Slashdot Mirror
Intel's Core M Performance Is Erratic Between Devices
An anonymous reader writes: AnandTech noticed some odd performance disparities with Intel's Core M CPU, a chip designed to bring high-powered processing to thin, fan-less devices. After investigating, they found that how OEMs build their laptops and tablets has a far greater effect on Core M performance than it does for other chips. "When an OEM designs a device for Core M, or any SoC for that matter, they have to consider construction and industrial design as well as overriding performance. ... This, broadly speaking, gives the OEM control over several components that are out of the hands of the processor designers. Screen size, thickness, industrial design, and skin temperature all have their limits, and adjusting those knobs opens the door to slower or faster Core M units, depending on what the company decides to target.
In the Core M units that we have tested at AnandTech so far this year, we have seen a variety of implementations with and without fans and in a variety of form factors. But the critical point of all of this comes down to how the OEM defines the SoC/skin temperature limitations of the device, and this ends up being why the low-end Core M-5Y10 can beat the high-end Core M-5Y71, and is a poignant part of our tests. Simply put, if the system with 5Y10 has a higher SoC/skin temperature, it can stay in its turbo mode for longer and can end up outperforming a 5Y71, leading to some of the unusual results we've seen so far."
In the Core M units that we have tested at AnandTech so far this year, we have seen a variety of implementations with and without fans and in a variety of form factors. But the critical point of all of this comes down to how the OEM defines the SoC/skin temperature limitations of the device, and this ends up being why the low-end Core M-5Y10 can beat the high-end Core M-5Y71, and is a poignant part of our tests. Simply put, if the system with 5Y10 has a higher SoC/skin temperature, it can stay in its turbo mode for longer and can end up outperforming a 5Y71, leading to some of the unusual results we've seen so far."
It's something some of us Apple fans have long figured out is that individual specs sometimes are completely meaningless.
Having a Core i7 will not actually feel more responsive in everyday tasks compared to a Core M if the i7 is paired with a spinning rust disk and the Core M has a PCI E SSD.
Similarly, just looking at the chip in the machine might not tell us everything if we don't know anything about how it's handling cooling or what specific design choices were made.
We're on the verge of reaching the 150HP car of computing. Don't really need much more for most tasks unless you're doing heavy lifting or looking to have fun, and even a lot of good clean fun can be had at 150HP.
Non impediti ratione cogitationus.
This has been the case for Intel CPUs for many years. Back in the Core 2 days they were already letting laptop manufacturers customize the power profiles (and therefore performance) of their mobile CPUs to suit the thermal load handling ability of their machines.
If you have one of those old Core 2 machines and don't install the Intel chipset driver for it (or install the generic one) it will get hot and loud pretty quickly. The only real difference now is that the CPU has better management built in and works okay without the driver giving it hints.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
I can't see any problem with that.
Just read a few reviews before buying to make sure you get a device that is properly designed.
If you can't take the heat, get out of the laptop.
Some settling may occur during posting.
Intel model numbering has often been a bit cryptic, and worse more recently as they've spawned new product lines and taken advantage of their lead over AMD by market-segmenting with incredible precision, producing parts that differ by a single feature enabled or disabled, or have the same clock speed but different 'turbo' speeds, or any number of similar permutations.
As though that isn't enough fun, now even expert level knowledge of the model numbers won't tell you how fast it is because the OEM can gimp it to suit their chassis design. It's a good thing that basically all modern CPUs are really fast, or this would be downright depressing.
I have noticed some wide disparities of performance in reviews of the new Macbook. One review I read put battery life at 7 hours, another at 11 hours of continuous video playback. Since there were no specifics, I don't know what to attribute these differences to. Who knew that the growth of mobile devices would direct manufacturers to focus on the design of the device? Intel's concern is now heat as much as power efficiency or performance. This point aside, if I were going to put down money for a new laptop, and using Mac laptops as an example, why would I plunk down $1300 on a Macbook when I can put down $800-$900 on a Macbook Air, and buy better performance and battery life? Sure, the Macbook is thin and light, but isn't the Air already pretty damn thin and light? This question applies to other PC makers. Are other PC makers watching Apple to see if there is that strong a market demand for laptops that have better screen resolution, weigh less, and have slicker design but poorer specs overall? Do these questions make any sense?
There, I fixed the headline for you.
Once notebook computers became a fashion accessory that happens to compute, this result was inevitable.
I suspect the Macbook makes better use of the Aluminum shell as a heat sink of sorts. One review said the laptop would become warm (but not hot) under load - and was not throttling..... There were issues with power utilities for Lenovo early on with regards to this line - something Apple has worked on the last few major releases of OS X - which gives me some confidence Apple is ahead of the curve there. The first few Macbook benchmarks seem to match what I would have expected (above others) which seems to bear this out. BTW, Are you sure you are not remembering PDP.... since that was what we did with the old PDP-?? (11 I think) was if it did not boot - lift and drop then boot.
