intel Replicating Intel's "200S Boost" in laptops

[SACbomber]

For those unfamiliar: https://game.intel.com/us/stories/200s-boost-overclocking-profile/

 

I have an 285HX and XTU lets me increase "SA fabric ratio" which I think is uncore/NGU. Didn't see a D2D option in XTU so that might be in my BIOS. My RAM is 6400 MHz(MT/s) CSODIMM and I wonder how overclockable they are.

 

Has anyone tried this?

Edited Wednesday at 01:25 AM by SACbomber
Adjusted title to reflect ongoing changes

[SACbomber]

 

Increased SA fabric ratio from the default 26x to 32x (for 3.2 GHz) on the default voltage (0.910 V). XTU stress test crashed the entire laptop (it automatically rebooted) immediately.

 

Increased SA fabric voltage offset to +0.025 V, for a total of 0.935 V since I saw that number somewhere. XTU stress test crashed the XTU program itself about 4 minutes into the stress test, but the laptop didn't reboot and remained responsive. The background process responsible for the actual stress testing (which is linpack_xeon64.exe) kept on running and saturating the CPU. Relaunched XTU and surprisingly, it didn't throw me a "This program exited unexpectedly" message. It did not recognize that a stress test is underway, however.

 

Figured I was on the knife edge of stability. Terminated linpack_xeon64.exe, lowered SA fabric ratio to 30x, stress tested again, everything passed, no crashes. Performed the Power and CPU tests in OCCT, everything passed. During all this time I had -50 mV on all cores and cache.

 

D2D ratio is available for tweaking in the BIOS. Haven't touched it yet. Going from 21x to 32x should provide tangible improvements but I have a sneaky feeling that performance improvements after applying 200S Boost come from RAM overclocking, not NGU/D2D overclocking. Intel's "big.LITTLE" chips are known to be sensitive to latency (RAM speed).

 

A reminder for those of you following along at home: Don't copy my values without performing your own battery of tests. Maybe you've won the silicon lottery and can push further. Maybe I've won the silicon lottery and my values will crash your system immediately.

 

More to come.

[SACbomber]

Set D2D to 30 (for 3.0 GHz) in the BIOS. Passed stress testing. Almost too easy. Will try pushing it further later.

 

Saw a reduction in RAM latency in AIDA64: 

(Ignore the Memory Bus value. It's reporting the instantaneous value right as the test began.)

 

Compare with stock D2D and NGU values:

 

A nice 13 ns reduction.

[win32asmguy]

I tried your settings for SA and D2D and they helped a little. Also set cache to 40x. You might try disabling SAGV + setting Gear 2 to see if that helps. It might hurt battery life though as memory clocks remain at the set frequency.

 

[SACbomber]

On 5/14/2025 at 4:05 PM, win32asmguy said:

I tried your settings for SA and D2D and they helped a little. Also set cache to 40x. You might try disabling SAGV + setting Gear 2 to see if that helps. It might hurt battery life though as memory clocks remain at the set frequency.

Are you undervolting at all? Think I'm OK with RAM downclocking. Also that's not much of an improvement. 98 ns down to 94, right? Might be because my RAM has such loose timings.

[win32asmguy]

On 5/15/2025 at 3:05 PM, SACbomber said:

Are you undervolting at all? Think I'm OK with RAM downclocking. Also that's not much of an improvement. 98 ns down to 94, right? Might be because my RAM has such loose timings.

 

If you intend to get that Gskill XMP kit once available you may have no choice but to disable SAGV as it usually increases stability (at least that was the case with raptor lake). Supposedly gear 2 is good up to 8400MT/s on the 285K. I could not get the Kingston Fury sticks to boot at 7200MT, probably due to the 1.25v limit on this bios version.