheartedly recommend the 4770K. If you want to build a "good enough" PC, we think that a Richland APU system would be better. "

You are giving reviewers too much unbias. They are going to read. OMG AMAZING INTEL CPU, LOOK AT THAT GPU PERFORMANCE COMPARED TO BEFORE AND IF YOU WANT TO DO WORK, CPU PERFORMANCE IS SO MUCH HIGHER THAN AMD ON ITUNES.

And Richland review will be: 6/10 "its ok" - Anandtech.

I'm with esrever on this one! HAHAHA!

 

[mayankleoboy1]

I can already read the Haswell reviews :

"One year after, and the CPU perf is incremental at best. The iGPU, much improved, still has lots of room to grow. It still cant compete with Trinity APU's from AMD. If you have to build a new system, and are going to add adGPU, we wholeheartedly recommend the 4770K. If you want to build a "good enough" PC, we think that a Richland APU system would be better. "

You are giving reviewers too much unbias. They are going to read. OMG AMAZING INTEL CPU, LOOK AT THAT GPU PERFORMANCE COMPARED TO BEFORE AND IF YOU WANT TO DO WORK, CPU PERFORMANCE IS SO MUCH HIGHER THAN AMD ON ITUNES.

And Richland review will be: 6/10 "its ok. Needs 2 generations to get it right." - Anandtech.

:lol:
+1

(Edit : Added the underlined part to original comment.)

 

[sarinaide]

"did you think that there might be reasons why core i3's hd4k performs better than core i7's other than improved drivers? i'll leave that to you to figure it out."
But, as we go larger, despite process, we lose the smaller, aimed more at gpu divy of power. Or more efficient, as in GCN or latter iterations, coming soon.
As well as catch up from better power resourcing, and also better cpu push/IPC/PPC whatever you want to call it.

Utilized cores yeah heard that pony before and drivers are not part biased HD4000 drivers are across the board.

Skyrim @ 1268 x 768 LS

5800K vs 3225 or better yet HD7660D vs HD4000

http://techreport.com/review/23662/amd-a10-5800k-and-a8-5600k-trinity-apus-reviewed/4

TH review on the i7 3770K with similar settings

http://www.tomshardware.com/reviews/ivy-bridge-benchmark-core-i7-3770k,3181-5.html

I account the i7's loss in FPS due to FRAPs so I would not say that the i3 is faster than the i7 at best the same.

Interestingly here we know that Skryim prefers a faster CPU and at 1280x720 the load on the iGPU is at its lowest so in theory this is best case scenario rather than median of performance of the iGPU itself, as can be seen the A8 3850 is nigh on 20FPS faster and the A10 5800K is as high as 30+ FPS which is severe, to stress that maybe HD4600 is not all it was said to be.

http://www.tomshardware.com/reviews/core-i7-4770k-haswell-performance,3461-7.html

1366x768 not a dramatic bump in resolution so it may only acount for 2-3 FPS so even a contemporary fastest processor with the best iGPU the company has made still is basically on par with a A8 3850, this is best case scenario, once the more demanding benches come out the FPS will tail off.

http://www.tomshardware.com/news/intel-Haswell-Flagship-TakeItNow-MaxICT,21594.html

This is a potentially scary article which lends to the sentiments that Intel will be across the board raising SKU pricing by what looks to be $30-50, if a i3 4XXX is priced around the $170 it will make it completely uninteresting knowing that APU Richland A10 6800k will be priced at $130 replacing the 5800k which will drop to around $115-120, if its iGPU is 15-20% faster then its basically made up the consumers mind.

Long story short JDJ, the HD7660D can play F1 2012 at ultra settings at 1080 around the 39-42 FPS mark

 

[jdwii]

I can already read the Haswell reviews :

"One year after, and the CPU perf is incremental at best. The iGPU, much improved, still has lots of room to grow. It still cant compete with Trinity APU's from AMD. If you have to build a new system, and are going to add adGPU, we wholeheartedly recommend the 4770K. If you want to build a "good enough" PC, we think that a Richland APU system would be better. "

You are giving reviewers too much unbias. They are going to read. OMG AMAZING INTEL CPU, LOOK AT THAT GPU PERFORMANCE COMPARED TO BEFORE AND IF YOU WANT TO DO WORK, CPU PERFORMANCE IS SO MUCH HIGHER THAN AMD ON ITUNES.

And Richland review will be: 6/10 "its ok" - Anandtech.

That's pretty much how every Amd review on their site is, they are incapable of saying a A10 is a better choice overall instead of just gaming then an I3 in a laptop or even a desktop. They always put dumb comparisons in their benches like a I7 to a A10 when its not even in the same playing field(or price with the A10 4600M they put 1 I3 in their tests no pentiums but several I7's i mean really).

