AMD CPU speculation... and expert conjecture

[sarinaide]

So I guess after much guessing the "Super FX" aka Centurion was actually true. From what I am told about this chip is it is likely featuring AMD test run of RCM, stock frequency is 5ghz @ 1.357v boost at 5.5ghz, rumor is overclocking bin on these is up on 6ghz using conventional cooling and rumor of an insane LN2 world record to be had here.

These are formed on golden sample silicon so this is like tray samples of old, exceptional overclocking and thermal performance. At 5ghz out the box the performance will be exceptional, what is not exceptional is the $795 price tag and $1200 for a tray chip beforehand. If its fast it will be a collectors item and insanely fat CPU.

This reminds me of the leaky phenoms that were sold primarily for ln2 testing purposes.

http://hothardware.com/Reviews/AMD-Phenom-II-Black-Edition-TWKR/

I have a Phenom II BE 42 TWKR, those were golden chips for enthusiasts, but yeah it may very well be a true overclockers part.

Are you sure? I believe all the Vishera, Trinity and Richland chips have them....and that was the reason given behind the significant reduction in power usage from BD->PD.

If RCM is present in Piledriver then its very passive, the major reason for the efficiency improvement is better process.

 

[anxiousinfusion]

Looks like Kaveri, which has two DDR5 units.

Somehow I don't think Kaveri would show up at an Intel Developer Forum in Beijing.

I didn't say that was Kaveri, I said it looks like Kaveri just with one eDRAM bank instead of two. The Kaveri leak happened a while back, and thats GDDR5 though.

I heard April 19 for Richland, May for Tamesh and Kabini

Yes, April 18th-19th last I heard. Richland may be released five days from now according to WCCFTech: http://wccftech.com/amd-a-series-apus-simcity-bundle-promotion-compares-richland-apu-performance-haswell-igpu/

The promotion would be expanded to Richland A-Series APUs (A10/ A8) when they launch on 18th-19th April 2013.

 

[esrever]

So I guess after much guessing the "Super FX" aka Centurion was actually true. From what I am told about this chip is it is likely featuring AMD test run of RCM, stock frequency is 5ghz @ 1.357v boost at 5.5ghz, rumor is overclocking bin on these is up on 6ghz using conventional cooling and rumor of an insane LN2 world record to be had here.

These are formed on golden sample silicon so this is like tray samples of old, exceptional overclocking and thermal performance. At 5ghz out the box the performance will be exceptional, what is not exceptional is the $795 price tag and $1200 for a tray chip beforehand. If its fast it will be a collectors item and insanely fat CPU.

It is still complete fud at this point. I don't see a reason anyone would buy that for that price. You can get a 6 core ivy extreme for that price. There is no way a 8 core piledriver could remotely beat a i7x even if its clocked at 5ghz. The other thing about it is the fact it would be highly unlikely to be manufacturable. The power consumption will be extreme even with binning.

Im calling bullshit on this one. Un-named sources to 1 website like hexus are ussually wrong.

 

[griptwister]

Hmmm... who predicted this 5Ghz barrier would be broken within a year or two? lol

Anyways, I'm curios to see how a 5Ghz chip compares to a 4 Ghz chip when people say they don't notice a performance improvement past 4.5Ghz on the 8350 chip. Also, how will this chip compete against Intel? Especially at that price point. I just don't see this chip beating a hexicore i7, even at 5 ghz. Unless they somehow changed the singlethread performance...

 

[noob2222]

That would make sense, but RCM is not present in any of AMD's current processors, AFAIK.

I suspect that RCM is in trinity. The problem with RCM is even slight imperfections in the silicon will cause RCM efficiency to not be detected. However when the silicon is in perfect harmony ...

the problem is all the conflicting reviews on wether or not they claim RCM is in the chip. This is nothing like the phenom II "HE" binning where the clock speed was reduced by 35% in order to achieve the maximum efficiency peak.

I don't see achieving this efficiency curve from 3.8ghx to 3.4 ghz without RCM, but only AMD really knows for sure.

