What's new

CPU Throttling on 15 inch Surface Book 2 after 1709?

I suspect they got a bad yield somewhere. I actually replied to your comment on the notebookreview forums [as Mindinversion]

The big problem here is that it's not like a hardware defect that can be traced in testing, as the chip actually WORKS-- but they're not binning them the way they should be for this application [if they're binning them at all] Silicon lottery is usually reserved for desktop CPUs, but I'm keeping a very close eye on this one with the advent of quad core ULV chips.... It could really give shape to how Intel is really doing with shrinking die sizes and CPU performance characteristics moving into 10nm process, etc.
Ahh Notebook review. I didn't even think to look at that forum at all to see if anyone else was having problems.
I tend to agree. Maybe something with the batch of chips that was used in some of them.
After a few tries running through IPDT, even with a fan blowing on the back of the machine from and air conditioner I still have in the window (despite it being Novermber), the thing still fails the load test withing 30 seconds of hitting the CPULoad test.
Even if the processor hits into the 90s when under load I think that would be okay with me as long as it doesn't thermal throttle like crazy as it is now. I have a Dell Precision 7520 for work that will pretty much run at 92 C or so when under full load with a I7-7700HQ but never throttles for thermals either. That was what I was expecting to see with this processor. Or at least that it shouldn't hit over 95.
When I get the replacement in the mail from Microsoft, I will be sure to post back a result of the IPDT to confirm if the new one is any better or not.
 
My SB2 15 i7 has the same issue, soon as I run any intense program it shoots to 100c within seconds.

On call with support last night and mine failed the Intel IDPT test twice in a row for thermal issue and said to check thermal paste.

Microsoft is shipping me another one Monday.
I actually wonder and hope that they will ship the replacements that soon. Just looking at their website it looks like the 15 inch models won't ship till the end of the month and the 1tb version won't be until 2 weeks into December. The 1tb one is the one I have since I will be putting raw video in it.
 
13 inch does have:

Surface Book 2 - Does it throttle? - YouTube

I like the program that they use (TS Bench from Throttlestop). I had to try it to my Surface Pro and was impressed that it literally maintained 25 watts (3.7-3.8 GHz on both cores @ 100% load) for the whole 1024 computation which took 455 seconds. If Surface Pro can maintain a 25 watt below 95 degrees (worst case summer no A/C) for 455 seconds, I believe SB2 should be the same. I suggest to bring this program too aside from Intel IPDT to check for thermals. I wonder how shorter is the computation time from SB2 15 on the same test.

Untitled1.jpg Untitled.jpg

I suspect they got a bad yield somewhere. I actually replied to your comment on the notebookreview forums [as Mindinversion]

The big problem here is that it's not like a hardware defect that can be traced in testing, as the chip actually WORKS-- but they're not binning them the way they should be for this application [if they're binning them at all] Silicon lottery is usually reserved for desktop CPUs, but I'm keeping a very close eye on this one with the advent of quad core ULV chips.... It could really give shape to how Intel is really doing with shrinking die sizes and CPU performance characteristics moving into 10nm process, etc.


Never thought that silicon lottery matters now on ULV mobile chips. Doesn't Intel even test their chips piece by piece? Maybe they went with AQL sampling for batches. Can't believe how a CPU that needs 1.37 volts to keep the target frequency passes quality control to be binned as an i7. They should've binned it as i3.
 

Seneleron

Active Member
Never thought that silicon lottery matters now on ULV mobile chips

I honestly never would have suspected it myself if Dolan's screenshot hadn't shown a peak of 1.396v, and his testing with intel XTU that wouldn't let him undervolt at all without instability. If the system requires that much voltage to remain stable, that *HAS* to mean there's an issue with at least one batch of processors, as there's NO WAY that's acceptable voltage in a design as thin as the surface base. The temps 110% confirm that. Darkahn's on the other hand... either there's a cover over the pre-applied thermal paste and someone forgot to take it off, or they forgot thermal paste completely. How that one made it out of the factory I have no idea.

