m3/m5/m7 difference under sustained load?

[Procfessor]

Apologies if this has been discussed before, can't find anything relevant with the search.

Seems that all Skylake-Y are/will be thermally limited in passive notebooks. Given that, at the same SDP of (say,) 3W, is there any difference in performance between the m3, m5 and m7? In other words at the same wattage is the m7 going to churn out more performance than the smaller siblings or when limited they are exactly the same? Or in other words what the heck is the actual difference between these 3?

Reason I'm asking is some vendors (looking at you, Dell!) charge insane premiums for m7 over m5 (~$350) whith otherwise identical systems and even though I'm not $$$-limited that's hard to justify. Especially if it means nothing during sustained loads (don't care about "snappyness" when the difference is 1.8 sec vs 2.0 sec).

Any and all replies appreciated, cheers!

 

[ShintaiDK]

It depends a lot on the design. One of the best designs in the Surface and Asus also got one. A bad designed m7 may perform like a good designed m3.

Example:

So in short, reviews, reviews and reviews.

 

[NTMBK]

Yup, all depends on how the tablet is cooled- and how the BIOS is configured.

For example my Dell Venue 11 has an i5-4300Y. Dell have configured a 6W long-power limit, so after ~20 seconds of a gaming load it throttles, hard. An i3 in a better-cooled tablet with higher power limit would no doubt perform much better.

 

[Procfessor]

It depends a lot on the design. One of the best designs in the Surface and Asus also got one. A bad designed m7 may perform like a good designed m3.

What about in identical systems with the same SDP/TDP limit? Would the m7 have any performance premium over the m5/m3 given identical enviroment?

So in short, reviews, reviews and reviews.

There are barely any reviews for m7 and couldn't find a single one for m5. Skylake both.

Yup, all depends on how the tablet is cooled- and how the BIOS is configured.

Same question, what about cases where the tablet/laptop is the same only with different CPU model installed? Would the m7 still have an advantage or at the same throttling parameters they perform identically?

For example my Dell Venue 11 has an i5-4300Y. Dell have configured a 6W long-power limit, so after ~20 seconds of a gaming load it throttles, hard. An i3 in a better-cooled tablet with higher power limit would no doubt perform much better.

Interesting. Is there a way to modify that limit (eg XTU?) or are you stuck with it? Also how badly does ambient temp influence the throttling? I'm planning to use it in 30c+ environments a lot so it does matter I guess.

 

[NTMBK]

What about in identical systems with the same SDP/TDP limit? Would the m7 have any performance premium over the m5/m3 given identical enviroment?

There are barely any reviews for m7 and couldn't find a single one for m5. Skylake both.


Same question, what about cases where the tablet/laptop is the same only with different CPU model installed? Would the m7 still have an advantage or at the same throttling parameters they perform identically?

Interesting. Is there a way to modify that limit (eg XTU?) or are you stuck with it? Also how badly does ambient temp influence the throttling? I'm planning to use it in 30c+ environments a lot so it does matter I guess.

Perhaps if the m7 chip comes from a better bin and can run at a lower voltage than the m3/m5, it could run faster within a given power limit; but I don't know if anyone has tested that.

Sadly temperature does not really affect this power throttle- as shown in XTU, a 6W package limit kicks in, and then everything is throttled, even when it is nowhere near the thermal throttle point. Only way I have seen to get around it is either hacking values in the BIOS (which I am not trying on a £900 tablet!), or using Throttlestop's "Powercut" feature. But this in effect lies to the CPU about how much power it is drawing, meaning it could e.g. end up drawing way to much power and ruin your battery. Again, I don't want to try that

You can mitigate slightly by undervolting your CPU to reduce the power consumption, so it can maintain higher clocks within that 6W envelope, but there is only so much you can do.

 

[Procfessor]

Hmm, so you're saying the m7 is not worth it as it's processor lottery at best, wasted money at worst. Thanks!

