♫ The Core i9 clock cycles go up. Who cares where they come down?
Owners of laptops fitted with Intel's Core i9 high-performance processor, including computers made by Apple and Dell, are finding that the machines slow down compared to the pace of older models. Chipzilla describes the six-core Core i9 as a "no compromises" chip aimed at gaming, VR and "next-level content creation". Intel has …
COMMENTS
-
-
Friday 20th July 2018 12:30 GMT JeffyPoooh
Re: The song the headline refers to...
I came here with this ---v in my copy-and-paste buffer, but I see you've handled it.
Don't say that he's hypocritical
Say rather that he's apolitical
"Once the rockets are up, who cares where they come down?
That's not my department!" say Wernher von Braun
...
-
-
-
-
Friday 20th July 2018 14:40 GMT GrapeBunch
Re: Optional Extra Purchase
So, thinking of the great Canadian Outdoors in January? Temperatures down to -40C not unheard-of? Good luck typing in woollen mittens. Over a decade ago there was a screen (which I associate with the IBM Thinkpad 600E and 600X models) which if used in extreme cold would adopt a pink (or perhaps magenta?) cast. Forever, though I think mine got less pink with time--and centrally-heated indoor use.
-
-
-
-
Friday 20th July 2018 13:28 GMT eldakka
These need a disclaimer that sustained maximum performance is not possible. The idea is sound that the processor can smash through a web page render or other short workload and then throttle back. BUT it's just not suitable for multi-hour rendering or gaming sessions.
That is the theory behind Intel's ultra-low processors like the m5-6Y57. It's a "finish the job as fast as possible and then go back to sleep" paradigm.
Which is fine for general purpose consumer laptops that follow that sort of workload - maximum speed while launching a program, or rendering a webpage, then go back to sleep.
However, these are 'pro' laptops, targeted at professionals doing video editing, rendering, other professional type workloads.
At the very least they should be able to sustain base clocks. But these laptops cannot even do that, they aren't just throttling turboboost, they are throttling their base clock speeds down.
-
-
This post has been deleted by its author
-
-
Friday 20th July 2018 14:23 GMT Anonymous Coward
Fake news.
I haven't posted here in a long time, but I had to, because apparently most of you are willing to believe whatever other says, still, despite the biggest theme in news for the past year or two being that of fake news.
As an owner of the latest Macbook Pro with the i9, I've ran benchmarks on it, I've also ran it against the last gen top end i7 that my colleague has on server and compilation work loads that utilises every core and it's faster in all circumstances, both single threaded and multi-threaded.
Then again, any "pro" using a Macbook Pro would know to put in a fan control app, Apple is well known to prefer silence machines than noisy ones blowing out hot air, a bit like this youtuber and the rabble on reddit.
Who in their right mind would believe in a youtuber, whose job is to blow hot air.
-
-
Friday 20th July 2018 14:50 GMT Steve Davies 3
Re: Dave Lee
Who in their right mind would believe in a youtuber, whose job is to blow hot air.
I wonder if this is the same 'Dave Lee', who is a BBC Tech Correspondent who lives in LA and is a serial subscriber and owns nothing.
https://www.bbc.co.uk/news/technology-44829976
Probably not but the same name got me thinking.
Anyway, time for a Pint of Firebird and to watch the end of todays stage on the TdF.
-
Tuesday 24th July 2018 09:55 GMT Poncey McPonceface
Re: Dave Lee
I wonder if this is the same 'Dave Lee', who is a BBC Tech Correspondent who lives in LA and is a serial subscriber and owns nothing.
https://www.bbc.co.uk/news/technology-44829976
Probably not but the same name got me thinking.
Ashamed to say but I've been on a bit of a YouTube binge of late and can confirm that (unless said chap has had a race transplant) we are *not* talking about the same Dave Lee.
https://www.youtube.com/channel/UCVYamHliCI9rw1tHR1xbkfw
Dude refers to himself as Dave2D and is a prolific tech vlogger who knows what he's talking about and is unbiased unlike yawn inducing Mr. Fanboy Anon above.
