Did anyone actually saw i9 boost to 4.8?

[leman]

You're paying for a design choice that is not capable of running turbo clocks for longer than 3 seconds. If a proper cooling solution was used than there would've been no problem. It's Apple's choice to make these Macbooks so thin, so I most definitely understand why the performance is so poor.

Sigh... You folks are very stubborn in your refusal to accept that turbo clocks are opportunistic and that different designs come with different balance of features. Your confusion probably stems from the fact that the Coffee Lake CPUs simply have such a wide performance range with their very high automatic overclocking ceilings. Equip them with a desktop-class cooling and they ill perform as a desktop-class CPU. Compare the turbo clocks on the i7-8086K — the fastest Coffee Lake desktop CPU, and the mobile i9 — they are practically identical. Put the mobile i9 into a large laptop that is able to supply and dissipate over 100W of CPU power, and you just might see that 4.8 Turbo. In a "normal" laptop designed for the "normal" 45Watt? Good luck with that.

The bottom line is that due to how the Coffee Lake is configured, the laptop manufacturers can target different ranges of performance — at the expense of form factor — using the same SKU. A laptop that is able to utilise the upper end of i9's performance will be 2-3 times larger (volume wise) and 1+kg heavier than the MBP or a comparable design. The MSI Titan for example has a power and cooling system of a beefy desktop computer and can treat the i9 as a desktop part — but its also 4.56kg heavy.

Want max performance? Go for a bulky workstation or gaming laptop. The MacBook Pro, which is — and always has been — designed and marketed as a thin and light machine, never offered that. Its selling point was a "perfect" balance of mobility, battery and still premium-level CPU performance. But machines like Dell Precision or HP ZBook have always offered superior performance configurations, but this performance came at a massive hit in mobility and battery. This was the case 10 years ago and it is still the case today.

The fact is that all laptops in the same size and weight category — the MBP, the Dell XPS, the Razer Blade — offer comparable performance. Neither of them is able to reach max. turbo clocks on those CPUs. Which is perfectly fine, since they are still significantly faster than their predecessor models and the i9 runs faster than the i7. And you claims of "poor performance" are simply ludicrous. The MBP is able to sustain clocks of 3.1-3.3Ghz under 100% 6-core utilisation, which is on par with its competitors and above Intel's guaranteed nominal spec. BTW, it would do you good to read the spec — it's in public domain.

If Apple made the MBP twice as thick and 3kg heavy (with power adapter) to make you happy, how many people do you think would buy it? Maybe Queen would, and maybe some other "pros" from these forums. That would be truly the end of the MBP brand and a sure way to kill the Mac for good.
[doublepost=1540329920][/doublepost]

still, my damn display shouldnt be flickering at 5grand price tag.

That for sure. There seems to be a bug in Mojave's Safari right now, I also see bad flickering when watching youtube in fullscreen. Is that what you experience? Anyway, I hope they fix it soon, its beyond annoying.

 

[Saeloe]

Using cpuSetter, disabling all but one core and no hyperthreading, logging with Intel Power Gadget using 1 ms refresh rate, maxing fans with iStat and then running short "yes > /dev/null &" burst or with Prime95 using one thread, checking the Intel Power Gadget log file I can see clock rate of 4700 MHz for periods of 1-3 milliseconds but not higher on i9. After 1-3 milliseconds, the clock rate drops to 4600 MHz and stays there.

Thermal Velocity Boost requires CPU temperature under 50 celcius and available power budget https://en.wikichip.org/wiki/intel/thermal_velocity_boost
iStat and Intel Power Gadget shows CPU temperatures of 35 C - 43 C between runs but peaks immediately to 90-100 C.

Can anyone else test this method?

 

[solouki]

Using cpuSetter, disabling all but one core and no hyperthreading, logging with Intel Power Gadget using 1 ms refresh rate, maxing fans with iStat and then running short "yes > /dev/null &" burst or with Prime95 using one thread, checking the Intel Power Gadget log file I can see clock rate of 4700 MHz for periods of 1-3 milliseconds but not higher on i9. After 1-3 milliseconds, the clock rate drops to 4600 MHz and stays there.

Thermal Velocity Boost requires CPU temperature under 50 celcius and available power budget https://en.wikichip.org/wiki/intel/thermal_velocity_boost
iStat and Intel Power Gadget shows CPU temperatures of 35 C - 43 C between runs but peaks immediately to 90-100 C.

Can anyone else test this method?

Yes...