However one question. In the Anand article about the m3 Asus and m7 Dell/Lenovo, it seemed like the aluminum housing of the Asus is a huge advantage. So, in other words, thermals at work.

 

[nismotigerwvu]

Yup, all depends on how the tablet is cooled- and how the BIOS is configured.

For example my Dell Venue 11 has an i5-4300Y. Dell have configured a 6W long-power limit, so after ~20 seconds of a gaming load it throttles, hard. An i3 in a better-cooled tablet with higher power limit would no doubt perform much better.

I've got one those and man did the BIOS edit that is floating out there online wake it up! It bumped up the throttle temp and unlocked the TDP. With the Intel extreme tuning utility (nice piece of software but an awful name) I can set it to tun at a 9 watt TDP and it still doesn't come close to the throttle point with a slight undervolt (-40mV core, -10mV cache, -10mV GPU).

 

[bhtooefr]

I don't think NTMBK was saying that the m7 is processor lottery, I think they were saying that the m7 may be better binned than the m5, and that's how it would be faster.

Unfortunately, looks like the only good way to test is to buy one of each in the same model, and compare in benchmarks.

 

[Procfessor]

I've got one those and man did the BIOS edit that is floating out there online wake it up! It bumped up the throttle temp and unlocked the TDP. With the Intel extreme tuning utility (nice piece of software but an awful name) I can set it to tun at a 9 watt TDP and it still doesn't come close to the throttle point with a slight undervolt (-40mV core, -10mV cache, -10mV GPU).

I would like this if there was an option to like. ) My main fear is this, btw, Dell and their long and glorious history with thermal management.. cough cough... Looking at the new latitude with core m.

I don't think NTMBK was saying that the m7 is processor lottery, I think they were saying that the m7 may be better binned than the m5, and that's how it would be faster.

Unfortunately, looks like the only good way to test is to buy one of each in the same model, and compare in benchmarks.

Oh I see. Well, "may be" still can't justify the difference of $350 between the m5 and m7.

 

[bhtooefr]

Another way to look at performance, BTW...

Base clocks:

        Improvement over:
Model   m3-6Y30 m5-6Y5x
m5-6Y5x     22%
m7-6Y75     33%      9%

0.9, 1.1, and 1.2 respectively.

Turbo clocks:

        Improvement over:
Model   m3-6Y30 m5-6Y54 m5-6Y57
m5-6Y54     23%
m5-6Y57     27%      4%
m7-6Y75     41%     15%     11%

2.2, 2.7, 2.8, and 3.1 respectively.

GPU turbo clocks:

        Improvement over:
Model   m3-6Y30 m5-6Y5x
m5-6Y5x      6%
m7-6Y75     18%     11%

850, 900, and 1000 respectively.

I'd expect those numbers to hold for the different SKUs in the same model of PC (with the caveat that memory bandwidth is also a factor, and a lot of these machines will have single-channel memory, even though all of these parts support dual channel LPDDR3-1866 - that'll hurt IGP performance such that the IGP clocks don't matter).

As far as feature support, per Ark, the m5-6Y57 and m7-6Y75 have vPro, TSX-NI, SIPP, and TXT support, which the m3-6Y30 and m5-6Y54 lack.

Myself, I'd say that a m5 is the sweet spot unless your vendor's m5 option is 6Y54 and you need any of those missing features (if you're a home user, you don't).

Disclaimer, and I didn't pay enough attention to the first post, so getting this out here: I work for Dell, but these comments are my own opinion, I'm not representing Dell on here, and I've got no familiarity with any of the Skylake-Y products that we sell (my client hasn't deployed any of them, and that's how I get familiarity with them, unless I buy them for myself).

 

[NTMBK]

Hmm, so you're saying the m7 is not worth it as it's processor lottery at best, wasted money at worst. Thanks!

However one question. In the Anand article about the m3 Asus and m7 Dell/Lenovo, it seemed like the aluminum housing of the Asus is a huge advantage. So, in other words, thermals at work.

