Slashdot Mirror
Intel To Debut Limited-Run Ivy Bridge Processor
abhatt writes "Intel is set to debut the most power efficient chip in the world — a limited edition 'Ivy Bridge' processor in the upcoming annual Consumer Electronics Show in Las Vegas. Only a select group of tablet and ultrabook vendors will receive the limited Ivy Bridge chips. From the article: 'Intel did not say how far below 10 watts these special "Y" series Ivy Bridge processors will go, though Intel vice president Kirk Skaugen is expected to talk about the processors at CES.
These Ivy Bridge chips were first mentioned at Intel's annual developer conference last year but it wasn't clear at that time if Intel and its partners would go forward with designs. But it appears that some PC vendors will have select models in the coming months, according to Intel.'"
We need to cut the power and heat of NOCs. Why only build these for the junk market of throw way toys?
Some "Limited Edition", doesn't even come with a free Gordon Moore in battle uniform statue....
Monstar L
Intel really needs to get its act together: It's Atom processors are a decent low power x86 solution, but as usual Intel has delivered them with a crappy 3D graphics to the point the graphical benchmarks can't even run on them, let alone any recent computer games. For the Atom Cedar Trail release they didn't even do DX10 drivers, and sheepishly back-speced it to the now outdated DX9. ARM tablets can deliver decent 3D, so why can't Intel? Even AMD can provide 3D graphics for low-power PCs. Why can't Intel? And Intel wonders why it's becoming irrelevant to the future of computing!?
No DX10 for you!
http://semiaccurate.com/2012/01/03/intel-thinks-cedar-trail-is-a-dog-reading-between-bullet-points/#.UOY58uRJNxA
Windows must live with DX9. Linux can't do anything at all...
http://tanguy.ortolo.eu/blog/article56/beware-newest-intel-atom
Oh and did I mention it doesn't work on Windows 8.
http://communities.intel.com/message/175674
http://www.eightforums.com/hardware-drivers/12305-intel-gma-3600-3650-windows-8-driver.html
http://answers.microsoft.com/en-us/windows/forum/windows_8-hardware/windows-8-on-intel-atom-d2700dc-graphics-driver/2a6015d3-af92-453d-b0c2-20cc56b764de
Once they can make the things in sufficient quantity they will undoubtedly make versions with server features. Most server buyers don't need or want on chip graphics, but do want ECC.
By a 'fuse' he's talking about the selective factory or post factory programming of a chip.
Intel has only 5 different pieces of silicon serving 150+ different Sandybridge and Sandybridge-E processors, the same is true for Ivybridge.
When the fabrication process is finished, the chips on each wafer are tested for quality. Chips that fail completely are discarded. Chips that have flaws in a core or cache segment will have that core or cache segment disabled disabled. This allows a faulty chip to be sold as a lower end model.
Similarly, if demand for a lower end model is higher than the supply of the lower end models, higher quality chips can have parts disabled so that they may be repackaged as a lower end product for marketing purposes.
All of this is done at the factory before the chip is inserted into a processor package. An additional step invented by IBM allows for firmware upgrades themselves to reprogram the chip, possibly reactivating parts that have been deactivated at the factory, or changing CPU parameters so that older firmware revisions cannot be installed (this is done with the PS3)
There are "crippled" products on the market, sold by others as well as Intel. Sometimes it's so they can build one part in their fab, cripple the mainstream part with a fuse, and then charge a premium for the un-crippled part.
Sometimes that actually turns our cheaper to make it this way. They could design a low end CPU to sell cheaper than their premium product, but the cost of producing an entirely different fabrication line for that CPU might actually more than just including a switch in their higher end processor to cripple the chip. The costs include having to reconfigure the production line to a different line of wafers.
Using the same chip design means that they can still sell the CPUs that fail quality control testing. If one of the cores fails in a quad core CPU, they can just turn that one off and sell it as a dual core part. So instead of increasing the price of the premium chip by having the "fuse" as you put it, they are making the chip cheaper because it reduces the wastage if having to discard the failed processors.
Intel is NOT crippling Ivy Bridge processors. Rather what happens is that minor variations silicon wafer mean that different chips come out with different characteristics. It doesn't take much to change things either, we are talking thins with features just 22nm wide, little things have large effects.
When you get a wafer of chips, you have to test and bin them. Some just flat out won't work. There'll have been some kind of defect on the wafer and it screws the chip over. You toss those. Some will work, but not in the range you want, again those get tossed. Some will work but not completely, parts will be damaged. For processors you usually have to toss them, GPUs often will disable the affected areas and bin it as a lower end part.