Saeloe, this is precisely how I had timed my runs listed in my earlier post in this thread, except that I used 100ms instead of 1ms time interval for the moving average. I found that using Prime95 on 1 thread with the MBP in a cool, low humidity, room with a fan blowing on the MBP, that the Intel Power Gadget would rise to 4.47 GHz momentarily. I'm fascinated that you found 4.7 GHz for 1-3 ms that then drops to 4.6 GHz. I'll rerun my tests but with a shorter interval to see if I get these speeds also.

Thanks.

P.S. I haven't used iStat in a long time, but I see you use it under Mojave to set the fan speeds?

 

[IngerMan]

I ran GB4 with Intel Gadget and iStat. Intel looks to be reporting lower turbo stats than iStat for the same run. My temps were below 50C for portions of the test which would allow it to max turbo on my i7 2.9 turbo 3.9 Kaby Lake.

 

[solouki]

Rerun of 1 CPU, no hyperthread, under Mojave...

Saeloe,

I did find a few CPU readings of 4600 but none at 4700 running 1 thread of Prime95. This rerun also differs from my earlier runs because my earlier tests were under HS+SU2 while this one was under Mojave 10.14.
[doublepost=1540336843][/doublepost]

I ran GB4 with Intel Gadget and iStat. Intel looks to be reporting lower turbo stats than iStat for the same run. My temps were below 50C for portions of the test which would allow it to max turbo on my i7 2.9 turbo 3.9 Kaby Lake.

View attachment 797812 View attachment 797813 View attachment 797814

Moving Average...

IngerMan, I don't currently have iStat, so I don't know what time interval they use to measure the moving average of the CPU frequencies. Can you specify the moving average time interval on iStat? Or does iStat tell you what time interval it is using? The Intel Power Gadget lets one specify the time interval (they call it "Sampling Resolution (ms)"), and so if you knew the iStat time interval then you could potentially set the Power Gadget time interval to the same value, then the CPU frequencies should be relatively close on both iStat and the Intel Power Gadget.

But since these CPU frequency measurements are moving averages over some time interval, if the time intervals differ then the moving averages will differ, even for the identical data stream.

In addition, I believe TVB was implemented in Coffee Lake and wasn't available to Kaby Lake?

EDIT: Also IngerMan, do you have a GB4 with the Stress Test? Mine doesn't have it and I thought GeekBench removed the Stress Test from GB4.

In summary, I would not expect iStat and the Intel Power Gadget to measure the same CPU frequencies unless, by coincidence, they are using the same time interval for their moving averages.

 

[1096bimu]

so much ignorance....
Max power in gen 8 processors only go for 500ms max, so you can't see it unless your monitor program refreshes in less than 500ms.

And yes it happens all the time in room temperature, not just in freezers. Putting a computer in a freezer isn't even an effective cooling method, just shows how ignorant that Youtuber actually is.

 

[Thysanoptera]

so much ignorance....
Max power in gen 8 processors only go for 500ms max, so you can't see it unless your monitor program refreshes in less than 500ms.

And yes it happens all the time in room temperature, not just in freezers. Putting a computer in a freezer isn't even an effective cooling method, just shows how ignorant that Youtuber actually is.

I'm really fed up with you calling everyone ignorant while you don't know squat, so let me explain:

1. Power limits are set by OEM, not Intel.
2. Times at max power are not set as constant time, but as a coefficient used in equation defining exponential weighted moving average power. And while this coefficient is published in seconds (not milliseconds) it doesn't by itself define how much CPU will spend at its power limit. Which again is different for each power state which each has its own max power.
3. The 8th gen Intel CPU consists of boatload of SKUs with each having different default PL and Tau limits. And they are configurable by OEM if they wish so. See Pt 1.
3. This thread is about frequency, not power.
4. People submit their result during single core operation, so even if it was about power it would't be about max.

 

[mdnz]

Sigh... You folks are very stubborn in your refusal to accept that turbo clocks are opportunistic and that different designs come with different balance of features. Your confusion probably stems from the fact that the Coffee Lake CPUs simply have such a wide performance range with their very high automatic overclocking ceilings. Equip them with a desktop-class cooling and they ill perform as a desktop-class CPU. Compare the turbo clocks on the i7-8086K — the fastest Coffee Lake desktop CPU, and the mobile i9 — they are practically identical. Put the mobile i9 into a large laptop that is able to supply and dissipate over 100W of CPU power, and you just might see that 4.8 Turbo. In a "normal" laptop designed for the "normal" 45Watt? Good luck with that.

