Intel "Haswell" Speculation thread

[Dresdenboy]

Ummm...What? Intel had an integrated GPU long before AMD did.

Yep, as did Cyrix 15 or so years ago.
All these were not really usable for computing tasks.

 

[tweakboy]

How would the Haswell OC. The hexacore version ? I think with a good OC that 5 to 7 percent will be 15 percent. Then another 15 to 20 percent for having 2 extra cores.

I think Haswell 6 core will be 25 percent faster then sandy or ivy OCed.

Dont forget we OC the haswell,, maybe 6GHZ ?

This chip being hexacore also, will pownz any desktop chip. You guys are crazy if you think its not gonna have a big difference then ivy or sandy. A OCed hexacore will crush any oced Sandy or bridge.

 

[Blitzvogel]

Ummm...What? Intel had an integrated GPU long before AMD did.

But wasn't it AMD that promoted the "proper GPU on CPU die" concept, and then turned it into a case for "GPGPU compute on CPU?"

If anything Intel was countering the graphics on CPU idea that AMD originally had. Intel wouldn't want high end GPGPU compute on their IGP because it doesn't tie directly into them expanding the capability of their actual CPU cores which are much more flexible.

 

[pantsaregood]

How would the Haswell OC. The hexacore version ? I think with a good OC that 5 to 7 percent will be 15 percent. Then another 15 to 20 percent for having 2 extra cores.

I think Haswell 6 core will be 25 percent faster then sandy or ivy.

Haswell will be 10-15% faster than Sandy Bridge per clock. The clock ceiling of ~4.6 GHz for Sandy Bridge and ~4.8 GHz for Ivy Bridge will probably land somewhere around 5.0 GHz on Haswell. Expecting six-core Haswell parts on a mainstream platform isn't likely to happen, however.

Based solely on history, Haswell units will initially ship at slightly higher clock speeds than equivalently priced parts from the previous generation.

Throw all of those together, and you're likely to see somewhere around 200 MHz higher clocks at stock (compared to current products) and an average of ~12.5% increase in performance per clock. This would place an i7-4770 (whatever they decide to call it) at 3.6 GHz - ~12.5% faster than a 3770 per clock, as well as being clocked 5.9% higher. The top mainstream part will likely outperform the current top mainstream part by ~19%.

 

[Phynaz]

But wasn't it AMD that promoted the "proper GPU on CPU die" concept, and then turned it into a case for "GPGPU compute on CPU?"

Nope, Intel was first there also.

Unless you are saying AMD opened their mouths like they did with their dual core fiasco, giving Intel the chance to out-engineer them and deliver sooner? That could be.

 

[Concillian]

Haswell will be 10-15% faster than Sandy Bridge per clock. The clock ceiling of ~4.6 GHz for Sandy Bridge and ~4.8 GHz for Ivy Bridge will probably land somewhere around 5.0 GHz on Haswell. Expecting six-core Haswell parts on a mainstream platform isn't likely to happen, however.

Based solely on history, Haswell units will initially ship at slightly higher clock speeds than equivalently priced parts from the previous generation.

Throw all of those together, and you're likely to see somewhere around 200 MHz higher clocks at stock (compared to current products) and an average of ~12.5% increase in performance per clock. This would place an i7-4770 (whatever they decide to call it) at 3.6 GHz - ~12.5% faster than a 3770 per clock, as well as being clocked 5.9% higher. The top mainstream part will likely outperform the current top mainstream part by ~19%.

This sounds like wishful thinking.

Haswell-E might fair better for enthusiasts, but the price will be an issue. I think SB was the start of the decline for enthusiasts like us.
SB - fully locked CPUs to force overclockers to pay up.
IB - TIM issue reduces OC capability vs. SB
Haswell - On package voltage regulation puts a strict upper bound on power consumption / overclocking.

I think we'll be lucky to have a Haswell that is the same performance as overclocked IB 3570k. Intel is focused more on the low power market, and will have to make concessions on top end power consumption to broaden their low power appeal. It may still have the same TDPs, but I think there will be less and less OC margin as they tweak the design to satisfy lower power designs, plus they'll have to bump speeds again by 100 MHz just to justify the new product like they did with IB.

They're fully into holding advancements in their back pocket and metering out only what is necessary to justify the next gen. This way when they reach the inevitable wall, they can pull these things out one at a time to extend a few more generations.

As has been discussed already in this thread, 10-15% IPC may be optimistic as well.

 

[Blitzvogel]

Unless you are saying AMD opened their mouths like they did with their dual core fiasco, giving Intel the chance to out-engineer them and deliver sooner? That could be.

That's what I meant

 

[Idontcare]

Hey, that's completely your prerogative. You can spend your money any way you want. I'm just giving a heads up straight up 2 core, 4 core, X core CPUs isn't where tech is going. Just look at the supercomputers, how did that happen right?

Too bad Haswell-E says otherwise too you.

I'd rather have 2-4 cores more and no IGP..and I'll pay for such a solution.

