Intel Demonstrates Haswell Processors at IDF
Intel formally demonstrated the first Haswell processors at IDF.
Haswell, which will follow Ivy Bridge as 22 nm processors with a new, fourth-generation micro-architecture, is almost entirely focused on power consumption and new computer form factors this will enable.
Intel said that Haswell will improve platform idle-power over the second-generation Core processor, Sandy Bridge, by more than 20x. The company also promised a portfolio of even lower-power processors based on Haswell for 2013. Low-power Haswell CPUs will initially target a power envelope of 10 watts.
In a product demonstration, executive vice president David Perlmutter showed that Haswell processors will deliver about two times the performance of Ivy Bridge processors at the same power, while consuming less than half the power of Ivy Bridge processors at the same performance level. Needless to say, Haswell will be Intel's power play for Ultrabooks and thin form factor PCs.
"The 4th generation Intel Core processor family and our new line of low-power processors will usher in an era of unprecedented innovation in mobile computing," Perlmutter said. "Our focus to deliver even lower power with the great performance that our processors are known for is as fundamentally significant as when we shifted our development focus beyond sheer processor speed in 2001. As a result, you'll see our customers delivering sleek and cool convertible designs, as well as radical breakthrough experiences across a growing spectrum of mobile devices."
In its original laid out and self-imposed tick-tock cadence from 2005, Haswell would be due for introduction late this year. This may be rather unlikely however and we should rather expect the processors to arrive in H1 2013, as a previously leaked slide indicates.
So about the same overall performance can be expected...
So about the same overall performance can be expected...
fagathlon.
I'm not seeing much info or claims about large performance gains like we did with Sandy. Aside from a big leap in efficiency and a good boost to the integrated GPU, is this all Haswell will end-up being about? Maybe we can lose the TIM and go back to flux-less solder for the heat spreader as well?!?!
for graphics it is a 2x increase for the same TDP, so the same performance at 1/2TDP of Ivy Bridge, which is highly impressive. But for the CPU performance we are not expecting a huge increase in performance, my shot in the dark guess would be somewhere in the 5-10% range, with some aspects not changing at all, while other aspects get much larger increases.
But think about it. PC sales are down, because for the bulk of users out there we have enough CPU power for that platform. Most people I know are more than happy with the performance of their C2Duo CPUs which are 5+ years old now (provided that they are paired with SSD and dedicated GPU solution). The demand in the market are for better battery life and performance in portable devices, smaller form factors for dedicated use machines (smart tables to vending machines), and lower TDP for fan-less operation in traditional devices.
Now I want the next gen of super computing just as much as anyone, but as a company they are facing a serious threat from ARM, which Haswell and Broadwell will probably remove if things go as well as Intel is hoping. Once the race to the bottom hits a major barier then we will see a new move on the ultra uncompromised performance end of things again. Development has always gone this way; Efficient single core development, to mass parallelism, then getting that more efficient, and then even more parallelism. Right now we are on an efficiency run for the next 2-3 chips (with modest performance increases as well), but in 2-3 years we will see a performance push again.
You got that wrong. Please creafully read the quote again.
So what actually says is that it will have twice the performance with the same current draw as the Ivy Bridge, OR it will have half the current draw with the performance as the Ivy Bridge.
Oh, and TDP = Thermal Design Power, it has absolutely nothing to do with power consumption. It describes how power the cooling system is required to diisipate at maximum current draw.
I think that here intel talks about performance a little like AMD does: I guess that when they talk about 2 times the performance they are referring to GPU bounded apps...
*ZERO appreciable performance improvement
*but hey, it uses less electricity
in other words... another 30 money-milking SKUs that have no value to desktop users
Monopolies suck
Too bad the rest of Windows8 sucks.
I plan to finally go it-3570K before Christmas...
Win7 runs very good on 4 or even 6 year old tech. More speed is always welcomed - most people won't notice it... gaming is more for consoles. So less heat and wattage = smaller form-factors (tablets I guess) is where the market is going.
Great that CPU's will use less power, but what about performance gains? The wild rollercoaster rush of performance I saw in my childhood has gone. I feel sad and downcast."
No kidding! I remember every new processor arc. being leaps and bounds faster than the last. Just a few examples; 286 to 386 to 486 to pentium. 68000 to 68020/30 to 040 to PowerPC 601/603/604. A 33mhz 030 might put out about 8 mips, but a 33mhz 040 put out almost ~33mips!
Now-a-days, you can expect a 5-10% gain (if were lucky) at the same clock speed when a new CPU is released. The reason for this? Companies don't change CPU designs anymore, they just "tweak" things. IE, a Core 2 is really a heavily modified Pentium Pro from ~1995. Heavily modified, but still very similar. It took Intel until what?... 2009 or so to come out with the Corei series? And even then, Corei seems to be nothing more than a core2 with an onboard memory controller and even more "tweaks". Hence the nearly identical int performance with HT disabled except for memory transfers and FP.
In short, there is no competition anymore. Intel and AMD dominate, and they both see no reason to change anything, just update 20 year old designs. I will give AMD props with Bulldozer for even trying thou. In short, expect things to get even worse in the future for home computing.
Man, I miss the 80's and 90's :-(
Assuming that Haswell does use half the power that Sandy does at the same CPU performance, that means that unless it is crap at stability with increased clock frequencies, K editions will not only have huge headroom, but Intel is very likely to release at least some models with CPU performance improvements over some of the Sandy and Ivy Bridge models.
There is no monopoly going on here anyway. AMD is plenty of a presence against LGA 1155 and with AMD making great improvements on a fast and regular schedule now (assuming that they keep to it, of course), the pressure is getting on Intel. Even without resorting to the core configuration tricks that can let FX-81xx/82xx compete with the Sandy and Ivy Bridge i5s and i7s (even the K editions) in overclocking versus overclocking comparisons, AMD competes very well with stock Intel CPUs in most consumer markets.
For gaming, even the few games that are very CPU bound and not very well threaded can be handled just fine on a Phenom II or FX CPU with the CPU frequency and the CPU/NB frequency (often overlooked because Intel doesn't have a comparable setting, this controls AMD's L3 cache frequency and can have a substantial impact on AMD's per core and highly threaded performance), so unless games and other software come out that the current K editions struggle with, Intel's competition problem isn't even AMD not beating the snot out of them, it's that not only is AMD good enough most of the time, but most games and even regular software simply doesn't need more CPU performance than we have available. If Intel wants people to get back into fast CPU upgrade cycles, they not only have to provide CPU performance improvements often enough, but they have to find a way to get incentive for this out.
Intel could try to get with game and software developers to get some things done such as new features and such that can take advantage of CPU performance more than most current software does. Getting things more well-threaded, increasing stuff such as AVX support and widening it or making similarly high-performance instruction set extensions for other purposes, and more.
I can literally take a five-year-old or three/four-year-old Core 2 Quad or Phenom II x4 and do pretty much anything that most consumers do and do it well, even if I need some overclocking to do it. No game is unplayable with such CPUs and other affordable hardware. Some compression, rendering, general productivity, and much more can all be done in reasonable amounts of time. It'll take more than small or even moderate performance improvements to stop this trend.
TDP and electrical power ARE related, albeit not with a 1:1 relationship. The TDP is set as the expected short-term average peak power the chip is expected to dissipate under normal operating conditions and a chip cannot dissipate more power than the amount of electrical power getting inside it. So the TDP does give a reasonably good idea of what the ceiling of electrical power might be.
when you have lower TDP it means you are converting less electric current into heat which means you are wasting less electric current which leads to less power consumption