Of the chips that do work, they'll have different characteristics in terms of what clock speed they can handle before having issues and what their resistance is, and thus their power usage.
What's happening here is Intel is taking the best of the best resistance wise and binning them for a new line. They discovered that some IB chips are much lower power usage than they though (if properly frequency limited) and thus are selling a special line for power critical applications.
They can't just "make all the chips better" or something. This is normal manufacturing variation and as a practical matter Intel has some of the best fab processes out there and thus best yields.
CPU speeds are sometimes an artificial limit (though often not, because not only must a chip be capable of a given speed, it has to do it at it's TDP spec) but power usage is not. It uses what it uses.
Intel has always been about Value Add... There are "crippled" products on the market, sold by others as well as Intel. Sometimes it's so they can build one part in their fab, cripple the mainstream part with a fuse, and then charge a premium for the un-crippled part. Sometimes it's so a crippled system can be sold, and then for an upgrade fee, be "enhanced" in the field. But in any case, it's all about revenue. The annoying thing about this is that they've gone to extra expense and effort to produce the crippled part - the premium part would actually cost less without the extra crippling capability.
While you're correct that Intel relies heavily on testing chips, disabling whatever doesn't work/lowering the clock speed until it works, and selling it as a cheaper product, that's really a cost-saving measure, not a revenue-boosting one. When a chip rolls out with half the cores broken, they'd much rather sell it as a cheap processor than throw it away. AMD does the same thing - even more, actually. As does Nvidia, and pretty much any company that produces enough chips.
These chips are likely a high bin, not a low bin. They're testing for stability at low clock speeds and low voltages, and these are probably the best they have. As you said, there's also a chance that the low power parts may be a deep sort out of the distribution, and there aren't many.
As a different perspective, Intel has also evolved into a performance-oriented company. I don't think that as a company they're very comfortable with this whole "power thing", and I think a limited production like this is probably the way to sell it to management and marketing.
That was the Intel of the early- to mid-2000s. Surprisingly enough, when they got their ass kicked in the Athlon XP/64 vs. Pentium IV wars, they learned from it. They obsess over instructions-per-clock now, not clock speed (old Intel) or core count (current AMD). They've actually been designing their microarchitectures for the laptop, not desktop, for years now, ever since Core 2. They build a 25W laptop chip, then scale it up to meet the 50W-130W desktop and server market (and scale it down to 10W for the even more power-conscious computers). And they've been maintaining a separate microarchitecture, Atom, for the sub-10W range for years now. It's worked pretty well for them.
I wish I could say the same for AMD - they bet pretty heavily on high core counts, disregarding power consumption, and they seem to be faltering on the desktop and laptop because of it. Servers seem to be holding up better, since 130W processors aren't unusual and server applications scale better to more cores, but for the consumer market, the *only* thing they have going for them right now is Fusion, having a powerful GPU on the same die as a half-decent CPU.
The top end non fusion CPU genreally comes between the i5 and close to much more expensive i7, sometimes beating out the i7 in multithreaded benchmarks. It's over 75% of the speed of the i5 single threaded now.
No, no it doesn't. It gets beaten by the i3 3220 for everything except for very multithreaded tasks, where it roughly draws with it:
http://www.anandtech.com/bench/Product/675?vs=677
Piledriver is not much better - it's the same architecture as trinity, but with the GPU stripped off.
There actually really aren't that many tasks where multithreading makes up the difference, as you can see from a comparison of a top end piledriver, against a cheaper i5 that consumes about half the power:
http://www.anandtech.com/bench/Product/697?vs=701
As you can see, even in a lot of the parallel benchmarks the i5 wins, and that's ignoring the i5's hardware video encoder, which utterly demolishes software mode. Overall, the i5 is a significantly better chip, at significantly lower power, and cheaper to boot.
No, I claimed that the A10-5800k *is* a piledriver. Which it is.
You made three assertions.
1. That fusion was between an i5 and i7 in terms of speed
2. That piledriver was faster still.
3. That you were talking about the FX line, not the fusion line all along.
All are false.
1. Is false because the fastest fusion chip (the A10-5800K) is only roughly as fast as the i3-3220, no where near as fast as an i5 or i7.
2. Is false because the A10 *is* a piledriver chip. There are faster piledriver chips out there, I compared the fastest available piledriver chip to intel's slightly cheaper i5. The i5 is faster than the FX-8350 in 23 of the 37 tests, some by a very significant margin. Also notably, the other tests include things like h264 encoding, which the i5 has a hardware unit for, which was not tested.
3. Is false as you can observe from the above quote.