Depends on the BIOS and the chip in question I guess! Remember I have an old Haswell part, these new Broadwell and Skylake parts seem much happier within these tight power limits. Or perhaps their BIOS is configured to be much more lenient with the power limit.

As Shintaidk says, read lots of reviews, and if you can try out the device in person to see if it meets your needs.

 

[NTMBK]

I've got one those and man did the BIOS edit that is floating out there online wake it up! It bumped up the throttle temp and unlocked the TDP. With the Intel extreme tuning utility (nice piece of software but an awful name) I can set it to tun at a 9 watt TDP and it still doesn't come close to the throttle point with a slight undervolt (-40mV core, -10mV cache, -10mV GPU).

Hmm, interesting! So can you set it up to normally run it with a 6W TDP cap in place to save battery life, then use XTU to uncap it when you want to e.g. game on it?

Tempted to try it out, because this thing is outside of warranty anyway.

 

[Raduque]

I have the HP Spectre x2 with the m3-6y30 and runs it's ram in dual channel (2x2gb). It's significantly faster than my Surface 3 Atom X7, and I don't know if benches will bear this out, but it feels as fast as my Surface Pro 2 with the i5-4200u. I don't have the Pro2 anymore to compare, sadly.

 

[Procfessor]

I'd expect those numbers to hold for the different SKUs in the same model of PC...

Well, yea, until you run into thermal and/or power throttling issues. At which point, I guess, all things considered the same they should be having equal performances, no? In other words the m7 is superior as long as you don't get throttled ("snappyness") but if you keep up the load (rendering, compressing video, gaming, anything) they'll pretty much fall back to the same point in the same machine (save for individual proc and case differences). Which was kind of my question.

Anyway I'm also leaning towards the m5, will probably buy that.

 

[IntelUser2000]

I have the HP Spectre x2 with the m3-6y30 and runs it's ram in dual channel (2x2gb). It's significantly faster than my Surface 3 Atom X7, and I don't know if benches will bear this out, but it feels as fast as my Surface Pro 2 with the i5-4200u. I don't have the Pro2 anymore to compare, sadly.

Intel boosted Core m3 performance by 30% over previous 5Y10 chips but from the few tests available, beyond that its just like Broadwell Core M in there is zero advantage getting higher SKUs. No real reason for m5 and m7 for Skylake.

I am interested to see if Huawei's Matebook can change that. They claim that they have a thermal design which cools the device that increases battery life to live up to their claim of 10 hours on a 33WHr battery. Maybe they did something different in performance wise and make m5 worth it? I don't know.

Seems that all Skylake-Y are/will be thermally limited in passive notebooks. Given that, at the same SDP of (say,) 3W, is there any difference in performance between the m3, m5 and m7?

I am not sure if there's much difference due to binning, if at all anymore. It used to. It's just like how process advancements offer mediocre gains. Probably because most of the gains you'd get from binning/process is used up by fancy circuitry tricks like Turbo Boost 2.0, and PCU(Power Control Unit).

I'd say at this level only m5 would show any gain. Now, that's an IF scenario. Maybe m3 is clock capped under TDP, and m5 would reach TDP so it would be faster.

 

[bhtooefr]

m3 is higher TDP, too, so I'd expect it to throttle even sooner than my numbers would indicate.

 

[Procfessor]

...there is zero advantage getting higher SKUs. No real reason for m5 and m7 for Skylake.

Yep, my thoughts exactly. Last time I cared about binning was in 2001 for my Northwood. Sadly some manufacturers really limit the options for the m3 systems (not only Dell but shame on them too anyway lol) so it's not really an option.

m3 is higher TDP, too, so I'd expect it to throttle even sooner than my numbers would indicate.

Can you elaborate on this please, maybe with a table for all 3 variants?

 

[bhtooefr]

Actually, I take that back, it's higher MINIMUM TDP (as in, cTDP down). At 600 MHz, it's 3.8 W, where m5 and m7 are 3.5 W.

 

[Raduque]

According to CPU-Z, the m3 in my HP has a max TDP of 4.5w.