-
-
Friday 20th July 2018 15:05 GMT ThomH
Re: Fake news.
Right now it seems to be Youtuber versus Twitterer* as far as data points go. Which person you've never heard of, publishing their results via the internet's various platforms for the attention hungry do you most prefer?
I think Ars at least has pushed back its review of the new Apple machine in order to test the claims and provide something like an evidenced conclusion; I hope other outlets are doing the same.
* to save everyone the reading: he seems to allege the problem is in the use of Adobe Premiere Pro for benchmarking, by posting figures that show it takes almost 2.5 times as long as Final Cut Pro X to perform the same task on the latest MacBook Pro. I think he's suggesting Premiere isn't well-adapted to modern processors. It's Twitter though, so mostly the word-based version of a Rorschach test.
-
-
-
Friday 20th July 2018 15:28 GMT naive
Re: i9 works great in a suitable chassis
This reply is 100% spot on, any laptop being sold as high-end gaming rig, and is not at least one inch high, will suffer from performance reduction due to heat constraints.
People buying thin, but supposedly fast laptops, fell for a trap, one can not beat physics.
MSI understood this when designing their GT-series gaming laptops, they stay cool and fast.
The extra cooling capacity due to the height, also reduces high and noisy fan speeds.
-
-
Friday 20th July 2018 14:58 GMT Halfmad
Laptop coolers won't be enough
No gaming laptop I've seen so far has been able to combine performance, good cooling and relatively low noise from the cooling solution.
Even with an external cooler it's not going to be much better, all this throttling does is expose how unsuitable the laptop chassis is for the CPU it houses. They can fudge a fix if they want, bottom line is that it'll still throttle and you'll still fail to get the full performance out of it for any sustained period because of the fundamental laws of thermodynamics.
-
Friday 20th July 2018 15:51 GMT hellwig
It needs a Blowiematron.
Saw those Canadian loonatics (lol) at LinusTechTips "install" an 11,000 RPM fan into a laptop to get adequate cooling for their Core i9. I think they also de-lidded it and added better thermal paste (remember, Intel's ease of assembly trumps your cooling needs). Anyway, it may run at 120dB+, but pushing several cubic feet of air a second seems to do the trick.
Remember when laptops came with "mobile" processors because manufacturers knew performance CPUs wouldn't operate efficiently in that enclosed, often battery-powered chassis? I guess marketing hype outweighs common sense these days.
Where is the fat cat businessman smoking a cigar and counting his money icon?
-
Friday 20th July 2018 16:49 GMT JLV
unless I am wrong
This CPU is the one that you have to get on a 512GB or up SSD MBP. For... $600 CAD extra to the 256/lower CPU model. That’s mucho $$$$ for 256G of disk, but hey, you get a faster CPU. Oh, wait.
So, if you want a “pro” machine without a thermally crippled CPU, that’s 256GB disk space max, bub!
Least Apple should do is revisit their SKUs and ship reasonably sized disks with CPUs that run normally within their TDP budget.
-
-
-
-
Saturday 21st July 2018 16:24 GMT JLV
Re: unless I am wrong
you’re a right genius, aintcha?
i thought about saying we were getting high prices and, at best, reasonable hardware (if you’re not often @ sustained peak CPU loading in your use that it)
but then I thought charitably that it’s obvious in context.
not to you i guess.
can type slower if it helps...
-
-
Monday 23rd July 2018 08:39 GMT Anonymous Coward
Re: unless I am wrong
@JLV & "I can make do with either of the 2 choices below:- high price/flawless hardware.- reasonable price/reasonable hardware"
This is Intel hardware so high price/broken hardware
Also best not to mention money to AppleFools®, they didn't opt for Apple because they understood what they were buying, the purchase was only to allow them to associate with other AppleFools®.
Personally I think the Wannabes should revive fur coats (wool being too obvious even for them), at least then their social entry ticket will have some practical value to the owner come the winter.