 

[truegenius]

Again clock rates between different uArch's are nearly meaningless.

yes i know that we can't say that 4ghz bd and 4ghz sandy are equal because they have same clock rate
every average joe kow this

As for the confidential information, that went way over your head.

The internal micro-instructions and how their processed is confidential. The 8086 didn't use micro-instructions, none of the x86 CPU's did until the Pentium and none did ILP until the Pentium Pro.

[strike]it is important to keep them secret otherwise amd can learn from intel's cpu and thus less R&D will required just copy paste and viola a new cpu (like arm arch vs qualcomm (eg cortex-a15 vs krait , krait is identicle to a15)[/strike]

btw, don't ya think that my naming scheme (ppc, ppcpg, ppcpw .......etc) are good

You are giving reviewers too much unbias. They are going to read. OMG AMAZING INTEL CPU, LOOK AT THAT GPU PERFORMANCE COMPARED TO BEFORE AND IF YOU WANT TO DO WORK, CPU PERFORMANCE IS SO MUCH HIGHER THAN AMD ON ITUNES.

And Richland review will be: 6/10 "its ok. Needs 2 generations to get it right. " - Anandtech.

:lol:

 

[sarinaide]

I find for reviews they are trying to place Intel's top end iGPU in perspective with the top end and most expensive AMD iGPU which is cheaper than all of intel's i3, i5 and i7 parts, so when they take the i7 3770 and 4770 and 3225 it is to ascertain strength only. On that basis the AMD wins all the time and quite substantially as well. Intels highest end part for DT namely GT is basically only on par with a A8 3850 best case still slower, probably around a A6 or A8 lower end part in true performance.


This is why the A10 5800k is priced to compete against the i3 3220 at the same bracket while the 3225 is $150 and there are better CPU options at that price tag. So if its i3 3220 vs A10 5800K, or HD2500 vs HD7660D its a complete miss match so they resort to pandering around where the APU proves to be a better value orientated buy with better chipsets and features and expansion capacity by simply focusing only on where the i3 can win and that is single threaded x86 operations and IMC.

So in passing the APU again I stress this represents the balance of price to performance, as a simply HTPC and SFF setup the A Series platform is copious with features and connectivity including up to 7 SATA 6GB/s Ports 6 USB 3.0, AMD Quick Stream technology, Lucid MVP, AMD Power Management Technology, The best iGPU solutioun rivaled by none but its own successors and more than enough x86 performance, Direct Compute and OpenCL support, limited HSA support which will only be relevent on Kaveri parts. If you need more graphics horsepower you have the Dual Graphics option and using a HD7790/7850 or 660 gives you high settings on Crysis 3 all while staying in the lower price tag.

 

[JAYDEEJOHN]

I liked their review of an AMD cpu, where on the first page they had a picture of a Intel cpu
Only on Anandtech

 

[sarinaide]

I liked their review of an AMD cpu, where on the first page they had a picture of a Intel cpu
Only on Anandtech

Now this I have to see

 

[JAYDEEJOHN]

Oh, he switched it, and had 2 articles, the "update" didnt have it, and excuses were made.
Depending on how you look at it, it was somewhat valid, with/including, a sideways admitted mistake
Even then, half hearted, and made you wonder just what was he thinking?

PS The second article came out within an hour or so of the first

 

[Cazalan]

We live in a world where judges take bribes, congressmen take bribes, police take bribes, presidents commit treason. Think online journalists are somehow kept to a higher standard for opinion pieces? LoL!

 

[de5_Roy]

Kabini successor has Jaguar+ cores
http://www.fudzilla.com/home/item/30959-kabini-successor-has-jaguar%20-cores

A look at Intel’s Latest PC Game Developer Relations Strategy
http://semiaccurate.com/2013/04/03/a-look-at-intels-latest-pc-game-developer-relations-strategy/
more wins(game devs) coming intel's way. looks like amd will have to make do with the console wins....

heh heh
http://www.tomshardware.com/news/APU-Bitter-Neal-Robison-Tony-Tamasi-PlayStation,21791.html

 

[8350rocks]

Kabini successor has Jaguar+ cores
http://www.fudzilla.com/home/item/30959-kabini-successor-has-jaguar%20-cores

A look at Intel’s Latest PC Game Developer Relations Strategy
http://semiaccurate.com/2013/04/03/a-look-at-intels-latest-pc-game-developer-relations-strategy/
more wins(game devs) coming intel's way. looks like amd will have to make do with the console wins....

Well, considering Far Cry 3, Bioshock Infinite and Tomb raider were all developed with AMD developer tools, I am pretty sure they will be just fine...