Some news on Piledriver

[Martini1991]

Its already 25% faster than Bulldozer, look, read and think about it

/facepalm.

Any fool can make an article with some written figures, need actual results and lots of them.
Also, AMD has it pegged at 10-15% themselves, while yes they can get better than that, I'm going to take everything with a pinch of salt, especially given AMD's track record.
Too many people hyping everything up, winds me up, get stupidly over optimistic.

 

[humbug]

/facepalm.

Any fool can make an article with some written figures, need actual results and lots of them.
Also, AMD has it pegged at 10-15% themselves, while yes they can get better than that, I'm going to take everything with a pinch of salt, especially given AMD's track record.
Too many people hyping everything up, winds me up, get stupidly over optimistic.

Well yes, i don't disagree on AMD's record with Bulldozer, but if you go back to the Toms Hardware link they clearly found AMD's promised 15% IPC, its also clocked at least 10% higher, add the numbers up.....

And its no good saying their numbers are wrong, but would be right for Intel, no one is that cynical.

 

[Martini1991]

Well yes, i don't disagree on AMD's record with Bulldozer, but if you go back to the Toms Hardware link they clearly found AMD's promised 15% IPC, its also clocked at least 10% higher, add the numbers up.....

And its no good saying their numbers are wrong, but would be right for Intel, no one is that cynical.

AMD's slides say 10-15%, nothing about clock speeds or IPC separately.

You're overhyping PD.

Trinity also IIRC is due to hit shops Q3, can't see Vishera (Spelling) doing the same personally.

Who said anything about Intel numbers?

EDIT : Same ****, different AMD CPU as far as it seems, people banged on about flipping agenda, then deneb, then Zambezi.
Not bad CPU's in their own right, but not upto the same standard that is available.

 

[humbug]

AMD's slides say 10-15%, nothing about clock speeds or IPC separately.

You're overhyping PD.

Trinity also IIRC is due to hit shops Q3, can't see Vishera (Spelling) doing the same personally.

Who said anything about Intel numbers?

AMD's slides say "Performance per watt improvements and 10 - 15% IPC improvments" that is exactly what it is

 

[Martini1991]

Lies ;

Don't even know why I bother.
That 10-15% is likely to be a mixture of clock and IPC, doesn't state any specific clock or IPC improvements, other than "Power and IPC improvements".
I wouldn't be surprised to see Vishera early Q4.

 

[humbug]

Good Grief..... Power improvements = just that, they are able to get higher clocks for the same power usage.

Look at this, FX-8150 3.6Ghz 8 cores TDP 125w

Trinity 3.9Ghz 4 cores + IGPU having approximately 50% of a 100w TDP + 15% IPC improvement over FX-8150.

Take off the IGPU, add 4 cores @ 3.9Ghz = approximately 100w TDP, add remaining 25w TDP = excess.

FX-8350 + 10% Clock + 15% IPC = +25%

Your supposed to be an enthusiast, its not rocket science.

 

[Martini1991]

You have absolutely nothing to back up your argument, you're branding around pointless figures and lying about what AMD have/haven't said. Softpedia article ISN'T official and SHOULD be taken with a MASSIVE amount of salt.
If you don't remember this article for example ; http://news.softpedia.com/news/AMD-Says-Bulldozer-is-50-Faster-Than-Core-i7-CPUs-177958.shtml They are far from reputable.

I've posted their slide which disagrees with you, and yet you STILL go on.
Where are you getting this 10-15% IPC increase from? AMD look like they're going for an overall 10-15% increase through IPC/Clocking, the power improvements could lower TDP/Yield higher overclocks or whatnot.

 

[humbug]

You have absolutely nothing to back up your argument, you're branding around pointless figures and lying about what AMD have/haven't said. Softpedia article ISN'T official and SHOULD be taken with a MASSIVE amount of salt.
If you don't remember this article for example ; http://news.softpedia.com/news/AMD-Says-Bulldozer-is-50-Faster-Than-Core-i7-CPUs-177958.shtml They are far from reputable.