 

[JAYDEEJOHN]

An i7 wouldnt stand a chance against an 8 core at 6GHZ.
They simply wont go that fast on air.
Perf wise, depending on clocks and scaling, the i7 may win, but then again, that wouldnt be to whom, or why these chips would be sold

 

[8350rocks]

So I guess after much guessing the "Super FX" aka Centurion was actually true. From what I am told about this chip is it is likely featuring AMD test run of RCM, stock frequency is 5ghz @ 1.357v boost at 5.5ghz, rumor is overclocking bin on these is up on 6ghz using conventional cooling and rumor of an insane LN2 world record to be had here.

These are formed on golden sample silicon so this is like tray samples of old, exceptional overclocking and thermal performance. At 5ghz out the box the performance will be exceptional, what is not exceptional is the $795 price tag and $1200 for a tray chip beforehand. If its fast it will be a collectors item and insanely fat CPU.

It is still complete fud at this point. I don't see a reason anyone would buy that for that price. You can get a 6 core ivy extreme for that price. There is no way a 8 core piledriver could remotely beat a i7x even if its clocked at 5ghz. The other thing about it is the fact it would be highly unlikely to be manufacturable. The power consumption will be extreme even with binning.

Im calling bullshit on this one. Un-named sources to 1 website like hexus are ussually wrong.

Well, Tom's Hardware reported it, so did techreport, wccftech, anandtech, and a few others...

Look here and tell me what you think about it's validity?

EDIT: Anyone else see that the newest version of catalyst drivers (13.1) supports A10-5800k X HD 7750 crossfire now? I saw a review about it somewhere, and there have been several people posting that it's compatible now...

6 core Kaveri + HD 7790 crossfire maybe?

 

[truegenius]

imo fx8350 needs ~6GHz to match i7-3970 on stock in multithreaded tasks
that is why it is priced below i7-3960/70

 

[sarinaide]

So I guess after much guessing the "Super FX" aka Centurion was actually true. From what I am told about this chip is it is likely featuring AMD test run of RCM, stock frequency is 5ghz @ 1.357v boost at 5.5ghz, rumor is overclocking bin on these is up on 6ghz using conventional cooling and rumor of an insane LN2 world record to be had here.

These are formed on golden sample silicon so this is like tray samples of old, exceptional overclocking and thermal performance. At 5ghz out the box the performance will be exceptional, what is not exceptional is the $795 price tag and $1200 for a tray chip beforehand. If its fast it will be a collectors item and insanely fat CPU.

It is still complete fud at this point. I don't see a reason anyone would buy that for that price. You can get a 6 core ivy extreme for that price. There is no way a 8 core piledriver could remotely beat a i7x even if its clocked at 5ghz. The other thing about it is the fact it would be highly unlikely to be manufacturable. The power consumption will be extreme even with binning.

Im calling bullshit on this one. Un-named sources to 1 website like hexus are ussually wrong.

Well, Tom's Hardware reported it, so did techreport, wccftech, anandtech, and a few others...

Look here and tell me what you think about it's validity?

EDIT: Anyone else see that the newest version of catalyst drivers (13.1) supports A10-5800k X HD 7750 crossfire now? I saw a review about it somewhere, and there have been several people posting that it's compatible now...

6 core Kaveri + HD 7790 crossfire maybe?

I am interested in the 13.1 supports 7750 crossfire part. Do you mean Dual Graphics? if so if you can find that linky that will be a great read.

 

[sarinaide]

That would make sense, but RCM is not present in any of AMD's current processors, AFAIK.

I suspect that RCM is in trinity. The problem with RCM is even slight imperfections in the silicon will cause RCM efficiency to not be detected. However when the silicon is in perfect harmony ...

the problem is all the conflicting reviews on wether or not they claim RCM is in the chip. This is nothing like the phenom II "HE" binning where the clock speed was reduced by 35% in order to achieve the maximum efficiency peak.

I don't see achieving this efficiency curve from 3.8ghx to 3.4 ghz without RCM, but only AMD really knows for sure.