The bottom line is that due to how the Coffee Lake is configured, the laptop manufacturers can target different ranges of performance — at the expense of form factor — using the same SKU. A laptop that is able to utilise the upper end of i9's performance will be 2-3 times larger (volume wise) and 1+kg heavier than the MBP or a comparable design. The MSI Titan for example has a power and cooling system of a beefy desktop computer and can treat the i9 as a desktop part — but its also 4.56kg heavy.

Want max performance? Go for a bulky workstation or gaming laptop. The MacBook Pro, which is — and always has been — designed and marketed as a thin and light machine, never offered that. Its selling point was a "perfect" balance of mobility, battery and still premium-level CPU performance. But machines like Dell Precision or HP ZBook have always offered superior performance configurations, but this performance came at a massive hit in mobility and battery. This was the case 10 years ago and it is still the case today.

The fact is that all laptops in the same size and weight category — the MBP, the Dell XPS, the Razer Blade — offer comparable performance. Neither of them is able to reach max. turbo clocks on those CPUs. Which is perfectly fine, since they are still significantly faster than their predecessor models and the i9 runs faster than the i7. And you claims of "poor performance" are simply ludicrous. The MBP is able to sustain clocks of 3.1-3.3Ghz under 100% 6-core utilisation, which is on par with its competitors and above Intel's guaranteed nominal spec. BTW, it would do you good to read the spec — it's in public domain.

If Apple made the MBP twice as thick and 3kg heavy (with power adapter) to make you happy, how many people do you think would buy it? Maybe Queen would, and maybe some other "pros" from these forums. That would be truly the end of the MBP brand and a sure way to kill the Mac for good.
[doublepost=1540329920][/doublepost]

That for sure. There seems to be a bug in Mojave's Safari right now, I also see bad flickering when watching youtube in fullscreen. Is that what you experience? Anyway, I hope they fix it soon, its beyond annoying.

Why don’t you read the posts of OP properly? He’s stating the clocks mentioned on Apple’s site are NEVER reached. I dont need a lesson from you how turbo clocks work.

And a laptop running 100c when doing a regular load of work is pretty poor performance in my book. Before you scholar me again, I know it is (barely) under the TJMAX, it doesnt mean it is a-okay to run like that. You’re free to think whatever about that.

 

[leman]

1. Power limits are set by OEM, not Intel

As you yourself wrote, the defaults are set by Intel. OEM can change the settings. Or the user can change the settings. To make it more confusing, Apple seems to completely bypass Intel's power management and regulates power delivery externally.

3. This thread is about frequency, not power.

They are related though. The i9 can draw a ridiculous amount of power at its max turbo frequency. I think MSI measured over 130 watts at some point. Not something that a normal laptop designed for a 45W SKU can deliver or dissipate, not that it should be able to.
[doublepost=1540363255][/doublepost]

Why don’t you read the posts of OP properly? He’s stating the clocks mentioned on Apple’s site are NEVER reached.

Why don't you read my post properly? They are never reached because you need desktop-level cooling to reach them, because Intel has set them ridiculously high. That CPU should have max turbo boost of 4.6 — would still have made it a premium SKU and prevented this entire dumb discussion.

I dont need a lesson from you how turbo clocks work.

Apparently I do since you don't seem to understand the relationship between power/cooling requirements and the dynamic nature of the CPU clock.

And a laptop running 100c when doing a regular load of work is pretty poor performance in my book.

The laptop is designed to hit its nominal TDP under full load while still running at a safe operating temperature. When you look at relationship between power draw and the temperature (I posted the graphs on these forums multiple times), you can see that it has been carefully designed this way. This is all about efficiency. Apple has designed the laptop to maintain sustained CPU at exactly 45W or slightly above while cooling it down only as much as it really needs cooling, and that's why it hits 100C when consuming this much power.

The things I would criticise with Apple's thermal design is a) they have difficulties cooling the CPU and GPU at the same time, b) the cooling system is not fast enough in removing heat that rapidly accumulates, which means that the CPU heats up very slickly when doing single-core turbo and hits the thermal limit almost immediately. A bit more buffer would have been nice, so that CPU could burst higher for a few extra seconds.

 

[leman]

That CPU should have max turbo boost of 4.6 — would still have made it a premium SKU and prevented this entire dumb discussion.

By the way, I am a retard. How could I have missed it? The i9 max turbo boost is 4.6.... the extra 0.2Ghz comes only from the "turbo velocity boost" marketing gimmick which will rarely have any relevance in real life, unless you really use high-end desktop-class cooling...