-
-
-
-
Monday 23rd July 2018 17:18 GMT JLV
Re: unless I am wrong
Ah, I stand corrected. I think the story I read @ Ars about this indicated that 256->512GB meant going i7->i9 and I didn't check. I had seen the 3100->3700$ bump before to get to 512, saw that the only thing that changed there was the CPU and assumed that $600 would mean a lot beefier CPU.
I wonder if I should be relieved it's only a greedy Apple move, without being into greedy and incompetent.
Almost $4K to get a laptop with 512GB? Linux on my next work laptop is looking likelier by the day.
IMHO, this is a really worrying incident about how much say the engineers have @ Apple if a flaw as obvious as this makes it to market.
-
-
-
-
Friday 20th July 2018 20:13 GMT J27
Re: Testing Standards...
Dell? For sure not. I've had access to the last 7 generations of XPS models and they all throttle under heavy load. You can complain to Dell all you want, they don't care because it doesn't seem to affect sales. It's an industry-wide issue, but I can only personally confirm it for Dells.
-
-
Saturday 21st July 2018 07:41 GMT Anonymous Coward
Re: Tick-Tock
I don't think so. Thermal throttling is nothing new, nor is reverting to low power states. What's going on here is that the marketing idiots at Intel tried to dress this up as a remarkable new feature, and now its come back to nip them on the bum. Compared to the Puma 6, IME, and Spectre fiascos, this is nothing, although it contributes to that nagging doubt that your laptop ought to say "AMD inside".
-
-
Friday 20th July 2018 17:58 GMT Anonymous Coward
Intel promised OEMs 10nm CPUs
Then pulled the rug out from under them a few months ago when they admitted they couldn't produce shit on that process. Supposedly they provided OEMs with some "golden samples" of 10nm CPUs earlier this year, but designing your laptop around something Intel can't deliver in anything remotely approaching production quantity won't end well. It is still on the OEMs to have backup plans.
There have been rumors for years about Apple switching to their own ARM SoCs for their Mac line, this fiasco is just giving them all the more reason to do so. After all, the A11 is faster than the i9 running at its default clock (the speed at which it apparently still throttles) They're still left with the problem that their ARM SoCs running x86 code will be far slower, but they've done 'fat binaries' and ISA transitions a couple times before so Apple knows how to handle that part.
Dell might need to finally have some serious talks with AMD, since it looks like they will have 7nm CPUs from GF well before Intel can hope to produce any of their 10nm stuff in quantity (foundry 7nm is roughly equal to Intel 10nm, so if they were both out they'd be fairly comparable from a process standpoint)
-
-
Saturday 21st July 2018 17:58 GMT Anonymous Coward
Re: Tempted to downvote...
Because the A11 is faster at equal clocks as Intel's CPUs. OK not in stuff that's heavily dependent on memory bandwidth like Linpack, so the GFLOPS won't remotely compare, but who runs Linpack on their phone? For the type of integer code dominated tasks that are common on both phones and PCs, Apple beats Intel clock for clock.
They aren't faster of course because Intel CPUs can hit double the A11's clock, as it should given its significantly higher power budget which allows the use of transistors optimized for high frequency rather than low power.
-
Saturday 21st July 2018 18:59 GMT Anonymous Coward
Re: Wait... huh?
[disengage the cloaking device, it is not hostile captain]
I'm still confused on the definition of "faster". You are comparing different workloads and different platforms. Comparing mobile to laptop. Which is fine, a compute is a compute. Except you then say we cannot take a general compute, because on mobile we don't often do that (we do web page and mobile apps, if I'm on the right page). But then compare it to a laptop made for "pro" and not "pro webpage surfing", but what people expect to be productivity. Thus general compute.
So either we need to say the A11 is not the same as the i9, and we cannot compare clocks or workloads. As my casio watch could overclock to 10GHz and be faster than both! XD
Or we say which type of compute, memory etc, and say it's "faster". I'm happy for that, to define it faster in prime, pi, content delivery, encoding or even just latency of a webpage lookup.
