AMD chief say BD will offer only 35% not 50% more performance than previous gen

[grkM3]

NO no no . If they do it as you say I call cheat. Intel IB will run on 1866 memory . Downclock IB leave AMD at stock.

well we dont know what timings were used and from what I read online is that intel will officially use 2133 with ivy,thats not a fair way to compare it if BD is on 1866 and its not fair to down clock ivy if its selling with support for 2133 ram.

 

[Idontcare]

So basically you're saying BD is a 35% improve over PhII and Sandy Bridge is 50% better than PhII.

PhII = 1
Sandy Bridge = 1.5
BD = 1.35
Sandy Bridge * 4 cores = 6
BD * 8 cores = 10.8
Sandy Bridge * 8 threads = ??? (2600K)

Where does the 2600K fall in your performance scaling list? Or SB-E? Or IB?

Now take the BD's supposed 5 Ghz capabilities and you got served.

/end_of_third_eye

Why limit yourself to waxing poetically about supposed capabilities? Just go for broke, bulldozer goes to 11:

Now take the BD's supposed 11 Ghz capabilities and you got served.


/end_of_third_eye

 

[Nemesis 1]

AMD is not gonna make a Performance market CPU the same speed as as an entry-level CPU and Llano, which already stepped in to completely replace (in the near future, that is) the Athlon II X4 and Phenom II X4 in the mainstream market, is MUCH faster than an Atom. It goes against the Core i3 quite nicely, especially in multi-threaded programs, so it's impossible for them to make a slower CPU than that with a higher price.

To me it seems like you're just playing devil's advocate for the sake of it while using your personal opinion as arguments. I suggest we stick to the facts; this is a technical discussion.

Iam not sure . Its rather confusing your argument is this . That BD is performance class cpu . Correct! So in all fairness we can't even compare SB-E 4 core to BD because it cost more. Correct! You almost make it sound like this is the way AMD intended it to be .

Let me help you and free you from your disenchant. AMD had every intention of BD being an enthusiast cpu . None here would say differantly . None who are sane anyway.

AMD missed the moving target. So they are back to being a cheap CPU . If AMD had created another performance monster . I would love to here what you would say under those circumstances.

The fact is AMD blew it yet again . and trying to pass it off as a 8 thread beast is pure hype on the desktop at this time in time. Your really need to change jobs , Because you have talent. You could easily be successful in marketing . IF you had a product like Intel to sell the sky would be your limit . So give that some thought , You really did well in this thread all things considered. Give my suggestion some thought . YOU would be good at marketing . Your better than 90% of the marketing people I know.

 

[Nemesis 1]

well we dont know what timings were used and from what I read online is that intel will officially use 2133 with ivy,thats not a fair way to compare it if BD is on 1866 and its not fair to down clock ivy if its selling with support for 2133 ram.

Thats all right don't worry about it . IB at 1866 will do just fine . Infact I doubt you would see .5 differance in performance on IB at those memory speeds , So really the Fair thing would be to run AMD at stock against IB at 1866. Beings how AMD going to get creamed by IB anyway.

 

[grkM3]

Thats all right don't worry about it . IB at 1866 will do just fine . Infact I doubt you would see .5 differance in performance on IB at those memory speeds , So really the Fair thing would be to run AMD at stock against IB at 1866. Beings how AMD going to get creamed by IB anyway.

Yup but I do think it will make a difference once they go 6 core,they will be able to make use of that extra bandwith.

Sandy E on the other hand is going to freaking kill everything with quadchannel 2133 lol

I hope they put out a 10 core 20 thread chip on that platfrom

 

[AtenRa]

"We don't have clue of what the actual IPC would be"

Correct.

"but it's safe to say that it should at least be improved somewhat over K10.5."

Incorrect.

There is no valid substantiating data to uphold your second statement. For all anyone outside of AMD knows, BD might not be outperforming Atom right now, or it could be outperforming IBM Watson with 7 modules blocked off. Although your deduction process is essentially valid given the extremely limited data it is based on... benchys talk and they are often full of surprises. Oh... Did I mention benchys talk?

We may not have actual bench data or real world data from BD but we DO KNOW that single core performance will be higher than Phenom II. (How much higher remains to been seen)

I will direct you to read (or re-read) Anand’s BD article


AMD Discloses Bobcat & Bulldozer Architectures at Hot Chips 2010

Technical data No1 : Better Front End

The decoder is now 4-wide an increase from the 3-wide front end that AMD has had since the K7 all the way up to Phenom II. AMD can now fuse x86 branch instructions, similar to Intel’s macro-ops fusion to increase the effective width of the machine as well. At a high level, AMD’s front end has finally caught up to Intel, but here’s where AMD moves into the passing lane.

Technical data No1a : Better Front End (Branch Prediction and a Deeper Pipeline)

In Bulldozer the branch prediction and fetch logic are decoupled. The predictor now produces a queue of future fetch addresses. Even if there’s a mispredict the branch predictor can continue to fill its prediction queue with targets. The fetch logic can then check this queue of addresses against what’s in the instruction cache to avoid future misses in L1.