I gather you have absolutely no idea who TuxDave is, what company he works for, and what he does to earn his paycheck from his employer

<- cannot put enough of these here to express my ROFL upon reading your "rebuttal" to a rather transparently obvious "heads-up". Wow

 

[meloz]

A 10% performance improvement over Ivy is acceptable for Haswell. Most people cannot peg their existing $200 IVB CPU at 100% usage in anything apart from benchmarks anyway.

The main focus -again- will be on improving the iGPU. If Intel can gain another 25-30% improvement in that area, it would be a big win. It would bring Intel very close to AMD, and then in a couple more generations maybe they can even dream of overtaking AMD in the iGPU department.

Intel will also be using a better TIM in Haswell if I heard the podcast right, so Haswell could be another good overclocker like Sandy Bridge and the enthusiasts won't have much to complain about either.

 

[Ajay]

Hey, that's completely your prerogative. You can spend your money any way you want. I'm just giving a heads up straight up 2 core, 4 core, X core CPUs isn't where tech is going. Just look at the supercomputers, how did that happen right?

If you're talking Cray - really fast vector processing...IIRC.

 

[sefsefsefsef]

The main focus -again- will be on improving the iGPU. If Intel can gain another 25-30% improvement in that area, it would be a big win. It would bring Intel very close to AMD, and then in a couple more generations maybe they can even dream of overtaking AMD in the iGPU department.

If Haswell really does have 40 EUs + a large eDRAM cache (hopefully like 32-64 MB) specifically for graphics, I would expect more in the 100-200% performance improvement range, not just 25-30%.

 

[Lonbjerg]

I gather you have absolutely no idea who TuxDave is, what company he works for, and what he does to earn his paycheck from his employer

*snip*<- cannot put enough of these here to express my ROFL upon reading your "rebuttal" to a rather transparently obvious "heads-up". Wow

He could be the tooth-fairy for all I care...don't alter my stance...nice "Appel to authority"....when we are talking about my subjective stance...:whiste:

I don't care if Otellini himself praised tie migdet-wanna-be-IGP...wouldn't alter my stance...or opinion.

I have no need for a crappy IGP...no matter who gets a boner on for the IGP...simple as that...nice fail though! :whiste:

 

[Nemesis 1]

But wasn't it AMD that promoted the "proper GPU on CPU die" concept, and then turned it into a case for "GPGPU compute on CPU?"

If anything Intel was countering the graphics on CPU idea that AMD originally had. Intel wouldn't want high end GPGPU compute on their IGP because it doesn't tie directly into them expanding the capability of their actual CPU cores which are much more flexible.

None of what your saying is true, The internet is a wonderful tool . If one uses good judgement. In 2004 Intel bought elbrus . A BIG DEAL. In the same year intel had 3 pilot programms going./ One was choosen which turned out to be like the larrabbee project. Both AMD and NV also ATI new what intel was up to . Thats way befor AMD even had dreams of fusion . The only thing that went wrong was AMD buying ATI rather than hooking up with NV . Had NV /AMD hooked up this game be way better . It would have also opened the door for intel to by ATI latter. I happy the way it turned out . As ATI stock holder AMD payed 2x what the company was worth. Now NV is in and AMD is on its way down to the tablet phone market . The problem is to much time going the wrong way. NV made good choices . Intel made good choices and AMD made bad choices its that simple , Hector the reckor choose ATI for fear of his own job . A personnal decision based on personnal loss over a companies future . Hector got everthing he had coming. AMD the company was trashed because of personnal greed at the highest level . End of truth story.

 

[Nemesis 1]

He could be the tooth-fairy for all I care...don't alter my stance...nice "Appel to authority"....when we are talking about my subjective stance...:whiste:

I don't care if Otellini himself praised tie migdet-wanna-be-IGP...wouldn't alter my stance...or opinion.

I have no need for a crappy IGP...no matter who gets a boner on for the IGP...simple as that...nice fail though! :whiste:

Its over we been in agreement for a time , But your wrong it will be all in one before 2020 time permitting that is. Everyone for the most part can clearly see this . with Few exception .

 

[Nemesis 1]

How would the Haswell OC. The hexacore version ? I think with a good OC that 5 to 7 percent will be 15 percent. Then another 15 to 20 percent for having 2 extra cores.

I think Haswell 6 core will be 25 percent faster then sandy or ivy OCed.

Dont forget we OC the haswell,, maybe 6GHZ ?

This chip being hexacore also, will pownz any desktop chip. You guys are crazy if you think its not gonna have a big difference then ivy or sandy. A OCed hexacore will crush any oced Sandy or bridge.

None here can make any call on how much faster it will be . But I real sure AVX2 will have 4 or five programms already recompiled for the launch of haswell . your smallish 25% increase is just that smallish . AVX2 is an ipc increase thats simply needs a recompile oand new programms made just for Intels haswell chip . Haswell will infact be the biggest step forward in Cpu history. We will see that on ATs review of Haswell as at least 4 AVX2 programms will be ready to benchmark. Those 4 programms alone will change the % improvement we see by alot more than 25% this time there will be NO throwing out the high and the low . Everthing will be averaged as a %. of total improvement.

 

[NTMBK]

My personal speculation- Intel will end the madness, and just call Haswell the Pentium V.

