Slashdot Mirror
Intel Shows 14nm Broadwell Consuming 30% Less Power Than 22nm Haswell
MojoKid writes "Kirk Skaugen, Senior Vice President and General Manager of the PC Client Group at Intel, while on stage, at IDF this week snuck in some additional information about Broadwell, the 14nm follow up to Haswell that was mentioned during Brian Krzanich's opening day keynote. In a quick demo, Kirk showed a couple of systems running the Cinebench multi-threaded benchmark side-by-side. One of the systems featured a Haswell-Y processor, the other a Broadwell-Y. The benchmark results weren't revealed, but during the Cinebench run, power was being monitored on both systems and it showed the Broadwell-Y rig consuming roughly 30% less power than Haswell-Y and running fully loaded at under 5 watts. Without knowing clocks and performance levels, we can't draw many conclusion from the power numbers shown, but they do hint at Broadwell-Y's relative health, even at this early stage of the game."
Meaningless number unless we know they are comparing at same performance level. You can get another IvyBridge CPU, downclock it, and you'll get 30% less power use.
Arm meanwhile has 8 core processors suitable for Smartphones (and yes they can run all 8 cores simultaneously).
What they need right now is an a chip *now* that is 30% less power THAN AN EQUIVALENT ARM, and more cores and cheaper, oh and it also needs to be SOC available.
Really saying you're next chip is 30% lower power than one you just launched, means the one you just launched is 30% too much power drawn. Which is true, but not something to point out.
Pretty huge.
1) means smaller design, which means you can pack more in for the same power
2) simpler cooling, which means you could fit it in smaller cases
Both of those are very good because you could fit both scenarios in to a production line trivially.
Larger procs go one way, smaller mobile ones the other way.
Hell, I am just surprised they are at 14nm. I never thought they could get down that low because of leakage.
Take a look at this slide, on the right is the system on a chip version of their Broadwell 2 core processor:
http://hothardware.com/image_popup.aspx?image=big_idf-2013-8.jpg&articleid=27335&t=n
See how much of the chip is assigned to crypto functions? It's almost as big as one of the processor cores. All that silicon used for crypto and it's completely wasted because it cannot be trusted because of the NSA. It wouldn't surprise me if some of that silicon is NSA back door functionality because that's one heck of a lot of transistors to assign to crypto functions.
There is a good comparison of ARM vs x86 power efficiency at anandtech.com: http://www.anandtech.com/show/6536/arm-vs-x86-the-real-showdown
"At the end of the day, I'd say that Intel's chances for long term success in the tablet space are pretty good - at least architecturally. Intel still needs a Nexus, iPad or other similarly important design win, but it should have the right technology to get there by 2014."
(...)
"As far as smartphones go, the problem is a lot more complicated. Intel needs a good high-end baseband strategy which, as of late, the Infineon acquisition hasn't been able to produce. (...) As for the rest of the smartphone SoC, Intel is on the right track."
The future for CPUs is going to be focused on power consumption. The new Atom core is two times more powerful at the same power levels than the current Atom core. You can see http://www.anandtech.com/show/7314/intel-baytrail-preview-intel-atom-z3770-tested:
" Looking at our Android results, Intel appears to have delivered on that claim. Whether we’re talking about Cortex A15 in NVIDIA’s Shield or Qualcomm’s Krait 400, Silvermont is quicker. It seems safe to say that Intel will have the fastest CPU performance out of any Android tablet platform once Bay Trail ships later this year.
The power consumption, at least on the CPU side, also looks very good. From our SoC measurements it looks like Bay Trail’s power consumption under heavy CPU load ranges from 1W - 2.5W, putting it on par with other mobile SoCs that we’ve done power measurements on.
On the GPU side, Intel’s HD Graphics does reasonably well in its first showing in an ultra mobile SoC. Bay Trail appears to live in a weird world between the old Intel that didn’t care about graphics and the new Intel that has effectively become a GPU company. Intel’s HD graphics in Bay Trail appear to be similar in performance to the PowerVR SGX 554MP4 in the iPad 4. It’s a huge step forward compared to Clover Trail, but clearly not a leadership play, which is disappointing."
you forgot the part about accomplishing an unknown amount of work on a benchmark with unknown results
The IPC has hit a brick wall. The proportion of time spent on cache misses and branch mispredictions simply is a limit.
After all IBM Power8 will have 8 threads/core (as announced at Hot Chips, but as far as I know, there have been no article about it on Slashdot). I'm not sure 8 is very useful, but apparently on many workloads, the 4 threads/core of Power7/Power7+ gives more throughput then 2 threads. Several threads per core increase aggregate IPC, but not per thread IPC.
The reason I'm doubtful on 8 threads/core is important on Power8 is that there are only 2 LSU (load/store units), which means that each thread can only access memory every 4 cycles on average. For a RISC processor with 32 registers and 2 LSU, 2 threads are an obvious way to keep the execution units busy, 4 threads can certainly increase throughtput, but at 8 I start to my doubts since they have to share the caches despite the 4(!) levels of cache: L1 and L2 are per core, L3 is mostly per core if I understand correctly, and L4 is completely shared among all cores. Fox x86, Intel has not gone (yet) above 2 threads, and they seem to go rather towards large and wide register file with AVX512, introducing a new instruction encoding which is even more baroque (tough job, but it's Intel we are speaking of) than everything they have piled on top of the original 8086, including a new 4 byte long prefix for a start. I'm also a bit doubtful about AVX512: the instruction to save/restore FPU context has been extended and now stores/loads over 2.5kB of data, this is more context than any other processor I know (including Itanic) and will certainly impact context switch and signal delivery latencies. It is also incredibly intertwined with the memory protection extensions (MPX) and adds complexities to supporting MPX.
