Dell XPS Speed Shift

Thanks for the guide. Can I restore it to default (disable speedshift) by typing setup_var 0xD8 0x0? In other words, is it reversible?

 

yes, you can set 0x0 to disable Speed Shift. It is fully working option, just Dell doesn't care about us to add it into bios interface. And they are assholes, because Speed Shift has zero requirements to bios or motherboard or machine vendor. It doesn't need any special implementation in UEFI code. It is fully hardware solution inside of CPU package. So it needs only one small marker from bios for windows to tell that it is supported in case if CPU supports it (and skylake cpu's supports it)

 

Thanks Goodwin,

Enabling SpeedShift in BIOS should permit one to adjust (the fantastic) EPP for max performance or max energy savings in the Windows "power options" via a simple dropdown box:

Control panel>hardware and sound>power options>[select a plan]>change plan settings>change advanced power settings>processor power management>processor energy performance preference policy.

If the "processor energy performance preference policy" is not viewable, go to the first page of this thread to enable (there are a few other dropdowns you can enable also). That is a dummy variable on my system (changing from 0% to 100% does nothing). But with your BIOS tweak, it might enable the dropdown. There is a huge difference in responsiveness between 0% and 100% so will be obvious in 2 seconds of internet surfing.

If you can't adjust EPP, from the Intel documentation, it seems OEMs would set EPP somewhere in the middle of the range. Regardless, you can always use ThrottleStop to enable SpeedShift and adjust EPP (note the EPP scale goes from 0-255 rather than 0%-100%).

FYI - I currently use ThrottleStop which requires that SpeedStep MUST BE enabled in BIOS. Since SpeedStep is laggy, I disable it via ThrottleStop (just uncheck the box). You can confirm that SpeedStep is disabled and SpeedShift is enabled via the free HWiNFO64 software (summary tab EIST=red and SST=green)

 

Thanks from me, too!
I just tested and it works without a problem. At first I had to disable UEFI secure boot. Despite an error regarding the size of the variable (it returns 0x00 instead of 0x0) everything is ok. The value I entered was 0x1 (which was reported back as 0x01). HW64 is green!

Thanks for your effort! So its kind of funny that Dell will refund my machine for the missing SpeedShift and now it works.

 

pressing, I tested the power-plan options you mentioned and it seems to work. If setting it to "100" and observing the multiplier with HW64, the CPU is stuck at around 800Mhz. Setting it to "0" the multipliers stuck around 3,2GHz with one core in Turbo. So setting the EPP by dropdown menu works.

(I edited this post, because I got pressing wrong).

 

I tried this, and it works, thanks!

However, the BIOS update to 1.2.16 failed with the error "Error accessing BIOS monothonic protocol" when I tried to flash the latest BIOS. I set it back to 0x0, ran BIOS update again and it worked.

 

Now that 1.2.16 is freshly out I tried it too. Got a bit confused due to the description being 'coarse', so to say. I used my Dell recovery USB stick, made from the original Dell Windows installation file, and replaced the efi file on it. Disabled Secure Boot in BIOS, and bootet straight into a command screen (EFI shell? I expected something different), with just saying 'grub'. No colors, nothing fancy. Typed in the commands, rebooted, and checked with hwinfo64 that SST is green.

Now I got a bit confused - is EIST/SpeedStep supposed to be still active or not? I disabled it in BIOS, but then Turbo is also disabled (shown by hwinfo64 in red), capping the performance by at least 10%, easily to see as the cores no longer reach peak CPU frequency. I used the old CPU Rightmark benchmark to verify that. So I enabled it again in BIOS - full performance is back.

I then added the new features to the advanced power plan settings. Yes, EPP indeed is controlled from there easily, and the changes can be observed in hwinfo64 in real-time. I have no clue what the time window setting is good for, and could not get it from trying values between 0 (default) and 10,000 µs.

