DigiTimes: Intel to launch Haswell on March 2013

[xVeinx]

The laptop parts command the better price, so the better parts will be used in laptops instead of desktop. There may be K parts that have the GT3 (or they may be released sometimes after during a speed bump).

[Brice MJ]

by the time haswell comes out, it will be facing steamroller based apu's. im pretty sure amd will still have the graphics crown, because those would be based on gcn.

Haswell quads+GT3 will be out by early Q2, maybe March 2013. According to Charlie there will be a Trinity 2.0 by the time Haswell launches and Kaveri is not coming till H2/2013.

[god_43]

Haswell quads+GT3 will be out by early Q2, maybe March 2013. According to Charlie there will be a Trinity 2.0 by the time Haswell launches and Kaveri is not coming till H2/2013.

i thought haswell was pushed back to late q2? well charlie may be right, i guess we will know more once the next analyst day comes round.

[jogshy]

I wonder if Hasswell uses that crappy TIM Intel used in Ivy Bridge...

[ajaidev]

The crappy TIM was used to keep performance in check using thermals, the effect that the TIM will have on the CPU was known to most if not all in Intel.

[god_43]

The crappy TIM was used to keep performance in check using thermals, the effect that the TIM will have on the CPU was known to most if not all in Intel.

....i dont understand this? what do you mean to keep performance in check? was performance running away (lol)?

[Iconyu]

The crappy TIM was used to keep performance in check using thermals, the effect that the TIM will have on the CPU was known to most if not all in Intel.

When the cover and TIM were removed the results didn't improve too much, if anything it was to stop people trying to push chips into their inevitable RMA zone. They just overheat 200mhz sooner than they would have normally, a lot of overclockers would have just used voltage and better cooling to go past that point and those chips would have died in weeks.

[JaD]

Yup. There are fields and fields of burned Intel CPUs near Santa Clara.

[Barr3l Rid3r]

New socket again? Intel suxxx! Where is the upgrade factor? ZIF Sockets? Why? They should just solder a CPU on it....

I've just bough a Z77 MB and it's already obsolete while my 1 year 990FXUD3 it's a piece of technology.

[zalbard]

....i dont understand this? what do you mean to keep performance in check? was performance running away (lol)?

Yeah, too close to SB-E.

[Iconyu]

Yup. There are fields and fields of burned Intel CPUs near Santa Clara.

You think I'm joking? http://community.futuremark.com/foru...600K&p=1681415

This idiot went through FOUR 2600k's, he just put them up to 1.5v and let his water cooling handle the heat, Intel replaced everyone of them without question. He didn't ask for any advice he just decided he needed that voltage to hit 4.9ghz. If he tried that on an Ivy Bridge, it would never let him do it, his water cooling would never overcome the TIM.

It seems crazy, but if I can find one idiot, I'm sure Intel found enough to make them go through with this change and incur our anger. So if Intel do the same thing again with Haswell, you know who to blame.

[ajaidev]

When the cover and TIM were removed the results didn't improve too much, if anything it was to stop people trying to push chips into their inevitable RMA zone. They just overheat 200mhz sooner than they would have normally, a lot of overclockers would have just used voltage and better cooling to go past that point and those chips would have died in weeks.

That was because of the thermal distribution limitation. The heat needs to be spread out with the IHS in order to deal with it better so replacing the stock TIM with a better one will no doubt improve temp's of the CPU. Better performance "Via OC" in IVB is a complicated affair, there is internal heat and core heat and sensors dont monitor the whole chip they monitor certain sections or certain surroundings. The chip can die if other parts are exposed to excess heat things Intel does not document openly.

....i dont understand this? what do you mean to keep performance in check? was performance running away (lol)?

Well kinda yes the gap between the SB-E and the IVB had to be maintained. Already there is a core difference but it may not be enough with IVB quads at 77W TDP its important to have a way to check the IVB performance. It was not about saving a buck using a stupid TIM, it was all about something more.

[god_43]

Yeah, too close to SB-E.

lol.

That was because of the thermal distribution limitation. The heat needs to be spread out with the IHS in order to deal with it better so replacing the stock TIM with a better one will no doubt improve temp's of the CPU. Better performance "Via OC" in IVB is a complicated affair, there is internal heat and core heat and sensors dont monitor the whole chip they monitor certain sections or certain surroundings. The chip can die if other parts are exposed to excess heat things Intel does not document openly.



Well kinda yes the gap between the SB-E and the IVB had to be maintained. Already there is a core difference but it may not be enough with IVB quads at 77W TDP its important to have a way to check the IVB performance. It was not about saving a buck using a stupid TIM, it was all about something more.

ahh i understand now. the man who was once blind; is blind NO MORE!

[Olivon]

2X graphical performance vs IB

[informal]

The only thing I don't understand in the above video is that the speaker mentions SB core for comparison and the slide/video says 3rd gen Core. But in any case, Haswell looks to be a massive upgrade over IB,bot in terms of CPU and GPU performance. All this at lower/same power. I'm impressed.

[Olivon]

Intel Developer Forum 2012 - Dadi Perlmutter Keynote Live Blog - AnandTech

[kl0012]

Additional info about Haswell CPU cores from Anadtech:

http://www.anandtech.com/show/6263/i...sure-live-blog

- 2xFMA per core, per cycle (2x FP throughput over IB) vs. 1xFMA per module in Bulldozer.
- 8 execution ports vs. 6 in IvyB (additional integer/branch port + store address port)
- Improved branch prediction
- Increasing size of buffers internally, giving us larger OoO window

[jogshy]

I want Hasswell-DT to use DDR4 omg !

I think they disabled tessellation in that presentation

[god_43]

I want Hasswell-DT to use DDR4 omg !

I think they disabled tessellation in that presentation

yeah i am not buying it until benches come out. the current 4k chokes pretty hard in "GPU" bound games, so if all they did was up the eu's....idk we will see.

[Gorki]

I'm really starting to love it...

[M.Beier]

I want Hasswell-DT to use DDR4 omg !

I think they disabled tessellation in that presentation

yeah i am not buying it until benches come out. the current 4k chokes pretty hard in "GPU" bound games, so if all they did was up the eu's....idk we will see.

Sorry for being a here, but do you really expect Intel to change something that central in the architecture, across same range of products?? I cant even imagine the cost of doing that, but it sure would not make sense...
Haswell is not DDR4, nor will the product after be DDR4, as its just a DIE shrink.... So forget DDR4 for both 2013 and 2014 on Intel, early Q2 2015 is more realistic.... But really, except the higher density, why would you want DDR4 for the time being??? DDR3 has much juice left, and the cost are at a pricepoint I only remember having seen once before in my life.... When SDRAM were becoming 'old' and RD were the new thing, soon followed by DDR...

Gatman John (Gautam), perhaps you can give your input to this thread? Not as an Intel rep, not any NDA info, just an engineers words on how complex and expensive an affair it would be to make one part of a range with a different IMC.

[B-Shot]

I wish they would omit the igp on the K series CPU,s, one less thing to fail and hold heat.....room for more cores
Yea this is XS more is never enough.