 

[Queen6]

@queen: sure, I’m a chassis that is over twice the volume and 1 kg more than the MBP. We’ve been over it over and over and you still can’t see that you are comparing different categories of machines. I care about performance, but I also care very much about mobility. I’m happy to pay 20% sustained performance for halving the footprint. Call me again when you find a 2kg laptop than can sustain 3.9Ghz, now that would be impressive.

And about “professionals forced to get gaming laptops”... what are you even talking about? Apple never made a high-performance workstation to begin with. Precision, HP ZBook etc. - all of them always came with significantly faster hardware, at the price of form factor. Anyway, why did you get a gaming laptop and not a real workstation to begin with?

Half the footprint volume maybe and at 800 grams over 2KG I'm fine with it even travelling internationally, and a tad more than 20% in performance compared to Apple's latest & greatest. Primary reason for the size is the 17.3" display, notebook being one of the smaller 17's on the market.

Why? I don't need Xeon processors or ECC RAM so why pay a premium and it's clearly just as powerful, with a good deal of scalability, nor is the 3 year warranty too shabby either

Q-6

 

[mdnz]

Apparently I do since you don't seem to understand the relationship between power/cooling requirements and the dynamic nature of the CPU clock.

Yeah I dont. You're calling hitting 100c "efficient" says enough (FYI the CPU throttles so it doesn't exceed 100c). Have a nice day dude.

 

[leman]

Yeah I dont. You're calling hitting 100c "efficient" says enough (FYI the CPU throttles so it doesn't exceed 100c). Have a nice day dude.

You obviously don't. What is relevant is the sustained maximal power limit (the TDP). If it's set to 45Watts, it doesn't matter whether the CPU is running at 100 degrees or at 20 degrees — it won't go any faster since it's already at the power limit. Example: Razer Blade keeps the CPU at 80C or under, but it's not any faster, exactly since the TDP is hit before the temperature limit is. Apple instead designed their cooling system so that it hits the thermal limit and the nominal TDP limit at the same time. This is efficient design, you get the same performance but don't need to cool the CPU that much. It shouldn't be that difficult to grasp.

Two quick notes: yes, if you have the cooling power, you can choose to set the TDP higher or remove it altogether. Then you will get better performance, but you will be also running the CPU above its spec. Apple decided not to do it, some OEMs do (especially in a large laptop). And yes, running hardware on high temps will reduce its lifespan, but I don't think that anybody really cares whether their CPU will live 50 years or 15 years.
[doublepost=1540371193][/doublepost]

Half the footprint volume maybe and at 800 grams over 2KG I'm fine with it even travelling internationally, and a tad more than 20% in performance compared to Apple's latest & greatest. Primary reason for the size is the 17.3" display, notebook being one of the smaller 17's on the market.

Fair enough, man Personally, I'm more than fine with sacrificing some performance if it means I can comfortably carry my 15" in a compact messenger bag. I suppose I am exactly the kind of user Apple aims at with their MBP. Different people with different needs, different designs.

 

[Saeloe]

Rerun of 1 CPU, no hyperthread, under Mojave...

Saeloe,

I did find a few CPU readings of 4600 but none at 4700 running 1 thread of Prime95. This rerun also differs from my earlier runs because my earlier tests were under HS+SU2 while this one was under Mojave 10.14.
[doublepost=1540336843][/doublepost]

Moving Average...

IngerMan, I don't currently have iStat, so I don't know what time interval they use to measure the moving average of the CPU frequencies. Can you specify the moving average time interval on iStat? Or does iStat tell you what time interval it is using? The Intel Power Gadget lets one specify the time interval (they call it "Sampling Resolution (ms)"), and so if you knew the iStat time interval then you could potentially set the Power Gadget time interval to the same value, then the CPU frequencies should be relatively close on both iStat and the Intel Power Gadget.

But since these CPU frequency measurements are moving averages over some time interval, if the time intervals differ then the moving averages will differ, even for the identical data stream.

In addition, I believe TVB was implemented in Coffee Lake and wasn't available to Kaby Lake?

EDIT: Also IngerMan, do you have a GB4 with the Stress Test? Mine doesn't have it and I thought GeekBench removed the Stress Test from GB4.

In summary, I would not expect iStat and the Intel Power Gadget to measure the same CPU frequencies unless, by coincidence, they are using the same time interval for their moving averages.