Deactivating SpeedShift (the autonomous mode) - I have no clue if that works or not. Simply because I also did not find a way to measure/demonstrate any differences on my system so far. And hwinfo64 still says SST green when disabling it via the dropdown menu.

I wonder if the reason that Dell doesn't add SpeedShift is that Windows isn't ready for it. Seriously, Dell will not add a function that requires the user to add complicated command lines into an admin box to be able to control that feature. Those controls should automatically show up as soon as SST is detected.

Also the not noticable performance advantages in real-world usage are disappointing. Why add something where users will later claim it doesn't work because they see no difference?

I know that user pressing has an example where it improves performance. I can't reproduce the same with other music apps and interfaces. I got 32 samples over USB runing quite well, with some crackling at 100% CPU from an outside source. That did not change with SST, not at all.

 

If you are using high performance power plan with AC plugged in, there should be no performance difference because cpu frequency is already fixed at highest. Speed shift is more for balanced and power saver plans.

Also, if Dell supports speed shift natively in BIOS, you will not need to use any command line to enable it, right? You got wrong logic.


Sent from my iPad using Tapatalk

 

Read again. Enabling SST in Bios (what we did) must bring up the advanced power plan options in Windows automatically. Which is not the case but instead requires the command lines typed in found on page one (post 10) of this thread.

Also I tested with Balanced mode. Meanwhile I found other posts in other forums claiming the exact same thing. Got it working but not difference...

 

Sorry, I didn't know you meant showing EPP settings in power plan options.

Yes, you will need these command to show these values.

However, the three built-in power plans have their default EPP values, which means most users don't have to change it.



These values are hidden just like many other hidden ones are.



EDIT

The attached screenshot cannot show up. Below is the defaul EPP values for High Perf and Battery Saver power plans.

High perf

Code:

Power Setting GUID: 36687f9e-e3a5-4dbf-b1dc-15eb381c6863 (Processor energy performance preference policy)
      GUID Alias: PERFEPP
      Minimum Possible Setting: 0x00000000
      Maximum Possible Setting: 0x00000064
      Possible Settings increment: 0x00000001
      Possible Settings units: %
    Current AC Power Setting Index: 0x00000000
    Current DC Power Setting Index: 0x00000000

Power saver

Code:

    Power Setting GUID: 36687f9e-e3a5-4dbf-b1dc-15eb381c6863
    (Processor energy performance preference policy)
      GUID Alias: PERFEPP
      Minimum Possible Setting: 0x00000000
      Maximum Possible Setting: 0x00000064
      Possible Settings increment: 0x00000001
      Possible Settings units: %
    Current AC Power Setting Index: 0x0000003c
    Current DC Power Setting Index: 0x0000003c

 

I'm not an expert but I have read that both EIST & SST can be activated simultaneously and the processor & O/S sort out the processes. But EIST is really laggy.

Turbo requires SpeedStep and C3 (plus C6?) states to be enabled in BIOS. The work around I use is via ThrottleStop. In BIOS I have both SpeedStep and C-States enabled. Then in ThrottleStop, I disable laggy SpeedStep but top turbo speeds remain. In ThrottleStop I disable the very laggy C1E States. There I also enable SpeedShift. These settings are verified in HWiNFO64.

EDIT - To be clear, real-time audio performance improved when I disabled C-States in BIOS. But that disabled the turbo and the processor maxed out at 2.2GHz. By enabling C-States in BIOS but killing laggy C1E states in ThrottleStop, I got the best of both worlds: turbo performance and removed the problematic C-States.

I never found a way to disable SST without turning off the computer. That said, the large Intel PDF on the first post might give you some insight.

For improved performance in real-time audio, I also noticed a difference turning off anti-virus software. Not via settings which didn't help. But in W10 pro going to Local Group Policy editor and "enabling" "turn off windows defender". I also tried removing Windows Defender for another lightweight antivirus program but that did not help. Use caution without internet!

Thanks for the insightful summary arshcaria!

 

Well we have been told enabling speedshift in throttle stop can be done but turning off can't so I expect this is a reboot to disable job.