The other factor here is that the APU, namely the 5800K, I can only achieve 4.5ghz stable at 1.4v, that being the lowest stable voltage for that clock, for complete stability you need 1.425v~. I think for the APU to run the CPU and GPU components at 4ghz and 900mhz it point to RCM but again its really passive.

I think Richland along with its AMD power management bake ins might show a better RCM hence why they are saying 1.325v runs the 6800K's, Boost frequency of 4.4ghz.

 

[$hawn]

Hmmm... who predicted this 5Ghz barrier would be broken within a year or two? lol

Anyways, I'm curios to see how a 5Ghz chip compares to a 4 Ghz chip when people say they don't notice a performance improvement past 4.5Ghz on the 8350 chip. Also, how will this chip compete against Intel? Especially at that price point. I just don't see this chip beating a hexicore i7, even at 5 ghz. Unless they somehow changed the singlethread performance...

Actually it would be able to catch up to a 3.3GHz 6 core SB-E chip, but IB-E should take a slight lead ahead.

If your not amused by theoretical guestimation, pls do NOT read past this point

Each PD core has only 2/3rds [66%] of the IPC that an SB core has in most applications on average.
A 5GHz PD core * 0.66 = 3.3GHz SB core.

If each PD core is taken as 100 speed units, then assuming ideal scaling, total compute power = 8x100 = 800units.

Now, for the SB-E chip, compute power without HT = 6*100 = 600 units.
Assuming HT brings in a 30% boost on average, 600*1.3 = 780 units

But IB-E should race ahead though

 

[JAYDEEJOHN]

So, SB is 50% faster eh?

 

[mayankleoboy1]

Those 20 point difference is easily covered by Intels Turbo Boost

EDIT: IIRC, i have heard from some AMD fanbois that AMD is much better than Intel at "switching between large number of processes" , which means that if you have many processes, some CPU heavy and some not, then switchinf between them is a much better experience on AMD processors.

 

[A Bad Day]

I wonder if the mobile Kaveri APU will come with DDR3-2133 support? Will they also start including more than two modules for the Kaveri APU?

 

[anxiousinfusion]

I wonder if the mobile Kaveri APU will come with DDR3-2133 support? Will they also start including more than two modules for the Kaveri APU?

There was talk of up to three compute units in Kaveri, which with AMD's module design would translate to six cores.

 

[8350rocks]

Review:

http://www.firstpost.com/topic/organization/amd-sapphire-hd-7770-hd-7750-a10-5800k-crossfire-benchmark-video-IfPVxUSkURw-48380-1.html

3dmark11 scores with/without dual graphics with/without HD 7750:

http://www.techpowerup.com/forums/showthread.php?t=180080

 

[sarinaide]

Interesting that, will give it a look seee

 

[$hawn]

So, SB is 50% faster eh?

Might be even faster sometimes actually, depending on the application.

 

[JAYDEEJOHN]

Just curious, links to real world apps?

 

[$hawn]

Just curious, links to real world apps?

All these numbers I've put up are a result of reading numerous comparison articles, some common sense and a little guess work. Do NOT treat this as scientific data!!

If you want to know how i arrived at the 50% faster value, have a look at this AAC encoding chart (audio encoding is my favourite single core performance measuring yardstick).
http://media.bestofmicro.com/Y/4/357628/original/itunes.png

SB takes 62 seconds @3.8GHz turbo.
As PD takes 84 secs @4.2GHz turbo, so it should logically take 93 seconds @3.8GHz

IPC ratio of PD/SB in AAC encoding = 62/93= 0.66. In other words, SB is 1/0.66 => 1.51x PD IPC, ie, its 51% faster.

I repeat again, all this is just my fancy guess work, and for all you know I could be very wrong. Do correct me if I am. My only intention was to put into perspective what a 5GHz centurion chip would look like against a SB-E chip. Cheers

 

[esrever]

So I guess after much guessing the "Super FX" aka Centurion was actually true. From what I am told about this chip is it is likely featuring AMD test run of RCM, stock frequency is 5ghz @ 1.357v boost at 5.5ghz, rumor is overclocking bin on these is up on 6ghz using conventional cooling and rumor of an insane LN2 world record to be had here.

