Octane VI / Clevo p775tm1-g | i9 9900k | Processor review for gaming and everyday use

debiruman665

Enthusiast
I have some of this information already in the "Show off" thread but since I'd like to talk more specifically on the topic of optimising the 9900k for different tasks which are moving further away from the topic I started and I don't want to be raising my old review everytime I want to update with my recent findings.

Preface

The 9900k at stock settings runs at very high temps, even when idle and doubly true when gaming. When first out of the box and hitting the default settings button on the Clevo control centre the processor will clock at around the 4900MHz range which is about as high as you'll get on this without doing something potentially dangerous. After clicking the default settings on Clevo control the speed will drop down to about 4600MHz.

Undervolting

-120mV

If getting the highest benchmark score is your goal then I've managed to get my undervolt down all the way to -120mV which was the absolute extreme. You can perhaps pop off a benchmark and upload for bragging reasons but this isn't really stable for too long.

-100mV

This mostly stable undervolt works for non CPU intensive everyday use fairly well gaming but this is actually unstable once you start playing games and you might only get about 20 minutes before your computer freezes if you try and play any CPU intensive strategy games.

-80mV

This is where I currently have mine at and seems pretty stable. Starting from -100mV I jump down to -80mV impatiently after a crash during gaming. There might be some more headroom but I haven't really tested it over the long period of time.


Clock Speed

I would say this area is subject to a lot of debate but here are my two cents.

Base speed 3.6GHz

To easily turn Turbo boost off without having to tinker with any overclocking you can create a power profile on your Windows computer and set the maximum processor state to 99%. This is a weird way of doing this I admit but I find it easier to switch between power profiles rather than changing the setting on XTU, BIOS or your personal overclocking software of choice.

Useful for: Plugged in everyday use, non-cpu intensive gaming

Clevo recommended setting 4.6GHz

The 4.6 is a bit misleading as this speed that shows up on your clock speed reads is actually the slowest core. Clevo recommends that the clock speeds gradually gradiate down from 50x down to 46x down the 8 cores allowing for faster single core and slower clock speed when doing multithreaded tasks. Using this setting will make your processor hot so expect a lot of fan noise. Not recommended for everyday use as you will suffer from louder fans for no perceivable benefit if you are just browsing the web. Jump back to base clock speed by doing the max processor state trick mentioned in the previous section. Use this mode if you want to host large 4x games you'll rip through the AI section of the turns in record speeds.

Usefol for: intensive use CPU-intensive gaming,​

Safely overclocked (kinda) 4.9GHz

By setting all the core multipliers to 50x the CPU will run at around 4.9GHz constantly. While this does stroke your ego and makes you feel like you're getting your monies worth the back of your laptop will be a constant hairdryer while you are gaming, even non CPU intensive games. When playing Metro Exodus or Anthem the RTX 2080 was comfortably around 70C even at max settings while the processor was hitting close to 100C even at max fans. It's a whole lot of extra heat of a barely minimal performance increase compared to the previous 4.6 setting which is only a 6.12% increase for around an extra 10C of heat on the processor. It was nice to see the higher clock speed and the result of many many hours of tinkering to get my processor going as fast as the laptop could handle but after a few weeks at this speed, I've personally gone back to 4.6GHz purely for the heat.

Underclocked 800Mhz-3.6Ghz

Without the ability to turn off the GPU some battery life can be gained by underclocking the processor. An easy way to do this is to just lower the maximum processor state in the power setting. You could set this to happen automatically when the computer is on battery. If you need up to 2 hours of battery life on a text editor (2.5 hours on a fresh new battery) on a somewhat long train journey then this is highly recommended if you want any longevity. I would propose that the i9 running at 900MHz feels just as snappy as an i5 with a similar power draw while its on turbo. I've even accidentally booted up a non CPU intensive game while I was on this power plan and the game was still playable if not with less framerate only to discover I was running at 900MHz.

useful for: when you need a laptop and not a portable pc, or want an absolutely silent machine. The heat sink can soak up a lot of juice before ever needing to kick in the fans.