As for entries in the power plan I assume windows checks what is available during install and adds the features, maybe someone due to reinstall can confirm this.

 

That might be. But obviously then Dell still has a reason not to enable/add this feature, as the majority of existing customers do not get those controls.

 

pressing, thanks for the detailed explanation.

Yeah, Windows Defender is a terrible performance brake. I configured its settings to not scan many things, including a lot of programs where it multiplies the start time (compressed exe) significantly. One day I might go back to Avira, where I never had these problems for the last 6 years.

 

I disable defender by group policy or it is never totally turned off, it's worst than a bloody virus!

 

I just did a performance test with different EPP values.
The test is done via building an Android project using Android Studio and measuring the consumed time.
Building Android projects involves reading/writing/compiling a huge amount of files so it can properly test the burst performance of different EPP values of speed shift.

Code:

EPP       0           128        192        224         255
#1        4.614       5.04       7.697      11.615      13.657
#2        4.518       5.69       7.842      11.641      13.932
#3        4.366       5.006      7.679      12.067      13.735
#4        4.27        5.025      7.859      11.899      13.6
#5        4.239       4.915      7.708      11.92       14.023

avg        4.442      5.190      7.769      11.806      13.731

ratio      1          0.856      0.572      0.376       0.324

Time unit: seconds

Hope this can be a guideline of choosing desired EPP values with proper tradeoffs.

 

As expected, between 0 and 128 there is no so much difference, but not bad battery saving. After 128 more aggressive power saving begins.

 

Arshcaria,

That is an interesting set of statistics - thanks for taking the time to run them.

Did you run them with both SpeedShift & SpeedStep enabled? You can see that in HWiNFO64. There were some interesting related questions on the ThrottleStop Guide thread (I think). For example:

1. how do SpeedShift and SpeedStep interact? When SpeedShift runs, does SpeedStep get superseded, does it positively-neutrally-negatively impact performance?

2. How do performance of SpeedShift vs disabling C-States compare?

If you have some extra recreational time, maybe you could run a few quick independent scenarios testing:

1. Try disabling SpeedStep by unchecking that box in ThrottleStop
2. Try disabling C1E in ThrottleStop. Those C-States seem to cause a lot of performance issues
3. Try fully disabling C-States in BIOS

 

Not sure how SpeedShift and SpeedStep interact, a pure guess is that SpeedStep is overridden when SpeedShift is enabled.

In my test, setting SpeedShift EPP to 0 and disabling C-States give identical performance.

1. Once you have enabled SpeedShift, disabling/enabling SpeedStep in ThrottleStop won't make any difference to the CPU frequency.
2. Haven't tried that, maybe later I will.
3. Haven't tried that, maybe later I will.

 

This is quite interesting! For my real-time music app you can see above that I had some differences.

Let us know if you have the opportunity

Thanks!

 

Though general benchmark/programs cannot show the performance difference, some time-critical programs are very sensitive to the clock cycle. I had some issues when using an ARM processor simulator for it requires very precise clock cycles.

In theory, SpeedStep and SpeedShift should not have impact on the CPU clock cycles. But once I turn of SppedStep and SpeedShift, the simulator becomes normal. It is really weird.

 

New ThrottleStop software just released. Speed Shift upgrades include EPP monitoring real time.

I don't know any other software that provies RELIABLE EPP monitoring (powercfg and control panel provide wrong data on the 9550). See below for details and link...

 

I notice that when I enable speed shift, I get less performance in gaming. During gaming I notice the CPU is not ramping up as high, but staying around 1.2ghz. If I set the EPP at 1, CPU stays too high, and 128 (default) gets me less performance in gaming.

On the Surface Pro 4, EPP at 128 does not give less performance.

Anyone else test this on gaming performance?

 

If you run EPP= 0 does that work better than EPP=1?

Is the "EPP" indicator green in your ThrottleStop main window (or on HWiNFO 64)?