I've posted their slide which disagrees with you, and yet you STILL go on.
Where are you getting this 10-15% IPC increase from? AMD look like they're going for an overall 10-15% increase through IPC/Clocking, the power improvements could lower TDP/Yield higher overclocks or whatnot.

lol... the IPC improvements did not come from Softpedia, it came from someone who benched a Trinity CPU, right here... http://www.tomshardware.co.uk/a10-5800k-a8-5600k-a6-5400k,review-32463-2.html

Trinity vs FX-8150 with one active core and running at the same clocks, Trinity beats the FX-8150 by 15%, that is exactly how your slide there envisaged.

 

[Martini1991]

The Trinity results are all over the place if you look at more than one page, ranging from better than Llano or worse CPU wise.
While it may have been X amount better in that example, it could be only Y better in another etc.
Extrapolating 15% in that one page is wishful thinking.

If Tom's had done a comprehensive review of that Trinity CPU in CPU tasks against the FX4 at the same clock speed, then we'd have a MUCH better understanding of how they compare.

 

[humbug]

The Trinity results are all over the place if you look at more than one result, ranging from better than Llano or worse CPU wise.

Yes, i have seen those results. Llano was always much faster than BD, just as K10 is, Llano is a K10 DIE shrink

There are 2 aspects in that where Llano is unbeaten, one is floating point performance, the other cache performance. Cache performance is not to indicative to real world use, whats more the benching software they used did not use SSE4 so the result is skewed in most situations anyway, even Toms Hardware conceded that.

floating point performance however, sticks. no its not a completely across the board everything beaten, its also unreasonable to expect that from AMD or Intel.

What matters is, the bulk of the performance, and what matters in the real world is up, significantly.

If Tom's had done a comprehensive review of that Trinity CPU in CPU tasks against the FX4 at the same clock speed, then we'd have a MUCH better understanding of how they compare.

They did put it up against the FX-8150 and it beat it convincingly.

But yes, i would also like to see a far more comprehensive test, and we will.

 

[Martini1991]

Either way, you've extrapolated an IPC improvement of 15% when you shouldn't have done.
Until there's a comprehensive review, stating 25% overall improvement is a hype fail.

Still think Tom's should have done an FX4 against Trinity review.
Llano has an IPC deficit against Deneb (Due to L3 cache) So the 3.8GHZ Trinity losing out against the 2.8GHZ, you've got to wonder.
Which is all more reason to why you can't make outlandish claims.

They did put it up against the FX-8150 and it beat it convincingly.

Two tests and a cache latency, woopy flipping do, lets make 25% better claims!

Bet there's instances where an Ivy can soundly beat an SB by 10%, but it's overall IPC improvement is lower than the 5% mark.

 

[humbug]

Either way, you've extrapolated an IPC improvement of 15% when you shouldn't have done.
Until there's a comprehensive review, stating 25% overall improvement is a hype fail.

Your disputing Toms Hardware's test results, not me. They say the IPC is up 15%, I'm quoting them. Get your facts straight.

Llano has an IPC deficit against Deneb (Due to L3 cache) So the 3.8GHZ Trinity losing out against the 2.8GHZ, you've got to wonder.

ok? if you like.... dispite this it still beat the Bulldozer CPU Which has L3, again your disputing Toms Harware (Trinity vs FX-8150) do that if you must, don't blame me.

Your acting like a child

 

[Martini1991]

Your disputing Toms Hardware's test results, not me. They say the IPC is up 15%, I'm quoting them. Get your facts straight.

/Facepalm.
In those tests,, I don't dispute the 15% gain.
But you're the one bleating 15% IPC gain as an overall thing and then with the clocking of a 25% increase.
Again, I go back to IB can best SB by 10%, but overall in tests its more like 4%, until you get a comprehensive review, stop saying "15% IPC GAIN AND 10% CLOCK GAIN FOR 25% IMPROVEMENT OVER BULLDOZER"

Yes dispite this it still beat it, again your disputing Toms Harware, do that if you must, don't blame me.

Wut.

EDIT : Acting like a child?
*Ragequits*.

 

[Phreaky]

Who gives a **** if its 15% IPC or extra clockspeed, as long as its faster overall than the equivalent Bulldozer at a similar price?

