Forget Intel Ivy Bridge, Haswell on the way

Hm. Improvements to QuickSync.

June-August seems later than I expected though. Not that I'm dying to upgrade from Sandy Bridge, haha.

 

Haswell to be 10+ percent faster than Ivy Bridge

Haswell comes with up to 3X faster 3D

 

Meh, more lofty claims from Intel, as usual, which very rarely if ever deliver on their promises. It's getting kind of old actually. That's not directed at you Jayayess at all but Intel!

 

3x faster graphics? Sounds nice but in the end benchmarks are all that matter.

Like talin said, not directed at Jaya who does a great job of delivering these news bits.

 

10% faster.... i was expecting a little more, but i guess without any competion from AMD, they are going to milk us 10% by 10% from gen to gen.

I might even consider Ivy Bridge now, thinking that Haswell is like a year away, for 10%...

 

At least 10%, not 10%.

 

So the mainstream about Haswell is all about the iGPU improvement. That's disappointing for laptop gamers and java users.

 

but i would think gamers and java users are the minority. most people dont care what graphics they have as long as they got graphics. I remember frys advertising this sony laptop with a gt 540m as a full fledge gaming laptop ready for bf3. With haswell (hopefully) they will be able to game

 

Is there supposed to be a big power consumption reduction with Haswell?

 

RumorTT: Intel's Haswell package options for desktop and mobile :: TweakTown USA Edition

Package options of desktop and mobile Haswell processors

 

GT3 with BGA only? That's pretty lame.

 

indeed! I wonder how many threads of non upgradeable we are going to see in next year

 

Well... taking into account that GhZ wise they already more/less reached limitations of what laptops and desktops can handle using silicon... its very likely that until that material is done away with (we know they had viable alternatives for some time now, but won't use them just yet because they can milk consumers for all they are worth), they will keep improving the architecture itself that will probably yield 10%, maybe 15% better clock/clock ratio on a reduced manuf. process - and more improvements being made on the IGP.

 

than IBM is doing something terribly wrong with their clocks speeds and different pipeline schemes

 

Says who? Clock speed has been increasing steadily ever since the Core architecture back in 2006. Every die shrink gives you more headroom to clock higher.

 

I don`t think we reached a limit with the CPU. But AMD and Intel is fighting a lame GPU war instead of focusing on CPU performance. So for people who give a rats behind about IGP, they will basicly get a marginally faster CPU with a hotter CPU due to the increased 3D performance.

 

The biggest bottleneck right now with the average notebook computer is the GPU.

A Core 2 Duo notebook runs Facebook, Windows Media Player, browses the web, does Microsoft Word well enough for 95% of people that use a computer today, granted it has enough RAM. What they would notice though is bad GPU performance. While the Intel GMA 3000 and 4000 HDs have vastly improved, there is still room for help in that department.

The two most important things Intel should be focusing on right now is GPU performance and battery life.

 

While the journey has been slow (because it can be...) Intel has been focussing on exactly those things: iGPU performance and longer battery life while still giving a more powerful system (from their last gen offerings) to boot.

AMD is who should be focussing on cpu performance and increasing battery life (that will make Intel make TOCK and walk like there's no tomorrow - like it did in 2006...).

 

no, the biggest bottleneck is the HDD. None of what you listed is GPU intensive as well, though java/flash over web browsing could get somewhat GPU demanding but not really.

 

indeed there was an article on ars, about how SSD works, here is the table

The table below lays out the memory hierarchy:

LEVEL ACCESS TIME TYPICAL SIZE
Registers "instantaneous" under 1KB
Level 1 Cache 1-3 ns 64KB per core
Level 2 Cache 3-10 ns 256KB per core
Level 3 Cache 10-20 ns 2-20 MB per chip
Main Memory 30-60 ns 4-32 GB per system
Hard Disk 3,000,000-10,000,000 ns over 1TB

basically, its a mess if you dont use SSDs, the slowest part is the HDD. However the gpu is also another culprit, its not one or the other, but a combination. I do prefer the trend that we are seeing now, fast storage (once you go SSD you never go back) and faster iGPUs

 

Did it now?
The Mobile Pentium 4 went as high up as 3.5GhZ.
The X9100 E0 rev. allowed for OC-ing to go as high as 4GhZ (with acceptable temperatures -whereas the C0 rev. was overheating unless undervolted and underclocked).
Also ... with a smaller manuf. process in newer cpu's, the temperatures rise because of a transistor number being increased (silicon inadequacies) and therefore, you are limited in how high you can go clocks wise.
Also i series of Intel offerings in the consumer segment didn't really change that much clock-wise.
The first quad i7 was clocked low (1.7GhZ per core if I'm not mistaken), and then subsequent releases all stayed on the lower end of 2.0GhZ margin with IB entry level increasing to 2.3GhZ base clock...
Extreme versions not really going over 3.8GhZ (if those speeds can be sustained in the first place with adequate cooling).

