HASWELL core I5: Performance Preview

[Arkaign]

These pre-previews always seem to use the most useless metrics for CPU performance.

I dunno. I seem to recall credible leaks showing BD to be a bit bad, as well as a few others that were right on target. I get what you're saying though, a full picture can't be had until competent folks like Anand get their mitts on the final hardware and test it right.

 

[mikk]

I suppose that in 3mdark11 the GPU might have been at 800Mhz instead of maximum of 1200Mhz

Of course it run with 800 Mhz lol.

but even at 1200Mhz it would score ~800pts in GPU subtest while (now old) Trinity gets 1360pts and Richland will score even more more(~10-15%) on stock.

3dmark11 doesn't reflect real world performance, same as Vantage where Intel graphics perform extremely good.

 

[Maximilian]

These pre-previews always seem to use the most useless metrics for CPU performance.

Yeah... this...

3dmark? You kidding me? What does xxxx points in 3dmark actually mean anyway, 3dmark, furmark, hurpadurpmark its all pretty useless for determining performance IMO.

How long does it take to encode a movie? What FPS does it get in a modern game engine (civ V, BF3 etc) using a discrete GPU? And of course the always interesting how much power is it using to do these things that's what i would like to know

 

[TuxDave]

I dunno. I seem to recall credible leaks showing BD to be a bit bad, as well as a few others that were right on target. I get what you're saying though, a full picture can't be had until competent folks like Anand get their mitts on the final hardware and test it right.

I kind of wish I had some paper like this:

http://www.spec.org/workshops/2007/...ance_Characterization_SPEC_CPU_Benchmarks.pdf

For every benchmark out there so that we know what each score actually reflects.

 

[sm625]

Performance per watt (PPW) is really the only metric that matters. We already know it will only be marginally faster than ivy in IPC terms. PPW, especially gaming framerate PPW is the big unknown.

 

[Arkaign]

Performance per watt (PPW) is really the only metric that matters. We already know it will only be marginally faster than ivy in IPC terms. PPW, especially gaming framerate PPW is the big unknown.

What? PPW doesn't matter that much to me personally. I'd rather have a CPU sucking down 300W if it's twice as fast as my 2700K @ 5ghz, than a CPU 3% faster with 2/3rds the power usage.

I do know that big performance increases seem to have gone by the wayside, and that most progress seems focused on efficiency, but I still want some legit performance reason to upgrade.

 

[fixbsod]

Looks like intel is gonna give AMD a breather this round, a healthy dose of competition might give the PC a nice shot in the arm.

 

[Homeles]

Looks like intel is gonna give AMD a breather this round, a healthy dose of competition might give the PC a nice shot in the arm.

Looks like you're reading too much into random benchmarks.

 

[IntelEnthusiast]

I havent seen anything on the performance of our next generation of processors but it seems to strike me a bit odd that it is running at 2.6GHz at 83w. We have the Intel® Core™ i5-3330S running at base 2.7GHz before turbo at 65w. Also the voltage is almost 25% lower than our 3rd generation Intel Core processors (1.1v down to .8v), it looks like it being downclocked but heck I really don't know.

 

[Homeles]

Haswell does have the voltage regulator on package, which raises the TDP of the package but should reduce overall system consumption.

 

[Phynaz]

Voltage regulator isn't going add .1w to the TDP. Voltage regulators are just a couple of transistors and diodes.

 

[aaksheytalwar]

Why is it so slow?

 

[GlacierFreeze]

Why is it so slow?

Maybe we should tape a cheetah to its back.

 

[fixbsod]

No, I just know that intel appears to be focusing more on perf/w to the point that IPC appears to be stagnant. While nice for mobile / low power uses this is moderately useless for the power pc market and considering the 8350 is close to an i5-3570k it will allow for some time for AMD to catch up.

I'm still an intel fanboy but I ain't afraid to call em like I see em.

edit--and this could be the reason for the rumored 8 core SB -- intel knows Haswell aint got much in the tank and is willing to toss an 8 core into the desktop ring to satiate the uber end battle.