Man, I got to use a PDP11 at my father's institute. I can assure you that lifting the machine would require industrial machinery. Also, the machine was rack-mounted with its disk... So that advice would not be very welcome :)
HDD is many orders of magnitude slower than SSD and quite a high latency, SSD is orders of magnitude slower than Memory, Memory is orders of magnitude slower than CPU cache etc. Getting a much faster CPU (GHz wise) does not increase the performance of the machine as much as many seem to be brainwashed into thinking. If you just upgrade a CPU with one 30% faster you will only get a fraction of that overall. It is all about having sufficient resources when you need them for the task at hand.
Performance for user computers (as opposed to servers) is very much about perception. If your application opens snappily to begin with, the user will feel the computer is faster. If there is sufficient CPU power when you need it, it will have the same effect. It is all about balancing and making sure that your slower components are not needed as much as your faster components. Most users have CPUs sitting 85% idle most of the time..... getting a faster CPU only will increase the amount of idle time and not give the user a better experience.
The greatest performance boost in recent times is the advent and rapid improvement of the SSD. The stock CPU in many computers is increasing in performance in the single digits...... The majority of applications don't tax even the Core-M CPU (for the majority of users).
Of the two systems that performed best with the bottom of the line Core M, one used a cooling fan while the other had an aluminum shell that acted as a heat sink. The machine with faster processor had a plastic shell with no active cooling. It throttled back when it got warm and would not run at full speed due to heat buildup. Hence the reason why a lower power CPU outperformed a higher powered one. Shell design is everything when it comes to quickly venting heat. Don't use enclosed plastic if you need to cool a CPU that is designed to run at 65C. Use a metal shell, or an active cooling system if the shell is plastic.
I just lived this - I started with a Dell Venue Pro 11 i5, TDP 11.5W in plastic case. It was a throttling and heat nightmare, slow, sluggish, very buggy graphics drivers. Next up was the Yoga 3 Pro in the review. Saw the same performance oddities in the review plus buggy graphics (frequent crashes with video, flash, very poor windowed video performance) and poor battery life despite removing all bloatware. Moved to a Asus Zenbook UX305 in the review, Signature Model sold by Microsoft. Felt much faster in use than Yoga 3. In my usage, twice the battery of life or more of Yoga 3. Very light, sturdy and fast, great value. But also buggy graphic problems which seems to be Intel's Core M variant of the 5500 drivers. That plus I just missed backlight on keyboard, returned (but still a worthy model). HP Spectre x360. This unit is a bit heavier but in my view is the winner. Fast, excellent battery life, much better graphics driver, nicest keyboard in my experience. Due to a shipping error got the QuadHD screen, but surprised that unlike Yoga 3 Pro, battery life does not take much of hit and performance is much better. Really like it and it's a keeper. It feels fast enough if you want to do some Photoshop or medium editing but light and sturdy enough that it's close to Macbook Air functionality (3.26 lbs vs. 2.96). Nice range of models from non-Touch. I
No it was the Apple III. It had a few issues because the design was really pushing board design.
PDP?
DEC Would never allow that. Frankly picking up and dropping most PDP 11s could cause death or a minor earthquake.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
Have fun with your slow ass MacBook after I warned you that their thermal solutions suck, ten thousand times.
[...] early Apple III computers where heat would cause chips to expand out of their sockets, [...]
“It’s not wise to upset an Apple III.”
“But sir...no one worries about upsetting a Droid.”
“That’s ’cause a Droid don’t cause people’s chips to expand out of their sockets. Apple IIIs have been known to do that.”
“I suggest a new strategy, Artoo. Let the Apple III win.”
temperature difference / thermal resistance = power dissipation
5K/W heatsink at 20K above ambient dissipates 4W. Same 5K/W at 40K above ambient dissipates 8W.
I had a Commodore PET that did that.
Solution, lift the lid (it pivoted) and push down on all the chips. Lower the lid and power it up.
I should use this sig to advertise my book ISBN-13 : 978-1501515132.
So... why exactly do you need a PCIe SSD for watching videos again?
Because any above-average part that Apple includes is what makes their devices superior to the rest of the market, and anything they exclude or go for below-average on is superfluous, of course.
I remember sigs. Oh, a simpler time!
If a buy a computer with a CPU that is rated at X GHz then that CPU had better be able to maintain that frequency, always. Otherwise it's a meaningless number. CPUs can already overclock themselves (Turboboost) above that frequency so if they can also legitimately underclock themselves then the 'rated frequency' is completely meaningless. I don't think that is acceptable. I encourage all slashdot readers to test their new computers under load and if they cannot maintain their rated frequency RETURN THEM! Or better yet, file a formal complaint for false advertising or fraud and then return them.
I blogged about this a while ago and I think the problem has only gotten worse. Lots of consumers are getting a crap experience because of insufficient cooling, manufacturers are selling rigs that can't do what they promise, and software developers waste time dealing with complaints about slow games/etc.
https://randomascii.wordpress....
Amigas with DIP 68ks were a bit fiddly as well, once you'd de-socketed the CPU once or twice. And especially Amigas with accelerators which were plugged into their DIP 68k socket...
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"