Regardless, if you are driving CPU in extreme situations, SpeedShift benefits will be reduced as CPU will be spending a lot of time in C0 states so less "shifting is required.. .

Disabling C-States in extreme situations might give the very best extreme performance. You can do that in the 9550's BIOS but I'm not sure that is optimal in such a thin laptop as the CPU may run hotter, providing less cooling capability for the GPU. Maybe not a safe option for CPU longevity. ThrottleStop does provide a checkbox to disable C1E states which I found to be hepful.

Also disabling C-States and/or SpeedShift capped CPU frequencies on my i5 (I just can't remember the combination that was problematic).

Intel guidance for laptops suggests EPP~128, which is a balance between performance and energy savings so performace with EPP=128 is probably too much different than a laptop without SpeedShift enabled.

 

Hmm, now that is astonishing. I updated two BIOS versions since I turned on SpeedShift, and HW Info still shows green. I would have expected that a BIOS update overrides those changes and resets SST to off...

 

Until you will do factory reset - SST will be enabled. In theory, bios factory reset should change it back to disabled state, but i wan't try it.

 

I think this is not correct with the Dell 9550.

On my 9550, restarting has always disabled SST (as long as I don't run ThrottleStop or have SpeedShift options disabled in ThrottleStop).

To verify, one can see that in the HWiNFO64 screen as the letters "SST" are red when SpeedShift is disabled and green when SpeedShift is enabled.

From ThrottleStop developer:

 

Of course it survives reboots - if you use the EFI shell to type in the 0x1 command, like I did.

 

That's true! You would need to take that out

 

Yes EPP is green. When running EPP=0, it's basically like running in High Performance mode, so any battery benefits will be negated. I'm still testing how batter life is on EPP=128. CPU ramps up fine during usual operations, but for some reason while playing a game like Overwatch, CPU doesn't ramp up that high.

Are others really benefiting in battery life and responsiveness from Speed Shift on our Dell XPS 15?

 

I created a thread on Reddit referencing this one, to figure out if this worked with the brand new XPS 9560 and Kaby Lake, but it doesn't seem to be working, alas.

Dell hasn't enabled speed shift on the XPS 9560 by default, as far as I can tell, and there's no settings in the BIOS in order to enable it. I wonder why - isn't it supposed to be further optimized with Kaby Lake? Why would they keep it disabled?

Edit: I should probably mention this in the SpeedThrottle thread instead though

 

Dell has not enabled SpeedShift via BIOS on the 9550, 9530, 9630 or 9660. The function is supported in the Intel chips and Windows 10.

Not quite sure why not. Maybe Dell is too busy with QC issues to test SpeedShift? SpeedShift runs fine on my laptop and I don't remember hearing about any performance issues.

It would be ideal to have Dell craft a BIOS switch for SpeedShift. That implies QC from Dell. Also should automatically enable SpeedShift after start or upon wake up.

I have not yet seen anyone document SpeedShift use on 9560.

9550 users, other Skylake computer users, and some Kaby Lake "U" computer users have documented several alternative ways to enable SpeedShift, including:

1. ThrottleStop software
- free, easy and stable
- Once SpeedShift is enabled, you can close out ThrottleStop

2. write to various MSR
- a bit more dangerous
- I don't know if the registers and/or coding changed for Kaby Lake
- bad case scenario - Skylake MSR instructions included a few pages back could brick a 9560

When using these two alternatives, SpeedShift needs to be reenabled every time a computer is turned on or wakes from sleep...

Finally, there apparently are other ways to launch SpeedShift but they just create dummy variables that the XPS seems to ignore (e.g. powercfg)

I would be daring enough to try ThrottleStop on the 9560 to see if SpeedShift works. I would not try writing to MSR without expert guidance...

 

Thanks for the information. I've only been able to persuade one person so far to try ThrottleStop together with the XPS9560, and he said that he was unable to toggle it. Not sure if nothing happened when he ticked the box, or if he was unable to do even that, as he wouldn't give any more information.

