Whats after Haswell?

[Martini1991]

read the thread

I did, you've just interpreted it differently (And wrongly)

 

[jonneymendoza]

I did, you've just interpreted it differently (And wrongly)

care to explain?

 

[Martini1991]

care to explain?

No one has stated the difference in clock for clock performance is 10% to Haswell, it's probably close to 20%, with Ivy being closer to 15%.
Then you've got the extra clocking potential, which is about 20%.
Haswell could be upwards of 40% faster per core than an i7 9XX.
And we don't know 24/7 clocks yet, it could increase the threshold.

 

[jonneymendoza]

I'm not quite sure where you get that idea from.

i7 920 vs i7 2600k

Even the 2600k is more than 20% ahead of the i7 920 in allot of those benchmarks

i7 920 vs i7 3770k

The 3770k just extends this lead, so the Haswell i7 will probably be knocking on to being 40/50% faster than the i7 920 at stock.

Don't forget the Sandy and Ivy i7's clock higher than the old i7 920 do as well.

here. says its 50% faster lol

 

[Martini1991]

here. says its 50% faster lol

That's stock for stock comparison, which doesn't contradict anything I've said.

 

[ubersonic]

here. says its 50% faster lol

Being X% faster when you have a 1Ghz speed advantage is not the same as being X% faster clock for clock.

 

[jonneymendoza]

So is it 20% faster clock for clock?

 

[ubersonic]

So is it 20% faster clock for clock?

Impossible to know until its released.

 

[Martini1991]

So is it 20% faster clock for clock?

Haswell should be, but I'm guessing, Ivy's about 15%.
But then it clocks higher, so as said, you could get upwards of 40% core for core over a 4GHZ 920.
But for all I know, Haswell can open up more overclocking potential to make it 45% core for core.

 

[jonneymendoza]

Haswell should be, but I'm guessing, Ivy's about 15%.
But then it clocks higher, so as said, you could get upwards of 40% core for core over a 4GHZ 920.
But for all I know, Haswell can open up more overclocking potential to make it 45% core for core.

i dont care about OC potential i care about clock for clock speeds at present.

if haswell is 15% faster then Ivy, does that mean Ivy was only 5% faster then my 920 clock for clock?

 

[Martini1991]

i dont care about OC potential i care about clock for clock speeds at present.

if haswell is 15% faster then Ivy, does that mean Ivy was only 5% faster then my 920 clock for clock?

How are you misreading so easily?
I said Ivy's about 15% better clock for clock than your i7.

Why wouldn't you care about OC potential? You won't be running the same clock speeds.

 

[WingZero30]

Jonney I am sure when haswell is released you can make comparison from the benches that would follow after the launch.
Iirc Sandy was 15-20% faster than nehalem reading from the reviews/benches. Ivy was considered to be 5-10% faster than Sandy. Now Haswell seems to be 10% faster than Ivy clock-for-clock.

So Haswell could be 30-40% faster than i7 920 clock for clock.

A lot of it also depends on what software you are running as you may get different performance numbers.

 

[Martini1991]

Aim low and avoid disappointment.

 

[jonneymendoza]

Aim low and avoid disappointment.

Can you please comment on the above post of yours stating it will be 30-40% faster yet you say it will be much lower.

clock for clock what is it then? nehalem is 2 gens old.

ivy/sandy what was the clock for clock increase? 10%?

 

[Spook187]

Just get ivybridge and be done with it, oh what's next after this and that.

 

[Martini1991]

http://www.anandtech.com/bench/Product/99?vs=287
That's as close clock to clock comparison as you're getting.
That's about a 10-15% difference, my aim low and avoid disappointment would have that as a 10% IPC difference.
SB to IB is around 5% IPC, Haswell isn't going to be much more.

If I was going to be a little less conservative, I'd say ~25% IPC difference between Haswell and Nehalm.

 

[Alasdairhurst]

Same boat as me, 4770k will be on the cards.

Count me in for that too

 

[tbh]

Count me in for that too

Me too most likely, depending on whether reports of them being "an overclockers dream" are true or not of course.

 

[Alasdairhurst]

Me too most likely, depending on whether reports of them being "an overclockers dream" are true or not of course.

Well Linus thinks that this refers to the VRM being in the CPU now instead of on the board, meaning that you can now overclock the same levels on any board. This just leaves expansion ports and looks down to the motherboard manufacturer

 

[Gripen90]

I'm interested to see how many system crashes Haswell will cause if never PSU can't do the 0,05amp 12v+ requirement for the new c6/c7 states.
I guess the motherboard manufacturer per default will disable c6/7 states in bios.