On x86, last time I looked, there were still only one load and one store unit (somewhat less flexible than 2 general purpose LSU since there are generally more loads than stores) but the big problem was with 32 bit code, which was spilling like mad because of two few available registers. amd64 (whatever claims Intel, they had to follow AMD's on this one, and the NIH syndrome shows) often gives 9 more available registers (in theory 8, but position independent code has to sacrifice a register on 32 bit) has made the memory traffic due to register spills negligible in practice. Intel stayed far too long minimizing importance of 64 bit support
(NIH syndrome?). However, they finally have admitted that 64 bit is mainstream and 32 bit fading away.
Trying to improve the IPC on x86 is a nightmare, because the instruction decoder is insanely complex (and the complexity is growing): they have gone from 3 instruction/clock on the PPro to 4 (Only 33% improvement in 17 years, I'm aware that this is a meaningless figure). I don't remember how instruction decoding is done on Intel processors, but I remember a description of an AMD processor in which they simply distribute the instruction stream to 16 decored, each shifted by one byte, and then
cancel the results of the decoders who were not fed a instruction which starts on an instruction boundary. That's gross, needs a lot of wasted power and transistors for no useful work. Actually some Intel processors have dual instruction caches: one encoded in x86, and one recoded to something easier to digest. Of course, this does not come for free (both in silicon area and in power consumption, coherency logic, etc.).
"led monitors use less power than ccfl at the expense of some colour quality"
What? Not even close. Go look at a spectrograph of a white LED versus any fluorescent. A white LED is only beat by an incandescent lamp as far as a complete light emission range goes.
Here you go.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
"Gone is the time when integrated Intel GPUs were worthless"
Actually, here's something funny about that. You want to know why Intel GMA945/950/X3100 sucked balls?
They were all deliberately crippled by Intel. Their original spec speed was to be 400 MHz. Every desktop and laptop that had these had them at 133/166 MHz speed. Unusable for even Quake 3.
But suddenly - if you fixed that clock speed issue, holy crap, you could play Q3 in OpenGL mode! Suddenly newer games like Faster Than Light run full speed instead of 2 FPS. I'm on a laptop that has an Intel GMA950, and since I installed GMABooster as well as reset the clock speed to the proper 400MHz, this system has suddenly become a LOT more usable, and it's still not getting any hotter than typical.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
Helps a lot. But there are many factors that affect power usage.
Power supplies used to be awful. I've heard of efficiencies as bad as 55%. Power supplies have their own fans because they burn a lot of power. Around 5 years ago, manufacturers started paying attention to this huge waste of power. Started a website, 80plus.org. Today, efficiencies can be as high as 92%, even 95% at the sweet spot.
GPUs can be real power pigs. I've always gone with the low end graphics not just because it's cheap, but to avoid another fan, and save power. The low end cards and integrated graphics use around 20W, which is not bad. I think a high end card can use over 100W.
A CRT is highly variable, using about 50W if displaying an entirely black image at low resolution, going up to 120W to display an all white image at its highest resolution. An older flatscreen, with, I think, fluorescent backlighting, uses about 30W no matter what is being displayed. A newer flatscreen with LEDs takes about 15W.
Hard drives aren't big power hogs. Motors take lots of power compared to electronics, but it doesn't take much to keep a platter spinning at a constant speed. Could be moving the heads takes most of the power.
These days, a typical budget desktop computer system, excluding the monitor, takes about 80W total. Can climb over 100W easy if the computer is under load. So, yes, a savings of 5W or more is significant enough to be noticed, even on a desktop system.
Intellectual Property is a monopolistic, selfish, and defective concept. It is "tyranny over the mind of man"
They had to, because at 50% efficiency, if you wanted a 500W power supply, you're talking about drawing 1000W. And that would be a practical limit because a typical 15A outlet would provide 1650W. Don't forget a switching supply also has horrible PF, so 1000W could mean 1250VA. Sure it's only consuming 1000W, but your circuit breaker has to handle the extra 250VAR (just over 2A), so your computer was drawing up to 12A instantaneously. At 80% PF, this limited consumed power to around 1300W, which at 50% efficiency meant your PC could draw up to 650W. (The thing with imaginary power is it has very real currents, so your electric meter will never register imaginary power, but all the power handling equipment will see its effects - because what really happens is you "give back" that imaginary power so your meter wouldn't record it).
At the same time, PC components were drawing more and more power and 650W wouldn't cut it.
So the situation was to be as-is and force everyone to install multiple circuits in their computer room, or to sharpen up efficiency.
Which they did, because at 95% efficiency and a power factor of 1.0, you can now have a power supply that nearly goes all the way - 1500W (a little less to account for component variations and such), and still be within 15A.