As for 'clocking higher'... those are puny increases at least in regards to clock speed over a relatively large period of years.
The biggest changes took place on the architecture level increasing clock per clock performance... but cpu's didn't really exceed 3.8 to 4GhZ (which I suspect to be their operational limit using current materials), and also, most of those speeds could not be sustained for very long before the CPU's downclocked a bit to prevent too high temperatures.

Oh and... Haswell it would appear won't really feature anything big apart from IGP modifications.
The architecture differences will probably yield 10% increase in clock per clock ratio... and as for base clocks... that remains to be seen, but I doubt that Intel will go higher than 2.5GhZ for entry level quads.
Oh and, the higher they go in base clocks, the less headroom they will have to turbo-boost - probably won't exceed 4GhZ either way (clocks might remain similar to IB across the board).

 

Are we back to P4 days? where clocks matter? Or that we pretend that those mattered?

In those days, intel looked at IBM, where they had a differently structured pipeline, where the high clock was essential (5GHZ), due to the longer pipeline, so that each processing bit could make the run and be done quickly.

Adding low level structures and putting hoops in the pipeline makes things even faster than those 5ghz.

But basically the GHZ is a useless metric unless you know where it needs to be applied to.

 

Mr MM, of course clocks matter!

Otherwise, why would I (or anyone else) be buying 'K' series CPU's so that we can overclock them by 1GB or more?

(Hint: greatly increased performance!!!).

 

Yes.

And Pentium 4 came out before Core (and had a massive TDP to boot). One of the premises of my statement was "after Core". P4 was designed to clock high whereas Core was designed to be more efficient per clock so that's a apples to oranges comparison anyway.

A stock X9100 ran at 3.06 ghz whereas the highest next gen dual (i7-640m) had turbo up to 3.2ghz for 2 cores and 3.4 for 1 core. Can't really make the comparison with overclocking since the i7-640m was not unlocked.

Uh, did you miss Sandy Bridge->Ivy Bridge? IVB is a die shink (CPU-wise) with a smaller node and has higher clocks than SB, so clearly there is something wrong with your statement. As history shows, nearly every die shrink has yielded higher clocked parts, since the Pentium Pro -> Pentium III days. (And I say "nearly" because I'm sure if you look hard enough you can find an exception, not because I can actually think of.)

They have. Off the top of my head most people could get the QX6600 to low/mid 3ghz. A Nahalem quad usually went to the mid/high 3ghz and Sandy i5-2500k could get low/mid 4ghz pretty easily. IVB doesn't clock as high, but that's really due to the crappy TIM that intel used between the die and the IHS.

It will have new instruction sets on the CPU side like AVX2 and TSX.
"Big" is a matter of opinion, but there will be vast improvements on certain workloads and none for other workloads.

IVB clocks will probably increase another .2ghz across the board for the half year refresh that Intel does (like the i7-2630qm->i7-2670qm). Given Intel's history, there's no reason to believe why they won't go past 2.5ghz base clock for the entry level Haswell quad. If they don't do it initially, they will almost certainly do it for the Haswell half-year refresh.

Anyway, my point is that there's no ghz limit that has been reached with silicon. The clock speed will be limited by the uarch and how good the foundry/process is way more than by just using silicon. If Intel wanted to build a 20ghz 22nm Pentium 4 they probably could, but they won't because it would perform like garbage and be horrendously inefficient. You may see 100ghz/thz/whatever graphene circuits or something on the news but those are not nearly as complex as a proper x86 CPU.

Maybe stuff with more desirable electrical properties like graphene will be used in the future, but right now the technology simply not mature enough. Intel spends billions just researching the next process node. How much do you think making new litho equipment and fabs for an entirely new material will cost? Not to mention training your engineers to work with it, and design CPUs for it's idiosyncrasies, etc., etc. It's not a matter shafting the consumer, it's a matter of the technology and commercial viability simply not being there yet.

 

after this statement of yours now I think you have no idea what you're talking about, but keep pretending that you do.

 

Thank you for contributing to the discussion.