Technical data No1b : Better Front End (Prefetchers)

With Phenom AMD implemented comparable prefetching logic to what Intel did with Core. In Bulldozer, AMD is ramping up the aggressiveness of those prefetchers. There are independent prefetchers at both the L1 and L2 levels that support larger numbers of strides and large stride sizes (both compared to what exists in current AMD architectures). There’s also a non-strided data prefetcher that looks at correlated cache misses and uses that data to prefetch into the caches.

Technical data No2 : Better Integer Execution Units

Each scheduler has four ports that feed a pair of ALUs and a pair of AGUs. This is down one ALU/AGU from Phenom II (it had 3 ALUs and 3 AGUs respectively and could do any mix of 3). AMD insists that the 3rd address generation unit wasn’t necessary in Phenom II and was only kept around for symmetry with the ALUs and to avoid redesigning that part of the chip - the integer execution core is something AMD has kept around since the K8. The 3rd ALU does have some performance benefits, and AMD canned it to reduce die size, but AMD mentioned that the 4-wide front end, fusion and other enhancements more than make up for this reduction. In other words, while there’s fewer single thread integer execution resources in Bulldozer than Phenom II, single threaded integer performance should still be higher.

After all that data, I dont understand why people think BD will have the same IPC with K10.5.

I dont want to make a performance guess, but with all the changes in BD Micro-Architecture a 15-25% more IPC is not impossible.

 

[Abwx]

AMD is using 6 cores to compete with a 4 core processor, and that's before considering their overclocked performance where 2500k will pull away even farther while consuming 87% less power.

.

Let s use this Xbit labs review as a bench to check your (in)competence.

Iddle :
PII X6 4G : 75.9W
2500K 4G7: 40.6W
That s 46% less , if it was 87% , the 2500K would use 9.9W.....

100% LOAD :
PII X6 4G : 305W
2500K 4G7 : 163W
That s 47% less , if it was 87% , the 2500K would drain 40W....

That s in line with the rest of your estimations about
the performances and features differences...

The only people who are incompetent here are the ones engaging in personal attacks. This is not OT guys, you MUST act better than this
-ViRGE

 

[AtenRa]

Let s use this Xbit labs review as a bench to check your (in)competence.

Iddle :
PII X6 4G : 75.9W
2500K 4G7: 40.6W
That s 46% less , if it was 87% , the 2500K would use 9.9W.....

100% LOAD :
PII X6 4G : 305W
2500K 4G7 : 163W
That s 47% less , if it was 87% , the 2500K would drain 40W....

That s in line with the rest of your estimations about
the performances and features differences...

ehm, no

40,6W + 87% = ~75,9W

and 163W + 87% = ~305W

PHII X6 draws 87% more power than Core i7 2500K


EDIT: Yes you are right, if you say that Core i5 2500K draws 87% less power, as you have said it is not correct

 

[Abwx]

After all that data, I dont understand why people think BD will have the same IPC with K10.5.

I dont want to make a performance guess, but with all the changes in BD Micro-Architecture a 15-25% more IPC is not impossible.

It has been explained ad nauseam in all BD related threads, yet there
is people still insisting that AMD prefered to release a new uarch that
do not as well as would a shrink of 8 K10.5 cores, despite having a die
area that is 30% bigger than the theorical 8C K10.5 easy route....

 

[Abwx]

ehm, no

40,6W + 87% = ~75,9W

and 163W + 87% = ~305W

PHII X6 draws 87% more power than Core i7 2500K

You are right and he is STILL wrong...
He said that 2500K draw 87% LESS power, wich is false.

 

[AtenRa]

yes, i have edited my post above

 

[Nemesis 1]

I am embarresed for both you guys , Shameful display of math skills,

I not going to explain it to you .My 4 year old grandson has better math skills. here is the proper usage to get correct ans ,

75.9 is what % of 40.6 = 186.9 which is 86.9 after you remove intels 100% = 87%

305 is what % of 163 = 187 again remove intels 100% = 87% . This was a shameful attempt at math. I will leave it to someone else to explain why its done in this manner. I won't because I find it extremely hard to believe you couldn't just look at the numbers and see how wrong ya are its so painfully clear to see, I not sure you 2 guys didn't do this by accident . But he was right.

Second edit attempt . I see one got it right . The other got it wrong even after he was corrected. I left my orginal post as it was, Just added the correction after rereading as he is still wrong part . Threw me . Sorry for the comprehension slip.

 

[Nemesis 1]

You are right and he is STILL wrong...
He said that 2500K draw 87% LESS power, wich is false.

Please explain how he is still wrong. SB draws 87% less power by the numbers given .

 

[Abwx]

Please explain how he is still wrong. SB draws 87% less power by the numbers given .

PII X6 consume 87% MORE than a 2500K wich consume 47%
LESS than the said PII X6....

So you re wrong as well, it seems...

