2018 MBP i9: Testing w/ Volta & Fan Control

[EugW]

So here is what I observed with the YES X 12 test

This is Default with no Volta or FAN control:
View attachment 772187

Wow, almost 80 Watts with this YES x 12 test. It immediately jumped to 100C.

Your so-called 45 Watt TDP mobile i9 looks better suited for an iMac. Ironically, my so-called 65 Watt TDP desktop i5 looks pretty good for your MacBook Pro, as it hovers below 45 Watts maxed out at 3.9 GHz.

 

[ofarlig]

Either way, it would be useful to see the results of a YES x 12 test with Intel Power Gadget. My guess is it would throttle at least as hard as the videos have shown with other applications, or perhaps would throttle even harder.

Yeah I agree, I could be completely wrong but that is just how I understood it.

 

[Queen6]

I guess this means that it actually does seomthing on your computer?

As I said, using intel power gadget, I confirmed that Volta didn’t actually make any voltage changes at all. When I showed the power gadget logs to the developer, he stopped responding to my emails

Apple likely locked the VID making undervolting impossible, unless you're handy with the firmware & Intel's CPU registers...

Q-6

 

[Aea]

I make the distinction because even if you modify a 2018 i9 MBP by putting better Thermal Paste, it won't improve your performance because the VRM is being maxed out before the heatsink/fan/TIM setup

I can't comment since I'm very much on the fence about trying this, but there are some reports saying it does improve performance to re-TIM.

Do you think I should always use Volta or just when performing the most demanding tasks?

Also, have you seen this other software VoltageShift ? Probably with it we can limit the power above the 45w volta threshold...

https://sitechprog.blogspot.com/2017/06/voltageshift.html?m=1

I am keeping it on all the time personally. You're still getting a turbo boost (except for it's ~3.5 Ghz vs 4.8Ghz) but the duration is so short I don't think it'll make any noticeable performance difference. This may differ in your workloads but the machine feels very stable / snappy with the 45W limit.

I have not tried VoltageShift yet.

Oh, don't misunderstand me, I will gladly pay for good quality apps, but I made a habit not to use system control apps like these if I can't review the code.

There are FOSS alternatives if that's a concern (see the quote above).

The downside of Volta is, that it will also slow down daily applications that only need burst speed for a short amount of time.

Not sure this will be noticeable, but it's hard to objectively test this. Even with the Power Limit it will turbo, but not as high as stock. The question is going to be whether 1-5s of 4.8 GHz vs 3.6 GHz makes a perceptible difference? Maybe if you have extremely bursty workloads it will.

I propose that Apple fire the current person handling whatever caused this issue and hire you.

I'm sure they have a room of people much more knowledgeable than me working on this issue.

I wasn't aware the CPUs could use more than their TDP. If that's the case, can the i9 periodically draw much more than previous 45 W i7s? If so, maybe Apple (and other manufacturers) haven't fully accounted for this in the design.

This shouldn't really come as a surprise to Apple or any other systems integrator for that matter.

The T2 chip is (also) responsible for Power and Thermal Management.

Running its own OS (eOS) it should be possible to implement quite complex power profiles, doing its work no matter if you are running MacOS or Windows... Maybe they released it before it was finished. But it can certainly be updated.

It seems very strange to me that they would not implement any limits whatsoever and just let the CPU do its thing, since Apple is usually all about control.

Here's hoping.

Those Volta-assisted scores are a touch higher than all of Aea's i9 scores. All of them.

Silicon Lottery I guess. This particular machine even idles hot so maybe I got a chip that barely binned i9.

I guess this means that it actually does seomthing on your computer?

As I said, using intel power gadget, I confirmed that Volta didn’t actually make any voltage changes at all. When I showed the power gadget logs to the developer, he stopped responding to my emails

What machine? Did you change the SIP properly? I could verify that power limiting worked. The undervolting testing I did was all over the place however, no statistical significance. I'm unsure if Volta currently supports undervolting on this chip.

I had time now actually. It isn't in some nice graph like the others but here are my results with the same methodology.