. If Surface Pro can maintain a 25 watt below 95 degrees (worst case summer no A/C) for 455 seconds, I believe SB2 should be the same

And my 15" SB2 does, but keep the following in mind: the 13" Surface Book 2 is PASSIVELY cooled [no fans]. The 7th gen dual core would probably do just fine, but the 8th gen has the same 8650u [albeit clocked and power gated differently] that the 15" has. Of course it's going to throttle. the 2017 Surface pro [i7] has fans as well, and being Kaby Lake and dual core, It SHOULDN'T thermal throttle.

the 15" SB2 has a larger surface area plus active cooling to dissipate heat, but to tie everything back to my first paragraph it seems like intel overreached with Kaby Lake-R and/or the processor manufacturing yields aren't as uniform as they should be. the 8650u is a brand new chip, so this isn't TOTALLY surprising, and the fact that other manufacturers can choose to dial down the performance would make the issue more difficult to detect. Add to that the fact that those ultrabooks that ARE running the 8650u are still a bit more robust [read: bigger/thicker where the motherboard is located] and it may just be that until this application nobody really noticed any problems. HOPEFULLY it was just one suspect yield, but if it's wider spread I'm sure it'll be all over the tech news in the next month or so.

I hate to say it like this, but that's sometimes the risk we take as early adopters. Obviously knowing MS will make it right isn't much consolation when you've just spent $2,500+ on a machine that doesn't work as intended, but there's always going to be a certain defect rate on products. All I can do is promise you that it's [mostly] worth the wait. I only say "mostly" because I had personally intended to use the SB2 in tandem with the surface dock connected to an asus ROG Swift PG279Q, which is made impossible by the inability of the dock to supply enough power under gaming to power the laptop without discharging the battery, but we'll see if THAT gets enough press to force MS to offer a fix or an upgraded dock.

I'm gonna stop rambling now, sorry about that: sometimes I get wrapped up in my own speculations.
 

Cimmerian

Member
Just ran AIDA64 for 20 min on my 15" 512gb Model and experienced no throttling, temps peaked at 85c and once fans were at full speed temps stayed in the high 60s low 70s. Voltage peaked just above 1.0 and settled around .8, I am however on 1703, updating now and will test again.
 
Just ran AIDA64 for 20 min on my 15" 512gb Model and experienced no throttling, temps peaked at 85c and once fans were at full speed temps stayed in the high 60s low 70s. Voltage peaked just above 1.0 and settled around .8, I am however on 1703, updating now and will test again.
Cool. Thank you very much. That was the confirmation I was thinking would be the real case. There must truly be something wrong with the one I got.
Even after downgrading mine with the recovery image back to 1703 I had the same problem. So it turns out that the upgrade wasn't the issue. At first the recovery wasn't available but showed up within hours of talking with Microsoft support.
 

Cimmerian

Member
Just finished testing after the update, so AIDA64 itself doesn't show any throttling but when monitoring the CPU directly it seems like there is. Before the test frequency was at 4GHz and voltage around 1.2, as the test went on that frequency dropped to 2.2GHz and voltage around 0.8. Is this not thermal throttling? Why wouldn't AIDA recognize it as such? Am I missing something?

AIDA64.png
 
Just finished testing after the update, so AIDA64 itself doesn't show any throttling but when monitoring the CPU directly it seems like there is. Before the test frequency was at 4GHz and voltage around 1.2, as the test went on that frequency dropped to 2.2GHz and voltage around 0.8. Is this not thermal throttling? Why wouldn't AIDA recognize it as such? Am I missing something?

AIDA64.png
There is current or voltage limits that the system should be using to balance available speed in comparison to battery usage and heat disbursement. Looking at the temperatures it never got hot enough to throttle for that reason. But more likely it just was because of the sustained draw. 4.0 is only max turbo and I don't think should be run constantly at that speed if I'm mistaken. Maybe regular turbo though which I thought was 3. something.
If you run the extreme tuning utility from Intel for monitoring, it has flags for these other two methods.
 

Cimmerian

Member
There is current or voltage limits that the system should be using to balance available speed in comparison to battery usage and heat disbursement. Looking at the temperatures it never got hot enough to throttle for that reason. But more likely it just was because of the sustained draw. 4.0 is only max turbo and I don't think should be run constantly at that speed if I'm mistaken. Maybe regular turbo though which I thought was 3. something.
If you run the extreme tuning utility from Intel for monitoring, it has flags for these other two methods.


Makes sense, thanks for explaining.

I typically benchmark my desktop not laptop and with decent custom water cooling and beefy power supply I don't really see throttling or power consumption problems.

Need to become more educated.
 
Makes sense, thanks for explaining.

I typically benchmark my desktop not laptop and with decent custom water cooling and beefy power supply I don't really see throttling or power consumption problems.

Need to become more educated.

Did you set the power plan to best performance? This keeps the CPU frequency at max turbo which is 4-4.2 GHz for the 8650u chipUntitled.jpg
 
Top