But just saying "the A11 [at 2.5GHz] is faster than an i9 [at 2.5GHz]" fails totally to pass logical syntax. As you say though, it is faster for some "type of integer code". That is always a "sometimes", never "it *is* faster". Thanks for the info though!
-
Sunday 22nd July 2018 15:02 GMT Anonymous Coward
Re: Wait... huh?
The A11 is faster on pretty much ALL types of integer code. Not because it is somehow better designed than Intel's CPUs or due to an advantage of ARM over x86. Its because it is designed for a clock rate of around 2.5 GHz or so, while Intel's CPUs are designed for a clock rate of around 5 GHz (because they use the same cores across their product line and want to be able to reach 5 GHz at least in turbo)
That implies caches that have more wait states, more pipeline stages etc. which leads to lower IPC. That's the compromise of targeting higher frequencies for max speed, if it runs at half that speed it has unnecessary delays in loads from cache, pipeline stages that don't get as much work done as they could, etc. If Intel designed a CPU targeted with a ceiling of 2.5 GHz instead of 5 GHz it would likely match or exceed the A11's performance at equal clocks.
The reason I said it isn't going to be faster on floating point is because a lot of floating point code is limited by memory bandwidth. Though an A11 used in a laptop would match x86 in FP, given the same memory speed.
-
Sunday 22nd July 2018 19:42 GMT Anonymous Coward
Re: Thanks!
No idea if you get this reply (does Reg do notifications?), but still, thanks so much for that extra info.
So, if I've understood, it's down to the chip silicone budget? The A11 is putting all the work at 2.5GHz to use. Where as, only the i9s running at 5GGHz would outperform it? Due to the i9's extra budget/code/transistors/cache (and wait times of all these systems).
I never knew that the A11 was actually a quicker compute too because of this. I can see where/why Intel is panicking even with the likes of AMD creeping up on "lots of slow" in their multicore systems. They don't need to be the fastest, if power and scale can still overcome the lead Intel has on pure silicone speed... which they are possibly about to also lose!
I knew some of the Apple and other chip makers had the low power chips outperforming Intel on everything (power use and compute power), but it's interesting to see they are now losing out in mid range, and possibly high end too!
PS, and all of which is partially rendered mute, if Apple (and by extension Intel) had the cooling to even run this at the expected 3.5GHz+. I also remember the stage 5GHZ they "showcased"! LOL!
-
Monday 23rd July 2018 16:42 GMT Anonymous Coward
Re: Thanks!
I'm sure the A11 could be pushed a little if you gave it more voltage in a form factor that allowed for more heat dissipation, but maybe not quite as far as 3.5 GHz since it would haven't been designed with such relatively high clocks in mind. If Apple plans to use their A* SoCs for Macs they might they might design a separate core for the Mac line or they might share it between Macs & iDevices and take a small hit on iDevice performance. If we see a new A* SoC that looks no faster or even a bit slower clock for clock, that will be a big hint that Apple is closing to putting it (probably not that exact SoC, but the big core from it) in a laptop/desktop Mac.
Of course Apple has the budget to design tweak their design for higher clocks so as not to compromise iDevice performance. Especially if they decide to use that same core in the mountains of servers they are filling their datacenters with. We might not see a server version of iOS running on ARM64 as a product, but that doesn't mean Apple won't use such a thing internally. That would help justify the cost of doing a separate core designed for higher clocks for the Mac.
-
-
-
-
-
-
-
Friday 20th July 2018 22:06 GMT The humble print monkey
Custom coolers
Silly idea, having spent the day in our air conditioned server closet (no room to sit, but hey, it is cold).
MBP (whatever model) has a known physical form. Ok, the thermal energy paste, and internal cooling are what has come out of copertino design shrink-farm.
Would it be beyond the whit of someone to design a form hugging water cooled dock/base that would drop the base ambient?
There are people who do have to use Apple applications (well, have to, are kind of tied to) fcpx etc.