I really have a hard time understanding why people need to act absolute *****s about a particular brand - do you people not understand how competition works? Its nearly as bad as the Graphics Card forum ffs.

 

[Martini1991]

Who gives a **** if its 15% IPC or extra clockspeed, as long as its faster overall than the equivalent Bulldozer at a similar price?

I really have a hard time understanding why people need to act absolute *****s about a particular brand - do you people not understand how competition works? Its nearly as bad as the Graphics Card forum ffs.

Clearly missing the point.

I understand clearly how competition works, I guess AMD currently don't, there's no options when you want a high end gaming CPU, my only option is Intel.
Unlike in the GPU arena where frankly AMD/Nvidia GPU's aren't that far away from each other.

 

[humbug]

/Facepalm.
In those tests,, I don't dispute the 15% gain.
But you're the one bleating 15% IPC gain as an overall thing and then with the clocking of a 25% increase.
Again, I go back to IB can best SB by 10%, but overall in tests its more like 4%, until you get a comprehensive review, stop saying "15% IPC GAIN AND 10% CLOCK GAIN FOR 25% IMPROVEMENT OVER BULLDOZER"



Wut.

EDIT : Acting like a child?
*Ragequits*.

IPC (intrusions per clock) is one of a few aspects in a CPU's performance, if a program uses IPC and it is up, then it is up full stop for every other program dependant on IPC. (because the IPC performance is up) If a program uses floating point and it is down, then floating point is down full stop, or if its a combination of floating point and IPC then its likely to be a mixed bag.

That is why IB ranges in performance increase of 2 - 10%, it is also why Trinity varies against Llano.

 

[elroberto]

The test on Tom's hardware did show that 15% improvement in IPC compared to bulldozer and if Vishera does show similar clock frequency increases to Trinity then it is perfectly feasible that it'll see a 25% plus performance boost to bulldozer. Put it at a good price point and it becomes a very interesting chip indeed.

 

[Martini1991]

IPC (intrusions per clock) is one of a few aspects in a CPU's performance, if a program uses IPC and it is up, then it is up full stop for every other program dependant on IPC. (because the IPC performance is up) If a program uses floating point and it is down, then floating point is down full stop, or if its a combination of floating point and IPC then its likely to be a mixed bag.

That is why IB ranges in performance increase of 2 - 10%, it is also why Trinity varies against Llano.

Thank you captain obvious.

Apply the same logic to Trinity versus Bulldozer, you can't possibly bleat on about an overall 15% IPC increase when we've had so little tests since there's so many factors.

 

[Martini1991]

The test on Tom's hardware did show that 15% improvement in IPC compared to bulldozer and if Vishera does show similar clock frequency increases to Trinity then it is perfectly feasible that it'll see a 25% plus performance boost to bulldozer. Put it at a good price point and it becomes a very interesting chip indeed.

Yes it's perfectly feasible if it's 100% scaling and it's an overall 15% IPC improvement.

But we've had TWO benchmarks to gauge the difference, you can't make assumptions upon that.

EDIT : If PD does turn out to be 25% better than Zambezi, halving the gap between Intel and themselves (Until Haswell recreates the gap) I'd damn well be interested in buying a PD.
Just can't see it happening.

 

[humbug]

Yes it's perfectly feasible if it's 100% scaling and it's an overall 15% IPC improvement.

But we've had TWO benchmarks to gauge the difference, you can't make assumptions upon that.

EDIT : If PD does turn out to be 25% better than Zambezi, halving the gap between Intel and themselves (Until Haswell recreates the gap) I'd damn well be interested in buying a PD.
Just can't see it happening.

I think you need to take a step back, breath.... and then come back to this 'much later'

What i said was IPC is up 15% and its clocked 10% higher, (which it is) i said that makes the performance +25%, i'm not being exact with that but its a reasonable proposition given that Trinity is the same arch as Bulldozer!

That's it.... don't agree? then Complain to Toms Hardware for (monipulating there own result? maybe?)

And i'm done with you until you clam down and look with a little more perspective.