Opinion: It's time we talked about throttling in reviews

[Redaktion]

Due to the prevalence of thermal throttling and new passive-cooling technologies, it is no longer enough to run a few short benchmarks and declare a machine's performance is satisfactory.

https://www.notebookcheck.net/Opinion-It-s-time-we-talked-about-throttling-in-reviews.234232.0.html

[I agree!]

I agree with everything you wrote, time to crack down on throttling laptops that can't sufficiently cool their high end advertised components!  Although, I am a bit confused by your article, what are you gonna change in the reviews that you do to highlight the points you made in this article, aren't you already highlighting throttling issues in your articles, how are you gonna make that more apparent?

[Rob Mc]

Thank you for your article, and all of your laptop reviews. I find them to be the most comprehensive on the internet.

Question: What is the difference (from a user perspective) between the results of the "stress test' which runs Prime95 and FurMark, and the other throttling test of running Cinebench several times in a row.

For example, on my Dell XPS 13 9360, I have noticed that when stressed, the CPU clock will drop considerably. However, I ran the Cinebench multi-core test for almost 30 minutes, and it always returned with a score of 330...so no drop at all.

[andregomes]

That's why I trust your reviews above any others. Keep up the good work.

[D16]

If you are able to get your hands on the Asus Zephyrus (top end model) i would love to see the throttling in that laptop design. Both in CPU and GPU. Thanks!

[edit1754]

Definitely agree. There are laptops I really want to like and recommend (Acer VN7-593G), but can't due to so many reports thermal throttling issues.

Another important issue I think needs to be addressed is Intel's confusing CPU branding. The consumer base has been made to believe that the Core i3/i5/i7 terminology is a defining aspect of a computer, and is a good descriptor of CPU performance, while in fact the different 'classes' of CPUs (Y-series, U-series, HQ-series) has much more correlation with it. The difference between a U-series "i5" and a U-series "i7", for example, is near negligible for most practical use.

This not only has consequences for individual users who aren't informed, it has consequences for the availability of well-balanced configurations in the first place. So even if you know what you look for, you might not be able to find it. It causes manufacturers to put cost-of-production budget towards minimally-faster CPUs at sacrifice of decent displays (ASUS K501UW), and it causes the 7700HQ + GTX 1050 combination to be more prevalent than 7300HQ + 1050 Ti, which is better for most GPU-dependent games.

I propose that U-series CPUs be capped at "i5" branding, and that all HQ-series CPUs earn the title of "i7".

[edit1754]

Meant to add -- and I propose that reviews call this out, from a point of view of firm grasp of the issue.

[Douglas Black]

Thank you for your article, and all of your laptop reviews. I find them to be the most comprehensive on the internet.

Question: What is the difference (from a user perspective) between the results of the "stress test' which runs Prime95 and FurMark, and the other throttling test of running Cinebench several times in a row.

For example, on my Dell XPS 13 9360, I have noticed that when stressed, the CPU clock will drop considerably. However, I ran the Cinebench multi-core test for almost 30 minutes, and it always returned with a score of 330...so no drop at all.

That will result in 100% CPU and 100% GPU load. It's an unrealistic scenario but it pushes the system to the absolute limits.

Cinebench is a CPU load, so it's actually less strenuous.

I agree with everything you wrote, time to crack down on throttling laptops that can't sufficiently cool their high end advertised components!  Although, I am a bit confused by your article, what are you gonna change in the reviews that you do to highlight the points you made in this article, aren't you already highlighting throttling issues in your articles, how are you gonna make that more apparent?

Since about 6 months ago we started doing the cinebench loop to check for CPU throttling, so this is something we have started watching a bit more closely. It was more focusing on the reviewing community/industry as a whole.

[ZM]

Well done. That's what I've been looking for. Notebookcheck doesn't disappoint me.

[GreyFox]

In a word, wrong.
The package has a 15W TDP, that's the design point. Turbo Boost is not continuous, the fact that it hits 25W or 30W during peak Turbo Boost is in no way intended to be sustained of it was the design point would be 25W or 30W. 

"I've established that a CPU should hold its turbo boost under load or it is considered throttling,"
Where, please quote the Intel engineering reference that establishes this. You can declare anything you want but that doesn't make it a fact.   

[Dude]

P0 is the highest clock state on any Intel processor.  The P0 states are now the Turbo Boost clocks, where previously they were simply the top clock speed before Speed Step.  Intel permits computer makers to freely code the EFI/BIOS to restrict access to the P0 state based on power, current or thermal limits.  Clearly, not all manufacturers choose the same restrictions.  Some only code thermal limits because the underlying power hardware (VRMs, mostly) are capable of sustaining extended P0 states without issues, likely most Macbooks with Turbo Boost that this site has reviewed are coded like this.  Other hardware will throttle (or exit P0 state) even if thermal limits has not been reached due to power/current restrictions. 

I think it should be published what the exact source of the throttle is, for each device tested (i.e., list the power/current/temperature limits and time set on the device, with perhaps Intel XTU).  In a few cases, the restrictions may not thermally related.

[I agree!]

Thank you for your article, and all of your laptop reviews. I find them to be the most comprehensive on the internet.

Question: What is the difference (from a user perspective) between the results of the "stress test' which runs Prime95 and FurMark, and the other throttling test of running Cinebench several times in a row.

For example, on my Dell XPS 13 9360, I have noticed that when stressed, the CPU clock will drop considerably. However, I ran the Cinebench multi-core test for almost 30 minutes, and it always returned with a score of 330...so no drop at all.

That will result in 100% CPU and 100% GPU load. It's an unrealistic scenario but it pushes the system to the absolute limits.

Cinebench is a CPU load, so it's actually less strenuous.

I agree with everything you wrote, time to crack down on throttling laptops that can't sufficiently cool their high end advertised components!  Although, I am a bit confused by your article, what are you gonna change in the reviews that you do to highlight the points you made in this article, aren't you already highlighting throttling issues in your articles, how are you gonna make that more apparent?

Since about 6 months ago we started doing the cinebench loop to check for CPU throttling, so this is something we have started watching a bit more closely. It was more focusing on the reviewing community/industry as a whole.

Well, keep up the good work!  As an idea, if you wanted to further highlight throttling/temperature issues in laptops then you could weight that variable more heavily in your overall laptop scores, thereby making this an issue of greater importance & visibility.

[Douglas Black]

In a word, wrong.
The package has a 15W TDP, that's the design point. Turbo Boost is not continuous, the fact that it hits 25W or 30W during peak Turbo Boost is in no way intended to be sustained of it was the design point would be 25W or 30W. 

"I've established that a CPU should hold its turbo boost under load or it is considered throttling,"
Where, please quote the Intel engineering reference that establishes this. You can declare anything you want but that doesn't make it a fact.

Directly next to the sentences that you quoted I pointed out that the CPU registers a throttling return when this happens, so it is considered throttling by Intel's own definition. :)

[Ashley Gaunt]

The situation is even worse now.. I have an hp spectre 15 X360. With a bit of fiddling the throttling can be kept so the 3.1ghz 8590 stays at 3.5ghz. I don't expect more as it's a 20mm laptop. It takes overclocking and undervolting a bit too making it a good laptop for non prolonged work like cad.

However, hp have made it throttle down to half the advertised ghz when in tent mode or tablet mode. This is unacceptable.