Announcement

Collapse
No announcement yet.

Intel Baseline Profile Yields Odd Power/Performance On Linux

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by sleeksorrow View Post
    I recently read some german news article about that topic that was mentioning something about mainboard default normally (among many other things) does a lot of undervolting to reach the same performance while using less energy. Under high load the affected CPUs then get unstable.

    I don't pretend that I fully understood that but I know your german is quite good. Perhaps you can verify that explanation yourself.
    On a side note, I saw this as a good test of the built-in newer Firefox translation stuff (built-in, meaning doesn't translate on a server external to you Firefox instance.) Have not read yet, but looks like it worked great.

    Comment


    • #12
      At least it should be stable now. The 7950xs ebay prices increased since

      Comment


      • #13
        Originally posted by archkde View Post

        Given that the baseline profile is the one actually from Intel, more like OEMs wanting to make their Intel offerings not look worse than the AMD variants.
        Given that Intel never forced motherboard manufacturers hands until now, we can say they were tacitly approving of what motherboard manufacturers were doing (because it made their products look better).

        Or at least, Intel turned a blind eye to what manufacturers were doing in their BIOSes. However, I can't see that being the case. They had to have known once public reviews were posted with numbers that didn't match up with their own "Default Profile" numbers.

        Comment


        • #14
          IgorsLab has 2 articles about this problem. ASUS is actually not the Intel baseline spec but Gigabyte does. ASUS uses the performance baseline according to the docs Igor shows.

          I have not yet commented on the now ubiquitous investigations and explanations of the problems with Intel's 13th and 14th generation K-CPUs, because pure specu




          Very interesting reading.
          Last edited by DRanged; 01 May 2024, 11:34 AM.

          Comment


          • #15
            Originally posted by raystriker View Post
            Intel's last 5 years summarized in one sentence- "Oh, how the mighty have fallen"
            Still no good ITX boards for AM5. Also no good enough APU with powerful CPU (7900-7950X3D) and 3-4 monitor support. I hope AMD will release some new chipset with better IO and better APUs before intel's 15 gen.

            Comment


            • #16
              Originally posted by mlau View Post
              You should also blame the mainboard makers: they all want to look good in benchmarks (best perf at lowest power consumption) so they cut corners wherever possible, and as long as windows boots to a desktop, they say "ship it"...
              Intel has a lot of say in which parameters they can tune and which they can't. Also, if they broke Intel's rules, Intel could cut them off as partners. Remember that these board makers sign many agreements with Intel and are dead in the water, without support from Intel. So, Intel holds all of the power in this relationship.

              Originally posted by archkde View Post
              Given that the baseline profile is the one actually from Intel, more like OEMs wanting to make their Intel offerings not look worse than the AMD variants.
              But Intel is the one who decided how much​ slack to give them. Intel is the main beneficiary of their "creativity". The motherboard makers don't particularly care about the Intel vs. AMD race, since most of them sell both Intel and AMD boards. Their main concern is just competition with other board makers.
              Last edited by coder; 01 May 2024, 02:23 PM.

              Comment


              • #17
                This whole "Intel default" thing is the reason why on the MSI PRO Z690-A Coreboot/Dasharo performs lower than MSI default BIOS, and was even documented. Is funny how now this is coming full circle.
                Lower power efficiency can be because higher AC_LL values (Which most likely the Intel baseline profile uses) increases voltage under full load thus it consumes more power than if it was tuned. At a worst case scenario on certain SKUs like my 12600K (Which MSI defaults to 80/80, whereas Dasharo uses 170/170) the power consumption difference was a whoping 23W on full load which actually makes it go above PL1 limits so sustained clock speed (After PL2 Tau time) tended to be lower, too. How amusing that now everyone will be facing these differences.
                Last edited by zir_blazer; 01 May 2024, 02:21 PM.

                Comment


                • #18
                  Hmm...interesting. This, plus the AnandTech article on same, makes me think the change has caused more cores to remain at turbo to complete the work. More overall heat and power, but it's spread 'safely' across some cores, when the issue seemed before to be running one core too long at turbo.

                  Comment


                  • #19
                    The good thing to do is stay away from Intel shit for a while... not like me, that recently purchased a i5-12600k and immediately regretted the purchase because Intel declared that socket 1700 is dead already There is no limit to the dirtiness Intel can immerse into

                    Comment


                    • #20
                      Disable XMP and retest. Afaik XMP forces some overclocking features that might not mesh well with the Baseline profile while the MB Default might be tuned to work with XMP better ...

                      Comment

                      Working...
                      X