I’m only suggesting the Apple water cooled dock, as it would be likely to out sell a custom dell/HP until, and there is a base level of aluminium forming part of the thermal envelope...
-
Saturday 21st July 2018 19:08 GMT Anonymous Coward
Good intentions, but you are just like me...
... prone to putting mouth before brain when it comes to ideas.
Think about it. Below ambient temperature = condensation.
Current form factor = limited thermal dissipation to even couple your customer option too (external air pumped in or external coupling of a heatsink)
There are already at least one water cooled docking station laptop on the market. But it requires the watercooling loop inside the case to begin with (then is docked to the res+pump+fans+radiator).
PS, I know I've probably also spoken before thinking long enough on this reply too! ;)
-
-
Friday 20th July 2018 23:32 GMT AZump
The solution is simple.
Download and run Throttle Stop. I have an ASUS Slate (EP-121) and it would throttle down to 667MHz before throttling up the fans. ...trying to be like Apple in the "noise" department. Well, this made the machine pretty much useless. Couldn't even watch Netflix in HD. Found Throttle Stop, locked multiplier in "Turbo" and ran at a constant 1.78GHz. Never saw temps above 80c and that was during extreme CPU loads playing KSP. Normally, working in my DAW, temps hang in the 50s, but, without Throttle Stop, it won't even get that hot before throttling. Machine is damn near useless without locking the multiplier because 1 minute after boot, it's down to .6GHz.
Screw throttling! We buy high end CPU's for their power, having them throttle down to 1999's clocks is beyond ridiculous. Especially when the damn things CAN run turbo clocks all day every day. My i7-k tower, 3rd gen, runs at 4.8GHz ALL DAY LONG. At that speed, it idles at 36c in a 76F room. I could push it to 5.2GHz, but, it won't encode video at that speed. Does everything else tho.
An old P4 runs averages higher clocks at the end of the day than modern CPUs. Hanging at 1.6GHz except when loading an app is pure BS. It's not like I have several thousand of them so what's the big f*ing deal if it eats a few more watts running full bore?
There is no difference, but it sure makes the huggie-feelie tree hugging brigade sleep better at night.
Manufactures need to wake up. A powerful CPU needs more power, more power means more heat, more heat means more cooling, more cooling means more heat, and all that adds up to LARGER laptops. Sure, everyone wants small, but those who buy the high end FAST stuff know and accept the thicker machines.
My last "desktop replacement" laptop weighed over 12 pounds. I took it everywhere. My ASUS eSlate weighs 1.1kg. Know what? Panti-waist reviewers complained about the weight. 2 pounds is too heavy!
...and that's the future folks. Processors that can run at 100c+ but throttling down before reaching half that temp, why? Because to do it right, the machine would: consume more electricity which pisses off the environmentalists, be thicker than Apples crap*, and everyone wants to be Apple, also a machine too heavy for our panti-waist youth to carry across the room.
It is what it is and its our fault for letting things get this far.
*You could have defended Apple in the past and I'd have agreed to their superiority in certain areas. ...right up until the day they went Intel. Now, it's just a regular PC with a custom "OS". Your i7 with whatever OS version is NOT superior in any way to my i7 running Slackware Linux. Actually, in many ways, it's inferior because frankly, I have 100% control over mine. ;)
-
Saturday 21st July 2018 19:11 GMT Anonymous Coward
Re: The solution is simple.
Sounds good. But kind of defeats the real problem. It's not the throttling down when not being used. This extends battery life, or lowers power use when actually idle. This also generally works fine on older models, unless... the software/hardware is buggy! So it is the over throttling up, of the failure to detect use that messes things up. :(
Glad you got KSP going though!
-
-
Monday 23rd July 2018 09:52 GMT Fading
Undervolt
If you are suffering the thermal throttling issue and are on a windows based machine then a bit of tweaking with Intel's XTU can go a long way in helping. If you undervolt the chip step by step you can get the same performance but with lower temps - hence avoiding thermal throttling. YMMV and note: only certain chips are supported.