Haswell Benchmarks?

[jones377]

Piledriver has AVX, FMA4 and FMA3, it's instruction set is the most advanced of any x86 processor. Kaveri will have AVX2 and will hit the market at about the same time as anyone starts using AVX2, so no advantage there for haswell there either.

Source on AVX2 in Kaveri?

 

[MisterMac]

Source on AVX2 in Kaveri?

Isn't really any.

PieSquared likes to make up stuff - and embed between claims.
I actually wonder sometimes - if pie is charlie introllcognito.


Would explain so much.

 

[Haserath]

Piledriver has AVX, FMA4 and FMA3, it's instruction set is the most advanced of any x86 processor. Kaveri will have AVX2 and will hit the market at about the same time as anyone starts using AVX2, so no advantage there for haswell there either.

There will be an advantage for Haswell.

It's called just being plain better in every single way.:whiste:

 

[piesquared]

Source on AVX2 in Kaveri?

No source to speak of.

 

[piesquared]

There will be an advantage for Haswell.

It's called just being plain better in every single way.:whiste:

Source that it will be better in every single way?

 

[CHADBOGA]

No source to speak of.

So you just made it up, expressing your dreams? :awe:

 

[CHADBOGA]

Source that it will be better in every single way?

https://twitter.com/FPiednoel/status/297005032992284673

 

[piesquared]

So you just made it up, expressing your dreams? :awe:

No to both.

 

[piesquared]

https://twitter.com/FPiednoel/status/297005032992284673

lol!! :ROFL:

 

[Haserath]

Source that it will be better in every single way?

Idk, I thought we were makng stuff up?

 

[piesquared]

Idk, I thought we were makng stuff up?

Who, you and Chad? Apparently so, yes.

[edit]

Anyway, you two lovebirds carry on. Us 'AMD fanboys' have to walk on eggshells round these parts so i'll bow out for now.

 

[CHADBOGA]

lol!! :ROFL:

Francois ain't no Randy Allen.

 

[DrPizza]

I'm led to believe this is a neutral thread about benchmarks of one particular processor. Please do not derail this into another AMD vs. Intel thread.

And please, personal attacks (e.g., "Amd fanboy") are unwelcome in this forum.

 

[Piroko]

To get this sort of back to topic, any educated guesses at how much power VRMs consume on a desktop board and how much can be saved by integrating them?

 

[ShintaiDK]

To get this sort of back to topic, any educated guesses at how much power VRMs consume on a desktop board and how much can be saved by integrating them?

Well depends on configuration. Here is an example:

So there is alot to be saved. Plus all the other benefits and quality assurance.

 

[Piroko]

Well yea that's the theory behind it, but what could one reasonably expect in practice? Just looking at the slides I'd expect savings mainly with light workloads (websurfing workload), but low gains at full load and full idle (workstation, gaming / always-on). Am I correct?

 

[ShintaiDK]

Well yea that's the theory behind it, but what could one reasonably expect in practice? Just looking at the slides I'd expect savings mainly with light workloads (websurfing workload), but low gains at full load and full idle (workstation, gaming / always-on). Am I correct?

Full load can be pretty high. It depends largely on what board you use today. But 30-40W savings isnt impossible. On the other hand it would also be 10-15W in the lower range.

Idle is anything from a few W to as much as 20W.

Again, mileage will vary alot between different boards.

 

[Piroko]

I don't think 30-40W savings on a 85W desktop processor are realistic when the efficiency of all the measured vrms is between 81% and 89% @~70A load in your own graph. Even 5W would be a feat imho. As I understood, power savings are expected when the processor shifts between power states a lot, but I'm curious how much Intel engineers or knowledgeable people would expect.

 

[TuxDave]

To get this sort of back to topic, any educated guesses at how much power VRMs consume on a desktop board and how much can be saved by integrating them?

Don't forget there's also power efficiency gains for the cores as well.

 

[Piroko]

As well as possible efficiency gains through different transistors compared to IB, yes, but those are factors which we can only speculate about whereas compact dc/dc power delivery kinda is a hot topic for small devices and more people might have knowledge in that field...

 

[ShintaiDK]

I don't think 30-40W savings on a 85W desktop processor are realistic when the efficiency of all the measured vrms is between 81% and 89% @~70A load in your own graph. Even 5W would be a feat imho. As I understood, power savings are expected when the processor shifts between power states a lot, but I'm curious how much Intel engineers or knowledgeable people would expect.

http://www.techpowerup.com/reviews/Gigabyte/X79-UD5/8.html

Well, think twice then.

Not to mention raw mobo vs mobo:
http://www.anandtech.com/show/6144/asus-p8z77v-premium-review-a-bentley-among-motherboards/6

 

[Piroko]

I would've liked to see the differences on the same board with different vrms (different production charges maybe) but that's probably asking too much. The Techpowerup article states "may not prove to isolate the CPU power draw in all instances", measures with 130W processors and that a bios update of the Gigabyte board pushes power consumption by 26W at load. So there's definitely a software part in the difference which also is exaggerated because of the different peak power draw.

Still, point taken, Haswell might lower power consumption on 'shoddy' boards by quite a bit more.