Apologies if some things I wrote don't make any sense, although I am a native English speaker sometimes my brain does not connect well to my fingers what I want to portray and I end up writing the wrong words or phrases. It's a bit like having autocorrect but it happens at the hardware level
 

SpyderTracks

We love you Ukraine
Worth adding that any undervolt of overclock is silicon dependent so what works for one person may not work for another.

But excellent breakdown and very informative, rep given!
 

Scott

Behold The Ford Mondeo
Moderator
Great post and fantastic information. From what you have wrote I believe that everything you have stated is the settings that you have used.... rather than the actual output values from the CPU?

The reason I mention this is, I believe, at full chat, the 9900k, in the Clevo chassis you will rarely manage to hold anything over 4.2ghz across all cores for any more than 30 seconds at a time?

Have you ran any benchmarks at your various settings to see what the peak hold & tail off is of the clock frequency during them? When I first got my 6700k Octane I happily managed a 4.8ghz overclocked. I was delighted..... until I noticed that it would drop to 3999mhz after 20 seconds :D
 

debiruman665

Enthusiast
Great post and fantastic information. From what you have wrote I believe that everything you have stated is the settings that you have used.... rather than the actual output values from the CPU?

The reason I mention this is, I believe, at full chat, the 9900k, in the Clevo chassis you will rarely manage to hold anything over 4.2ghz across all cores for any more than 30 seconds at a time?

Have you ran any benchmarks at your various settings to see what the peak hold & tail off is of the clock frequency during them? When I first got my 6700k Octane I happily managed a 4.8ghz overclocked. I was delighted..... until I noticed that it would drop to 3999mhz after 20 seconds

Good question, the speed I'm referring to is the one that you'll see advertised in the Task Manager Performance pane or the Clevo Control center which would be the normal clock speed when its not under load and throttling in any way.

From my tinkering in XTU its always current throttling or power throttling once you have the undervolt applied. I'm not brave enough to go into the bios and increase these though.

I decided not to use benchmark as example as they don't really replicate real life examples.

Got a benchmarking tool of choice? and I'll run it on my profiles and ammend the information.
 

Scott

Behold The Ford Mondeo
Moderator
I tend to just run firestrike personally. The generic one. Purely because it allows me to compare current gen to older gen. It's reasonable for checking actual gaming performance too.

All the benchmarks with 3DMark are quite handy though, useful tool.
 

debiruman665

Enthusiast
Just an update that since the scotster mentioned the cpu does not stay at 5Ghz I can confirm it only manages around 4.2GHz under full load. I need more time to keep tweaking my settings. Even when at Clevo recomended settings when playing games on ultra the processor sits at around 90-100C and constantly throttles.

I'm currently working on getting a better processor ratio sorted in order to try and get the most and an optimal max temperature
 

Scott

Behold The Ford Mondeo
Moderator
This is the grind with a high end laptop unfortunately but very worthwhile IMO.

It will be great info for other users though so definitely appreciated.
 

debiruman665

Enthusiast
This is the grind with a high end laptop unfortunately but very worthwhile IMO.

It will be great info for other users though so definitely appreciated.

There are a lot of pains involved. Clevo control and XTU seem to fight over the control, yet... XTU wont work without clevo control working. Its a lot of saving profiles, and restarting involved.


I also checked out the obsidian recommended settings from their control panel. Which is quite different and seems to value low temps, somewhat lowest boost and avoiding the throttling at all cost.

The CPU multipler ranges from 41 to 50 on the obsidian setup with maxed out power draw and short term boost settings. Since the processor seems to cap out at around 4200 at max load this seems like it might give slightly better consistent performance at the expense of the boost.

