yosarianilives

Assuming his pretesting doesn't lead to later massive scores lol

Ycruncher is really heavy on the memory system, so depending on cpu could be too fast mem settings, too high fclk, too high cache. Only sometimes is it too high of core if you can't run off the bat. Stability in any other bench is irrelevant for ycruncher stability

If you are worried a bench is not worth pts then the best thing to do is bench it more, the more people who bench it the more pts. Seems pretty straightforward to me

That Gma booster tool basically set mobile chipset to run at desktop clocks, it essentially just set a register. Mobile was at like 200 mhz so boosted it to 400 mhz

I find certain Gma chipsets clock pretty well, but then certain ones are terrible. The good ones probably would be well worth vmods

Interesting as I have seen it where lower chips can't push FSb as hard but score is almost the same until mem goes higher with higher FSb.

Typo?

Is Gma popular this week? Lol

Gpu is overclockable but it's weird. Basically stock the gpu always runs at a certain frequency, I think 400 mhz, no matter the cpu you put in. What it does is has a divider or multi which is selected by the stock FSb, so lower stock FSb means higher multi. It then increases with FSb, although no sensor shows that. There is also no way to set this multi aside from changing the cpu or perhaps doing a bsel mod. Memory also has a huge impact on score, especially mhz because most boards give you extremely limited timing controls. But anyways this way that it scales with FSb despite no reading showing it would be why cpu oc seems to increase score a lot. I had to read through documentation from Intel with the help of mickulty to figure out how clocks work for Gma.

Should be a box to tick for cores unlocked right?

I've always saw it as separate by model number and arch, not step. Once you separate by stepping then you also can separate by tdp which is an even bigger mess. There are chips with same model number, same arch, same process, same stepping but different tdp rating which means likely different binning and they have a different amd part number on the ihs technically but it's the same model number so same category. I think trying to separate those out would be a mess, and stepping (which is already a mess) would lead to that.

I think you have it backwards, the idea is you donate out of the goodness of your heart, then they added the comp to thank people for donating. Whether or not you agree with it I think that was the intention.

Xtu profiles are well broken, basically everything about xtu is.

Well said, hopefully getting second can get enough interest that we can get a meet organized in the US.

Blame lack of sleep, uploaded valid while subbing then never attached to score https://www.3dmark.com/3dm/55888989

@Leeghoofdplz remove, subbed to wrong category

Scrolling screenshot didn't take well, so made scrolling gif https://i.imgur.com/kQh4BeU.gif

https://i.imgur.com/kQh4BeU.gif tried scrolling screenshot and it was weird, so here's scrolling gif, insane imc and fclk on renoir, 2500 is probably possible

The pll is on the mobo not the cpu, you would have to see what pll chip is on your mobo. It should be near a crystal occilator.

Summed it up perfectly

Seriously?

Tse is similar, just heavier and scales to more thread. Maybe uses avx, I know you can make it use up to avx512 with custom settings but idk what it does stock. Ycruncher is mostly heavier than x265 4k, which isn't a bad thing, especially cause it uses avx512 which no other bench on the bot does currently afaik. At higher core counts its mostly about how fast is your mem, on 3175x it was zero core scaling on water 100% mem scaling. And it was so held back by mem that core temp went up 10c from a mem oc just from extra core utilization. Lower core counts scale more with core clocks

Its supposed to be

Personally I have no issues with it cause it's hardware, also it isn't necessary faster than ssds anyways. But not sure how leeg feels. Hopefully storage isn't that big a deal anyways as it's only part of a subtest, and gpu shouldn't matter much as no game test in express.

Also a side note, I'm sure you've done it but we'll make sure ramcache is banned on pcm10 right?