 

[Nemesis 1]

It has been explained ad nauseam in all BD related threads, yet there
is people still insisting that AMD prefered to release a new uarch that
do not as well as would a shrink of 8 K10.5 cores, despite having a die
area that is 30% bigger than the theorical 8C K10.5 easy route....

You can explain it till your blue in the face. Your basing your assumption that AMD engineers are = to intels . Your assuming AMD because of so called improved front end got a performance gain . Your assuming that sharred resources won't degrade performance . Your assuming that long pipes = higher clocks . Intel thought they could get to 10ghz on P4 we all know how that played out . Your also assuming that SOI @32 will be as good as SOI@ 45nm. Your also assuming that gate first is as good as gate last. Annealing can cause all sorts of problems with metal gates.

Didn't AMD buy a company called Next ?

 

[AtenRa]

Your assuming that sharred resources won't degrade performance .

Irrelevant, we are talking about SINGLE core/thread performance.

 

[Nemesis 1]

PII X6 consume 87% MORE than a 2500K wich consume 47%
LESS than the said PII X6....

So you re wrong as well, it seems...

Show the math> my god man.

Lok at the numbers.

305watts AMD 163 Intel .

I will make it easier for ya with wrong numbers but so what .

300 watts AMD 150 watts intel . What % is this

300watts is what percent of 150 watts. = 200% Subtract intels 100% = 100% more watts .

2x 150 = 300. or we can go 300-150 = 150 150 is 100% 0f 150

 

[Nemesis 1]

Irrelevant, we are talking about SINGLE core/thread performance.

OK thats fine by me . Are both cores in an amd module = ?????

 

[AtenRa]

Nemesis, take off the blue class and you will see the correct math.

2500K has 46% LESS power consumption than PHII X6

PHII X6 has 87% MORE power consumption than 2500K

Like it or not, that's the correct expression

 

[AtenRa]

OK thats fine by me . Are both cores in an amd module = ?????

what are you trying to say ??

 

[Nemesis 1]

PII X6 consume 87% MORE than a 2500K wich consume 47%
LESS than the said PII X6....

You can't do it like this . Whats is the lowest power usage . Thats the base it has to be done like this , 305 is what% of 163.

It can't be written or figured like this .

163 is what % of 305 .. This math is wrong period . You can spin it any way ya want .


LETs do this. What percentage of core increase does AMD 16 core processorBD have over 12 core MC . 33% 0r 25% You can get both results but only 1 result is correct. I choose 25% only cut AMD slack . Make it looka little better for them . But 33% is the only correct ans.

Because MC was released first it has to be figured as 12core base . You cannot use 16 as the base . Ever unless its worded differantly .

16 is what % of 12 1.3333333= 33% It has to be worded like this.

Now if AMD went from 16 cores down to 12 cores than you do this

12 is what percent 16 =75% = 25% fewer cores as the base changes . You can play on words all day long but your wrong and if you insist your not than maybe a mod should PM ya.

 

[Nemesis 1]

what are you trying to say ??

ANS the dam question

You said this .( Irrelevant, we are talking about SINGLE core/thread performance.) Now ans the question.

 

[Troll Trolling]

Your username fits the argument that Atom can be faster than Bulldozer well.

Also, what do you mean "it's not a matter of making sense"? Everything in an argument is a matter of making sense, especially here since we're talking an entry-level CPU vs a performance CPU.

First thing, I don't really believe Bulldozer is going to be slower than Atom.

The thing is, there is no data. It could be slower than Atom, as it could be 10x faster than Sandy Bridge. This is my point, as is Bridito's point.

 

[TemjinGold]

You can't do it like this . Whats is the lowest power usage . Thats the base it has to be done like this , 305 is what% of 163.

.

That's where you went wrong. The base in a math equation isn't always the lowest number, it's the point of reference that you start at. If A uses 87% more than B, B does NOT use 87% less than A because in the former, A is the base, while in the latter, B is the base. Let me explain it to you with a simple example:

If CPU A consumes 50 watts and CPU B consumes 150 watts, you can say that CPU B consumes 200% MORE power than CPU A, which is NOT the same as saying CPU A consumes 200% LESS power than CPU B. I hope you realize that it is IMPOSSIBLE to consume 200% LESS power.

In this example, you can say CPU B consumes 200% MORE power than CPU A or you can say CPU A consumes 66.7% LESS power than CPU B.

 

[Nemesis 1]

I see you refused to ans the question . Puts taill between legs and runs as he new he was going to get hammered. You relized AMD BD has shared resources and if you can only use one core kind screws the performance of shared resources now doesn't it.

Each core has 1 128bit FPU. kind bites doesn't it. So your sharred resources just won't go away and you can not compare single core results as performance would take a big hit . I seen ya used the thread word . But you used the single core only. Ya see lmpw matter what. You have to compare 1 module 2cores to intels 1 cores. You can't compare a single core SB to single core BD as the performance takes pretty big hit ; So no matter what We will allways comparred 2 intel cores to 1 AMD module . But we can do single thread apps. But single BD core against SB core will never work as 2 cores are required in resource sharring .