 

[Nemesis 1]

99% of users will be fine with an APU, the real boys(1%) will always need more CPU performance/threads and discrete GPUs

No matter how fast the APUs will become(fused, more apps to take advantage off etc), a CPU + Discrete GPU will always be faster in both apps and Games.

Your dreaming . THE APU isn't going to win . The IGPU is . AMDs APU name will be disgarded. As Intels cpus will be better at everthing including graphics as iGPU name also disappears. Intels cpu will be a cpu that does it all . NV coined the GPU name what was it that ATI coined theirs. Its been so long I forgotten those 3 small letters from ATI . APU is bound for the same result. Because INTEL will never use APU.

 

[Dufus]

OC Haswell! If it's going to have integrated VRM control then maybe OC'ing is going to be somewhat limited.

 

[Nemesis 1]

Haswell will be 10-15% faster than Sandy Bridge per clock. The clock ceiling of ~4.6 GHz for Sandy Bridge and ~4.8 GHz for Ivy Bridge will probably land somewhere around 5.0 GHz on Haswell. Expecting six-core Haswell parts on a mainstream platform isn't likely to happen, however.

Based solely on history, Haswell units will initially ship at slightly higher clock speeds than equivalently priced parts from the previous generation.

Throw all of those together, and you're likely to see somewhere around 200 MHz higher clocks at stock (compared to current products) and an average of ~12.5% increase in performance per clock. This would place an i7-4770 (whatever they decide to call it) at 3.6 GHz - ~12.5% faster than a 3770 per clock, as well as being clocked 5.9% higher. The top mainstream part will likely outperform the current top mainstream part by ~19%.

What are you talking about . What clock ceiling of SB at 4.6 ghz. LOL on air . Than your IB limit is a joke . for air maybe your correct . But thats not the ceiling not even close. I am intel fanboy but your a fantasy fanboy making AMD better than it really is ON average . and you lower intels performance on average , You are in a fantasy world of your own creation .

 

[BenchPress]

I'm just giving a heads up straight up 2 core, 4 core, X core CPUs isn't where tech is going.

I'm afraid you're not aware of the revolution that is about to hit us... It's certainly true that the increase in core count has slowed down in the past few years, but that's because with today's CPUs it's quite hard for developers to juggle many threads (because these CPUs only have very primitive and slow means for synchronizing between threads).

Haswell however will mark a major milestone in the development of multi-threaded software. It brings two key pieces of technology as part of its TSX extension: hardware transactional memory, and hardware lock elision. Both are invaluable to efficiently scale to more cores/threads.

Furthermore, the AVX2 extension shows that Intel realizes the unique benefits of homogeneous throughput computing. Haswell will no less than double the throughput per core. And to continue scaling the throughput in the future, they will most definitely start increasing the core count again.

So both these technologies seem to be part of a longer term plan to empower application developers to create multi-threaded high-throughput applications.

 

[sefsefsefsef]

BenchPress, I'm afraid your language isn't EXTREME TO THE MAX enough. Merely calling this a "revolution" is under-selling it, isn't it? What's more EXTREME than a revolution? I'll let you know when I think up some words that are EXTREME enough to describe it.

 

[Dufus]

While all these new instructions are very nice, it's probably going to be years before they are applied to mainstream software. Look what happened when SSE came out, took about 5 years.

Anyone know how much software utilizes AVX which has been out since the release of SNB? Hell, even Windows didn't support it until W7 SP1 and there will still be a lot of older OS's and CPU's that will not support AVX for quite a while. So IMHO there seems little incentive for software houses to produce software that incorporates the newer instructions until it is mainstream, except maybe in the case of a few niche programs.

IOW while the technology is exciting, early adoption just doesn't seem that appealing.

 

[AtenRa]

Your dreaming . THE APU isn't going to win . The IGPU is . AMDs APU name will be disgarded. As Intels cpus will be better at everthing including graphics as iGPU name also disappears. Intels cpu will be a cpu that does it all . NV coined the GPU name what was it that ATI coined theirs. Its been so long I forgotten those 3 small letters from ATI . APU is bound for the same result. Because INTEL will never use APU.

No matter how fast the CPUs (Intel or not) will be, the combination of a CPU + Discrete GPU will always be faster in applications and gaming.

 

[Fjodor2001]

Haswell detailed & release dates:

http://wccftech.com/intel-haswell-d...-launching-q2-2013-core-gpu-details-revealed/

And some pricing info:

http://www.fudzilla.com/home/item/28523-haswell-starts-from-$184

Interesting that they will produce a mobile 2C GT3 chip @ 15W TDP. I thought the die area and hence TDP for such a chip would be quite large, considering the GT3?

Also mainstream desktop 4C GT2 @ 65W TDP (I assume they correspond to IB 3570/3770K?). Somehow they have been able to lower the TDP from 77W->65W compare to IB without a node shrink?

 

[TuxDave]

Also mainstream desktop 4C GT2 @ 65W TDP (I assume they correspond to IB 3570/3770K?). Somehow they have been able to lower the TDP from 77W->65W compare to IB without a node shrink?

Better design team.

I kid I kid... don't fire me. XD