I don't know the sample rate of iStat but it shows the average of 6 seconds, so it's not very precise tool to check the peak frequency. I use Intel Power Gadget logging on 1 ms sampling resolution and then I filter the log with Excel. In the log I can see the 4700 MHz for few milliseconds (there is row per millisecond), but mostly it stays at 4600 MHz. In the log you see the temperature and power too per millisecond.

iStat Menus 6.2 is excellent for monitoring temperatures and frequencies of whole MacBook Pro (CPU, both GPU's, Battery, charging, different voltages etc). Give it a try, it has 14 days trial.

 

[Thysanoptera]

They are related though. The i9 can draw a ridiculous amount of power at its max turbo frequency. I think MSI measured ove r 130 watts at some point. Not something that a normal laptop designed for a 45W SKU can deliver or dissipate, not that it should be able to.

I wanted to wait until he comes back and calls me ignorant, but have better things to do for now. They are dependent, of course, linearly at constant temperature to be exact. But we’re not in 1980’s anymore and there is a myriad of scheduling techniques that CPU makers employ, the main CPU clock modulation is just one of many. That's 8700k, which for all intends and purposes is the same as 87xxh series, locked to 4.3GHz, while idle @4.3=20W, stress test @4.3=70W.

 

[leman]

View attachment 797914
I wanted to wait until he comes back and calls me ignorant, but have better things to do for now. They are dependent, of course, linearly at constant temperature to be exact. But we’re not in 1980’s anymore and there is a myriad of scheduling techniques that CPU makers employ, the main CPU clock modulation is just one of many. That's 8700k, which for all intends and purposes is the same as 87xxh series, locked to 4.3GHz, while idle @4.3=20W, stress test @4.3=70W.

Why would I call you ignorant? I agree with most of your post, just wanted to comment on few points.

I am not sure what your message is in this last one, probably because I have difficulty reading that graph as it lacks axis labels. Are you really sure that the CPU is at 4.3Ghz when idling? It looks to me very much like it was in a lower power state...

 

[Queen6]

Fair enough, man Personally, I'm more than fine with sacrificing some performance if it means I can comfortably carry my 15" in a compact messenger bag. I suppose I am exactly the kind of user Apple aims at with their MBP. Different people with different needs, different designs.

To be fair the MBP has always been focused towards portability certainly more so in recent years, although for me I think Apple has gone to far and compromised the usability for some. I think Apple will just try to hold out for Intel to get the power consumption down again, then the performance will rise, or switch the rumoured ARM CPU's For most the MBP's performance is likely adequate although the heating potentially annoying.

I'd just do what I did in the past and manually take control of the fans, I'm not sure much else can be done...

Q-6

 

[Thysanoptera]

Why would I call you ignorant? I agree with most of your post, just wanted to comment on few points.

I am not sure what your message is in this last one, probably because I have difficulty reading that graph as it lacks axis labels. Are you really sure that the CPU is at 4.3Ghz when idling? It looks to me very much like it was in a lower power state...

Not you, come on, you don’t do this, I talked about the guy I quoted in the post before, Mr. Scientist bent on educating ignorant masses.

Yes, it was at 4.3 GHz, and sure as hell it was at lower power state, but still running the same base frequency. That's the whole point, you don't have to lower the frequency to get lower power consumption, you can shut down parts of the core, turn off cache, lower voltage etc.

 

[Ploki]

Sigh... You folks are very stubborn in your refusal to accept that turbo clocks are opportunistic and that different designs come with different balance of features. Your confusion probably stems from the fact that the Coffee Lake CPUs simply have such a wide performance range with their very high automatic overclocking ceilings.

If Apple made the MBP twice as thick and 3kg heavy (with power adapter) to make you happy, how many people do you think would buy it? Maybe Queen would, and maybe some other "pros" from these forums. That would be truly the end of the MBP brand and a sure way to kill the Mac for good.
[doublepost=1540329920][/doublepost]

That for sure. There seems to be a bug in Mojave's Safari right now, I also see bad flickering when watching youtube in fullscreen. Is that what you experience? Anyway, I hope they fix it soon, its beyond annoying.

Flickering happens when using "auto brightness" and calibrated profiles - its replicable on a clean system and I've been on the phone with apple support for three weeks already - fwiw i'm not yet frustrated, i understand they need to rule out every possibility before issuing a replacement.

Unrelated; im just salty because i already have 2 RMAs with equipment i bought with this computer... bought a black magic mouse (defective) paying 25€ for black color, CalDigit TS3+ also defective - been out of accessories since i bought it.

As far as CPU is concerned, I'm just a bit annoyed that 2.9 i9 under multi-core load cant squeeze an ounce more of performance than 2.2.