 

Hi
Guys i was reacting on the title of thread and here is mine title then;
Forget Intel Haswell , Broadwell on the way..
OR lets just forget on even broadwell and lets just wait on next one or
you know what Jayayess1190 ? lets just wait fooor eveeerrr that is the best n?

exactly!!

 

I have that thread made already

The point of these threads is just to discuss future platforms. I have made an AMD Fusion one too. This is what I like to do, discuss and inform people on future cpu generations (as well as plan for my next purchase).

 

jayayess, you've been planning on the next purchase since I joined NBR in 2009! lol

 

Of course the clocks matter, they are the afterall the speed that each cycle of processing that occurs, however not in the way that it mattered for P4 cpus, where they were based on the pipelines of the IBM cpus of the time

 

DDR3 SDRAM - Wikipedia, the free encyclopedia

not sure which column you use but RAM is 3-10ns or 10-15ns depending on which you use. SSDs are 100ns(.1ms) and hardrives are 3,000-15,000ns (3-15ms) Depending on drives and what type of seeks your doing.

People have gotten the 920XM to 4GHz but trying to cool that in a laptop is nearly impossible. it uses like 150-180 watts from what i have heard. Also SB gets to 4.5-5+GHz IB has gotten to 5GHz. Go read reviews on newegg....i was skimming around for fun like 30 mins ago before i read this...only reason why i know those numbers off my head right now

EDIT: side note that DDR4 will help IGPs a lot....any info if IGPs will be better...they are still a little short from what i would have hoped. Especially in the ULV area :/

 

Im currently in london, when I get to brazil I see what values and how old that article is, but it was a basic copy and paste from their table.

1 ns = 10 elevated to -9 seconds
1 ms = 10 elevated to -3 seconds

 

Intel Haswell Desktop and Mobile Processors Launching in Q2 2013, Core and GPU Details Revealed

 

damn, intel is really going to turtle with the dual cores this time around

 

why do they turtle with XM chips so much? I hate how they release over 6-9 months so dumb!!!! hell it has been almost 9 months since release and no new 3930K! And what is up with Xeon chips being locked? Might as well buy 3930K for 1/4 the price and over clock it to compensate for the lack of 2 cores

 

Intel 4th gen Core chips to slash power consumption

Intel's power-efficient Haswell processor targets thinner laptops with new 10-watt TDP

 

I really don't care what is coming next until programs or games come out that I'm no longer to use efficiently on my current setup.

 

So... ??? lol...


All that means is that you'll be 'lost' in the new jargon when/if you have to upgrade (because you won't know which end is up).

 

I may get lost...hell...I'm already lost.

 

AnandTech - Haswell at IDF 2012: 10W is the New 17W

 

10W SOC will be crazy good if it has at least GT2 graphics plus performance at around core i5 ULV. That will be around half of IVB ULV TDP(considering SB consumed around 3W). we should see super thin convertibles and ultrabooks with 15+ hours of battery life. I hope broadwell reduces it even further.

 

So this is pretty sick, but how soon will we get to the point where the CPU power savings, for laptops, at least, are pretty much insignificant? I always seem to hear that it's the screens that draw a lot of the power.

 

when screens switch to AMOLED that is when you'll see the difference...expect i dont know if that is possible. There may be not enough material to make everything that. I read for LEDs along that the manufacturing was going to reach the max possible production around this year unless they build more plants. So a lot of factors have to be considered. Some can be over come others can't

EDIT: the big thing is that you'll be getting powerful devices in the same run time in a smaller form factor. not more run time. I care more about power over run time. This is great news

 

AMOLED doesn't even consume less power. It uses less power to display blacks, but more for whites so it's a wash. Plus there's burn in and all the other issues with it. The real power savings will come from IGZO displays which are due next year.

 

indeed the screen will always be the more consuming thing, unless under load

 

hmm what about speakers? I get more of a battery drain when blasting speakers vs full brightness

 

Haswell's GT3 Will Work 40% Slower Than HD4000 - Softpedia

 

This article makes GT3 sound like it will be slower than HD4000 on the whole when it is actually comparing the clock speed of a SINGLE execution unit on each chip.

 

yea i was worried for a min there lol. I would be happy with a hotter CPU with better GPU performance to be honest.

 

Gotta have something for Broadwell, am I right? Anyways, I wouldn't be surprised if the truth of it is that Haswell's iGPU will range from 600/800MHz to 1100MHz depending on the form factor.

I'll love when we get to the point where an integrated GPU is good enough for production apps in the way CPUs kind of are now.

Really interested to see how much more of the low-to-mid-range gaming market Intel will take with Haswell, too.