Benchmarks:

Seem to put them at around the same score range in XTU with around a 50 point margin of change which causes them to sometimes be better than each other but I think this is due to cpu temps never starting the same/

Obsidians settings don't perform as well with cinebench with clevos settings performing better.

Obsidians firestrike actually got the second best score I ever had coming close to my own overclocked setting which would run at around 100C.

While gaming Obsidians settings get more sensible temps also. I'm trying to convince my wife to help me with her maths skills to figure out the sweet spot because adding 500HZ onto speed can cause drastic temp changes so I'd like to science this one a bit better

attached is the screenshot from XTU for the obsidian settings on the right, this'll save anyone needing to install their software to get their settings. For some reason its hard to get the 96 second turbo boost time window to stick after a restart which I think might be clevo control flicking it back when it boots.
 

Attachments

  • Capture.PNG
    Capture.PNG
    32.7 KB · Views: 786

ubuysa

The BSOD Doctor
I'm wondering whether what you're seeing at the limits here is a chaotic relationship rather than a linear one?
 

debiruman665

Enthusiast
Update on the obsidian settings.

Was getting bluescreens "kmode exception not handled", I tried bringing the undervolt down to 50 bit was still periodically getting them. I'm not entirely sure but it seems like after a bluescreen the short term power limit time and power limits were being changed so I suspect it was something to do with them. I'm going to try with no undervolt for another week while before abandoning completely.

Overall: Much better thermals and fan noise | worse stability (random crashes even using web browser, not just gaming)



Although this is an alienware laptop when he is setting up the XTU i get the exact same problems with the 9900k. Asking for reboot then not applying the changes etc.
 

debiruman665

Enthusiast
I've had a eureka moment when testing the core multipliers and I think I have managed to figure out some of the voodoo behind it.
I was up to 3 am last night doing these tests and I'm having to summarise by memory at the moment because it still needs a bit more tweaking.

In order to get a single thread stress test to get the cpu to run at 5GHz for a single thread, I had to set the first 6 out of 8 cores to x50. I assume this is because the single thread does not always live on the same core and it seems to get moved around to keep the temps down on each individual package.

In order to test this, I set all multipliers down to 36 and ran a single threaded stress test. I incremented each core up to x50 one by one and the speed gradually started to increase till the sweet spot of 6 cores at 50x.

So at this point, I had my multipliers as so:

x50
x50
x50
x50
x50
x50
x36
x36

unless I had the -80mV undervolt the single thread could not reach the 5Ghz without current throttling. So my single threaded stress was stable and non-throttling.

I then started incrementing the thread count until the frequency dropped back down to 3.6Ghz to where I would increase the multiplier up to where I started getting throttling.

At a certain point I had my multipliers as such:

x50
x50
x50
x50
x50
x50
x46
x36

and I booted up anthem and the game was running on all cores at 3.6Ghz at around 80C and my framerates were in the 70-100 region instead of the 50-60 when all cores were running at 4.2Ghz throttling.

After increasing all cores using the previously described method I finally ended on

x50
x50
x50
x50
x50
x50
x46
x43


Some observations:

Only the last three multipliers actually make any noticeable changes.

Anything over 8 threads is controlled by the last multiplier.

The laptops idle clock speed under no load will be at the 8th multiplier


When running XTU's benchmark I was still getting some throttling and the last multiplier of x43 had the best CPU only performance for 8+ threads, it's still too hot for gaming because once the GPU is on it will start getting hotter.

I was feeling brave since the single thread stress tests weren't near any throttling and I managed to get a single thread to overclock peaking around 5.1-2GHz if setting the first 6 multipliers to x52,

x53 caused an immediate reboot and I figured I would leave it alone at x50 for now while I get the rest of it stable.

I feel like this is pretty significant information as I haven't read anything like this online. If anyone else is brave enough to test their own CPU concerning the multipliers and fixed thread stress tests to confirm my findings I would deeply appreciate.
 
Top