CB Scores on 2.2 base 15":
982
982
980
977
983
981
979
981
982
978

Average: 980.6

Nice scores. Looks like you might have saved $500 on the CPU for better performance.

So it seems Apple designed the cooling system for a TDP of 45 Watts, and the system actually works decently well for a TDP for 45 Watts.

However these chips use way more than 45 Watts for sustained loads. I could understand if it was with power viruses or something, but these results are with actual and common applications, what people might actually use their MacBook Pros for in real life.

IOW, it looks like Apple took the MBP chassis, and essentially just dropped new 45 W TDP chips in there... except some of the higher end chips really shouldn't be rated as 45 W TDP.

The decision to let the i9 run above 45W is in Apple's hands.

I want to see one more column of test with factory settings but in an extremely cool environment (ie what if chasis could properly dissipate all heat), to see whats the most performance one could get.

If I had a walk in freezer I'd help test that

 

[Queen6]

Wow, almost 80 Watts with this YES x 12 test. It immediately jumped to 100C.

Your so-called 45 Watt TDP mobile i9 looks better suited for an iMac. Ironically, my so-called 65 Watt TDP desktop i5 looks pretty good for your MacBook Pro, as it hovers below 45 Watts maxed out at 3.9 GHz.

Your confusing Intel's definition of TDP and power draw 80W is to be expected given the performance these CPU's can deliver in the right chassis.

i9 can pull over 150W and put many a desktop to shame, the rest is on Apple...

Q-6

 

[EugW]

Your confusing Intel's definition of TDP and power draw 80W is to be expected given the performance these CPU's can deliver in the right chassis.

i9 can pull over 150W and put many a desktop to shame, the rest is on Apple...

Q-6

I’m not confusing TDP vs power draw at all. The point is the TDP is simply too low. YES x 12 is not real world but it’s far too easy to far exceed the the TDP with common real world workloads. Furthermore, judging by the tests the MacBook Pro can handle 45 W workloads just fine.

This is not inherently an Apple design problem. This is an Intel 8950HK-45W-TDP-is-bullcrap problem.

 

[Nozuka]

But it was Apple who tested it and decided to ship it like this anyway...

 

[Queen6]

I’m not confusing TDP vs power draw at all. The point is the TDP is simply too low. YES x 12 is not real world but it’s far too easy to far exceed the the TDP with common real world workloads. Furthermore, judging by the tests the MacBook Pro can handle 45 W workloads just fine.

This is not inherently an Apple design problem. This is an Intel 8950HK-45W-TDP-is-bullcrap problem.

The TDP & power limits will be documented in Intel's data sheets which will be made available to all OEM's with engineering samples well in advance.

Same will be made publicly available in time. Apple simply either got it's maths wrong or doesn't care...

I own a notebook with the 8750H and it has zero throttling tendencies pulling close to 90W under max load. Then again the OEM clearly spent more time with Intel's specifications, than it's sales & marketing department....

Intel clearly state the TDP derived from base frequency, the rest is in the data sheets. People want thin & light, people want big number's you got it, pity the performance is so clearly compromised.

Q-6

 

[hajime]

I completely agree, but I think it would have been less of a loss if they had limited the power before hand instead of having people being disappointed with the performance after getting the MBP
[doublepost=1532288216][/doublepost]
Yah, and it should not have been so. I do not know why, but the quality control at apple seems to be getting worse

Perhaps it is Apple's way of making us "part" of the Apple family. Users are now involved in testing and finding solutions for their products.

We also help with PR. Now the i9 issue is big. Later somebody find a solution (if lucky) and make the MBP2018 great. Then, more people buy it because of big publicity.

 

[Aea]

But it was Apple who tested it and decided to ship it like this anyway...

 

[afir93]

The downside of Volta is, that it will also slow down daily applications that only need burst speed for a short amount of time.

Yeah and it seems you also have to keep System Integrity Protection disabled permanently in order to use Volta which is another big downside IMO.

I have no problem disabling SIP for quickly installing an app and then re-enabling it afterwards, but permanently disabling such a basic security feature in order to use this is just not great. Of course this is not the developer's fault, it's just how it is – if we want to get the really tempting additional performance from setting a power limit then we have to give up some of our Mac's security.