I won't receive my unit until the 9th, so I'm out of luck trying myself.

 

Odd, speedshift is enabled by default on my wife's 9350 with the latest bios (no idea if it was before that bios).

 

It isn't odd, just enabled on the XPS13 because Dell are you know, wankers.

 

lol, that's true. I figured they would do it on the 9560 since it's mo betta. Gotta love Dell.

 

Does she have SpeedShift toggle in the 9350 BIOS? Maybe some other new toggles appeared as well?

 

I haven't looked in the bios for the toggle, but I know throttlestop shows it enabled by default.

 

Does she by chance have "enable speed shift when throttlestop starts" box checked (in "Turbo Power Limits" window)? That could be activating SST when ThrottleStop is opened. Interestingly, SST will stay activated until the computer is turned off or goes to sleep.

One way to test it is a BIOS change is to restart computer without launching ThrottleStop. Then launch HWiNFO64 (summary only) and look for the SST indicator. Red = disabled. Green = enabled.

 

No the option is greyed out, I do know how to enable speedshift on ThrottleStop. She is skiing so I'll have to check in the bios when she comes home tonight.

 

Thanks custom90GT.

I forgot to mention, ThrottleStop seems to have a few "on" switches. I think if any are checked, ThrottleStop will enable SST:

- "enable speed shift when throttlestop starts" box (in "Turbo Power Limits" window)
- "speed shift" box (in "Turbo Power Limits" window)
- "speed shift-EPP" box (in "main" window)

 

Right, you forget I've got my 9550 that I had to enable SST on, I didn't have to do anything on the wife's 9350. I'll also check hwinfo to see what it says for SST.

 

EDIT- Above it looks like Uncle Webb is noting some updated BIOS are being released that activate SST automatically behind the scenes. That is, there is no user toggle in BIOS screens. That may be a recent change with the 9350. Not currently the case with the 9550.

Also see that EPP=128 (which Intel seems to recommend as a midpoint between Performance & Efficiency) seems to cap CPU max performance on UncleWebb's Kaby Lake i5-7600 (FYI not a Dell XPS). Uncle Webb finds EPP=80 is a better "balance" on that chip as it allows CPU to reach max performance. I use EPP=0 for max performance. YMMV.

 

Hi,

after getting a refund for my 9550 I ordered a 9560 which is to arrive tomorrow

I wonder if SpeedShift really is still disabled. I think this is not quite clear, as FluffyDroid mentioned about reddit...
At my 9550 I enabled it by setting the bit via UefiShell and I would like to do it again on the 9560 But I gotta have some Information, if its still the same bit to avoid bricking it. Do any of you guys know, where to find this?

 

Rockstar - I trust you to give us a definitive answer, as you did the UEFI hack on the 9550 and understand SST.

Until the "pros" do some UEFI hack testing, you can easily enable SST via ThrottleStop.

Also, if you have a moment, please confirm here (and at the ThrottleStop thread) if ThrottleStop indeed properly enables SST & EPP on your 9560.

 

It must have been windows updates or something that enabled speedshift on my wife's 9350. There is nothing in the bios, but it SST is enabled even without throttle stop running. Odd that the update didn't enable it on my 9550, I just did a fresh install last weekend on it too...

 

I've been trying to figure out which update it was, if there's some documentation somewhere, but can't really find anything about it.

Seems strange. So Speed Shift hasn't been enabled on laptops prior to this update? Or is the update in question specific to certain OEM:s?

 

No idea sadly. I just bought her 9350 last week and I did a fresh install on it before even boot into the factory install. I figured when I had checked throttle stop that SST was on because dell enabled it in the bios, I didn't even give it a second thought.

 

Speed shift has to be enabled in the BIOS for windows to make use of it so no amount of software update will work on the 9550/9560 unless the BIOS has it enabled.

 

The fact that HWinfo shows SST as being enabled means nothing? What about throttle stop? Adjusting EPP values shows differences while running applications, I find it hard to believe that nothing is happening...