Looks like you're reading too much into random benchmarks.

 

[Face2Face]

maybe we should tape a cheetah to its back.

lol :d

+1

 

[Haserath]

I Haspoor benchmarks.

 

[Stoneburner]

The pre-release information on CPU's has been pretty accurate since Conroe. The last time it was dead wrong with when k8 came through, knocking down buildings.

 

[parvadomus]

specs look impressive on paper but they only changed data structures for bigger ones and added new instructions. This wont account for a big change in IPC, however it will help to have better instruction parallelism inside each core. Then we have the decoupled L3 cache, it is a step back from IPC point of view, but a good step for energy efficience.
I think IPC gains will be minimal, if any in some applications. AMD will have a great chance this time with SR, if its really as good as they say (at least 20%+ IPC).

 

[inf64]

I think IPC gains will be minimal, if any in some applications. AMD will have a great chance this time with SR, if its really as good as they say (at least 20%+ IPC).

IPC is relative term and varies wildly in various workloads. AMD will not only need "IPC", they will need "moar cores" (insert hand drawn image/chart please) to compete with 6c/12t+ IB parts . My estimate is that they will need 12T parts on desktop/1P server as basis, just like FX8xxx series is today. Adding 2 more modules on 28nm process will just slightly inflate the die area (given that cache parts scale well with shrinks), by about ~10%. Power will most likely stay the same for 12T parts with better "ipc" and folks who need performance and want amd won't mind it much.

 

[Puppies04]

What? PPW doesn't matter that much to me personally. I'd rather have a CPU sucking down 300W if it's twice as fast as my 2700K @ 5ghz, than a CPU 3% faster with 2/3rds the power usage.

Unfortunately you don't matter to intel. They aren't making CPUs for you they are making them for the other 99.9% of the market. Sad but true.

 

[mrmt]

IPC is relative term and varies wildly in various workloads. AMD will not only need "IPC", they will need "moar cores" (insert hand drawn image/chart please) to compete with 6c/12t+ IB parts

Oh, this strategy worked wonders on servers. Did you check what happened with their market share?

 

[sm625]

What? PPW doesn't matter that much to me personally. I'd rather have a CPU sucking down 300W if it's twice as fast as my 2700K @ 5ghz, than a CPU 3% faster with 2/3rds the power usage.

If it were even remotely possible to make a 300W chip that was twice as fast as a 2700K @ 5GHz (in IPC terms) then you can bet intel would be selling it. But that is like asking for an 7 GHz wolfdale back in 2007. (7 Ghz is about how fast a wolfdale would need to be to produce a 5GHz sandy bridge super pi single thread score.)

If you were to design a custom 200W peltier which perfectly fit a 4770K's die (in 3 dimensions), I bet you could theoretically run it at near 7Ghz, giving you a ~300W chip that is nearly twice as fast as a 2700K. It would be close, but surely in the ballpark of twice as fast.

 

[exar333]

This makes me feel better and better about my 3930k purchase. At this point, with the rumored IB-E offerings, I will get another 18 months (at least) from my current setup.

November 2008 - November 2011 - i7 920
November 2011 - November 2014? - i7 3930k

Feeling pretty good about my rig choices lately.

 

[Grooveriding]

lol

 

[parvadomus]

IPC is relative term and varies wildly in various workloads. AMD will not only need "IPC", they will need "moar cores" (insert hand drawn image/chart please) to compete with 6c/12t+ IB parts . My estimate is that they will need 12T parts on desktop/1P server as basis, just like FX8xxx series is today. Adding 2 more modules on 28nm process will just slightly inflate the die area (given that cache parts scale well with shrinks), by about ~10%. Power will most likely stay the same for 12T parts with better "ipc" and folks who need performance and want amd won't mind it much.

I think a 10-core SR will match a 6-core 12-thread IB part. However, AMD wont make it. I doubt they will continue with FX series. 6-core Kaveri will probably walk over a 3570K in MT workloads, why bother with FX..