These are formed on golden sample silicon so this is like tray samples of old, exceptional overclocking and thermal performance. At 5ghz out the box the performance will be exceptional, what is not exceptional is the $795 price tag and $1200 for a tray chip beforehand. If its fast it will be a collectors item and insanely fat CPU.

It is still complete fud at this point. I don't see a reason anyone would buy that for that price. You can get a 6 core ivy extreme for that price. There is no way a 8 core piledriver could remotely beat a i7x even if its clocked at 5ghz. The other thing about it is the fact it would be highly unlikely to be manufacturable. The power consumption will be extreme even with binning.

Im calling bullshit on this one. Un-named sources to 1 website like hexus are ussually wrong.

Well, Tom's Hardware reported it, so did techreport, wccftech, anandtech, and a few others...

Look here and tell me what you think about it's validity?

EDIT: Anyone else see that the newest version of catalyst drivers (13.1) supports A10-5800k X HD 7750 crossfire now? I saw a review about it somewhere, and there have been several people posting that it's compatible now...

6 core Kaveri + HD 7790 crossfire maybe?

a bunch of sites reporting rumors don't make it true. They don't have a real source. So yea. Its retarded.

 

[truegenius]

I repeat again, all this is just my fancy guess work,

i see that you missed the part where we need to multiply by ipc factor (i.e, multiply fx core by 0.66 i.e, 800x0.66) and where we neet to consider clock diffence too

here i offer my theory which resembles to yours
http://www.tomshardware.com/charts/cpu-charts-2012/-01-Cinebench-11.5,3142.html
here we can se that
SB @ 3.9ghz scores 1.57 i.e, 1.57*5/3.9 = 2.38 @ 5GHz
(formula is points * target speed / speed at which initial point is calculated)
PD @ 4.2ghz scores 1.11 i.e, 1.11*5/4.2 = 1.322 @ 5GHz

so SB/PD = 2.38/1.322 = 1.8
means at same speed, SB is 80% faster than PD in single threaded

combine it with cores+ht
then SB-E @5 ghz = 2.38 * 6 * 1.2 = 17.136
(1.2 is for 20% performance from hyper threading because i see max only 10-20% improvements via ht (unless it is not atom))
fx8350 @ 5ghz =1.322 * 4* 1.6 = 8.461
(4 for module count and 1.6 for performance of a full module mens module loses 20% performance on each core)
SB-E @3.3ghz = 17.136 * 3.3 /5 = 11.3

so Sb-E is still 11.3/8.461 = 1.34 i.e, 34% faster than fx8350
but 4 core SB will be 11.3 *4 / 6 = 7.53 @3.3Ghz
so fx8350@5Ghz is better than any 4 core intel cpu but will be 34% slower than sb-e

i am using cinebench r11. 5 because i think it is good indicator of performance for general

 

[JAYDEEJOHN]

What I find interesting is, its much like C2D vs the old K8s, as tho differences were much the same, which means, even going backwards with BD, AMD has kept pace with Intel, the difference being theres more growth finally for AMD, which will start to narrow the gap, abd since this is rumors/conjectures, there you have it

 

[8350rocks]

Cinebench uses ICC compiler...which does not accurately reflect real world values for AMD because of the ICC genuine intel flag.

 

[truegenius]

taking x2 555 as base
http://www.anandtech.com/bench/Product/120?vs=404
(4 818/3 724)*(3.2/2.9)= 1.43 (formula is => (score of 555/score of 850)*(speed of 555/speed of 850)

cinebench showed ~40-45% more per core per ghz performance in sb in comparision to k10
so k10 should be clocked at ~40-45% higher speed than sb
means 2.2ghz sb = 3.2ghz k10
so here it is
http://www.anandtech.com/bench/Product/120?vs=407
showing predictble results in many tasks
but i am saying that cinebench is good but not elite