 

[CodeJoy]

Same will be made publicly available in time. Apple simply either got it's maths wrong or doesn't care...

It could also be that they were faced with only undesirable options. Redesign and lose on the economy of scale in manufacturing. Or not redesign and lose performance. They may simply have -- again -- painted themselves into the thermal and power corner.

Ultimately, they end up in this situation because of poor judgement and poor design decisions. Not because of maths. Not because of engineering skills. There are management issues fully and completely. Or rather -- these look like poor decisions from a customer perspective, but given that Apple is printing money like nobody else, they may be fully competent decisions. They're just not good engineering decisions, and not good for customers.

 

[parsonsmike]

[/QUOTE]

The undervolting testing I did was all over the place however, no statistical significance. I'm unsure if Volta currently supports undervolting on this chip.

Yes, I followed the install directions, which included disabling SIP

How did you get undervolting working, if not with Volta or voltageshift?

 

[ESA]

Is there now good statistics as you had with the 2.2 or 2.6 ?

I saw with the 2.2 it got the same results under 45W.
Hope we can test with more W soon.

 

[karanlyons]

I’m not convinced that you can undervolt this CPU at all, at least currently. Intel XTU has all voltage parameters grayed out, and I don’t think Volta is actually undervolting either.

 

[Aea]

I’m not convinced that you can undervolt this CPU at all, at least currently. Intel XTU has all voltage parameters grayed out, and I don’t think Volta is actually undervolting either.

Volta almost definitely isn't, 99% certain.

 

[karanlyons]

Yeah, I can also confirm that you can’t write to the OC mailbox at all. So it looks as if undervolting is not going to be possible, which is unsurprising but also pretty ****** given the potential upside it would’ve given.

 

[Sparkel]

Yeah and it seems you also have to keep System Integrity Protection disabled permanently in order to use Volta which is another big downside IMO.

I have no problem disabling SIP for quickly installing an app and then re-enabling it afterwards, but permanently disabling such a basic security feature in order to use this is just not great. Of course this is not the developer's fault, it's just how it is – if we want to get the really tempting additional performance from setting a power limit then we have to give up some of our Mac's security.

I Just removed Volta for this reason. Anybody know how to enable system integrity protection again?

 

[karanlyons]

I Just removed Volta for this reason. Anybody know how to enable system integrity protection again?

`csrutil enable` whilst in recovery mode. But note that the only portion you had to disable was kernel extensions.

 

[The Mercurian]

@Aea check this guys solution out it seems better than Volta:
https://forums.macrumors.com/thread...-explanation-of-apples-engineering-f.2128936/

 

[maflynn]

I'd love to know how you got the voltage control working. I followed the directions on the developer's website, and I can't make adjustments to the voltage.

I didn't say I got the voltage changed, I said that I adjusted the wattage and that keeps the heat down.

 

[M.Rizk]

@Aea Volta got updated and allows up to 90W now btw if you feel like squeezing some extra performance

 

[parsonsmike]

I didn't say I got the voltage changed, I said that I adjusted the wattage and that keeps the heat down.

You said "I've confirmed its working" in response to my post saying "I confirmed that Volta didn’t actually make any voltage changes at all."

So you could see how, following standard forms of logic and the English language, I might have been confused.

 

[maflynn]

You said "I've confirmed its working" in response to my post saying "I confirmed that Volta didn’t actually make any voltage changes at all."

So you could see how, following standard forms of logic and the English language, I might have been confused.

I think we were talking past each other. You said it wasn't working and a waste of money, I responded it does work. I should have been more clear, that I was referencing the power limit setting, not the undervolting, but if you use that the power limit portion of the app, it does work. Whether you agree with me or not on the usefulness is up to personal preference.

 

[shavou]

@Aea any chance you can do some tests with the new version of Volta which allows to get to a 90W power limit?

My MBP i9 should be delivered this Thursday and I'd love to know whether I should keep it or not....

Thanks a lot for the hard work!