chew*

Nice i figured it was a round 10 second difference.

Ok I guess 5:36 eludes me for now on non v cache untuned OS. Will revisit later when I have a proper OS tuned for this benchmark specifically. Use v cache for 5.0 from now on gents. I guess non v cache for anything 6 gig.

V cache is definitely faster. But so am I.

chew lives

Merely showing possibilities. c24 32m stable is one of them.

Test to see where 64g stack up, HwI open entire time to monitor clock speed avg 5040 can't control this cpu. no tweaks shit os just using core 2.

Probably gonna want to drop speeds a tad as CL24 is a reality......just not sure if its a reality at 6600 yet.....but 6400 is done.

Finally logged in. What's up infra. Don't think I have a 6gig cpu. Where should we start? OS? 7,10,11?

safedisk can you add poor mans impact to the front page? ASUS ROG X570-I

lol yah that was a zen 1 thing for launch agesa.

3/8 used to work better for me on R1 as well. Not sure if your sticks can do it but 10-9-9 showed a lot of promise in the past, don't bother with cas 9 though it was slower than c10. dropping twcl -2 of CL is an intel thing. AMD likes them =. Honestly I ran some aggressive settings the other night vs my daily mem timings. I'm not seeing anything spectacular. 100 mhz is worth 15 secs on avg. im doing 6:00 at 4850mhz on prime stable settings and 6:15 and 4750mhz. Honestly the times im seeing are where they are supposed to be for the cpu speed and the memory tuning gains are like pulling teeth for any gains. I see what you meant about certain dividers. fwiw at least yours can boot. 5900x boots 3733, 07 hang @ 3800, boot 3866 all the way to 4066 so lol..... That's on the dark but my 5800x is fine......

My guess would be aggressive error correction. 2000 is not really any more stable than 3933 or 3966 all produce whea errors when not cutting OS memory down via maxmem. 3800 and lower is perfectly fine if there is no post hole which apparently even on DARK my 5900 has at 3800 with 07 post code. 3800+ boots fine but wheas.

I've played with the feature. It's viable if your not setting your baseline at "prime stable clocks" but AMD does a better job at it if your target fixed clock is "prime stable clocks" Right now other than that feature and or want the ability to SHOW a number higher than 2000 fclk or require the slightly improved vrm for benching on ln2 or just want a fanless heatsink solution I see no reason to upgrade to DARK. It's not bad, I like it but overall experience wise it runs and performs the same in the 24/7 stable department.

probably a similar situation as this. Fmax scalar vs PBO. 4.6 avg clocks through CB 4450 avg clocks through CB Although I noticed something when pushing dual CCX chips to the ragged edge that score is effectively half what you would expect leading me to believe 1 CCX has dropped.

define working are you looking to be stable or looking to bench. I'm not saying its impossible but i am saying its possible that it will never be stable due to the whea reports being cpu bus/interconnect. booting and running does not denote stability which both hero's lack at anything over 1900 in my testing.

High pll buys you about 200 mhz more mem clocks linked. without it the chVIII non dark is capped at 3800/1900. If you experience slowdown crank it till you don't....... Board choice seems to matter alot as to how high the fclk can go linked to mem, as long as the vendors aren't already playing games with pll you should be able to gain 200 mem/100 fabric.

Jack up your PLL voltage to about 2.1v. Benching only........do not take this advice anyone running 24/7.........

probably error correction feature in cpu causing slowdown. bclk on asrock x370 was limted to like 110 previously have to kill all devices no nvme best to physically unplug the bluetooth/wifi from board as it knocks it out will probably break it. pci e gen 1 etc etc all the AMD bclk tricks will be needed for it to maybe work.

yep davis was a tad faster than this on the giga which was faster back then. This was one of the fastest 4 gig results back then. looks like 5000 series does it easily. Now ill have to break it down see if cpu model impacts results or we don't need the highest cache cpu to play the game. I imagine at your experience level you tested 2x8 also loud so coldbug was bad regardless i gather?

now AMD is faster than r1 finally. hows the 4 gig clock results loud?

yah it's not that causing this problem speed. These are neos. they are from the new bins. In fact they are very close to the non existent 1.4v c16 4000 bin as they can do 4000 16-16-16-36 4000 @1.4v if I leave everything else auto. I'm only running 1.45v to 100% ensure they can handle my subtimings. 1T is really not the problem anyway. the problem exists without even throwing 1t in the mix and the problem lies somewhere in the training when trying to manually tune shit. If B0 and B1 is shit then A0 and A1 should be shit as well. I have 4 8gb sticks of old B die so I will compare them versus my royals 2x8 gb which are new bdie but tbh i'm not expecting much as this board currently has an artificial wall. I already found a very convincing bug and can do 4600 with only 1.35vddr ( because if you change that it fails boot ) Almost full auto boots and runs tests fine 4400/4500/4600 but copying settings and inputting them identically fails training so it's not possible manual. Honestly I know how they test and i'm testing the same way so they probably think nothings wrong board does 4600 it's good. Go manual and not so good

That is not gonna kill anything. I've run up to 1.7v and i'm sure the hardcore benchers run more.

Just requires a little balls and good air/water, @ 4300 he was probably under 1.4 vccsa and 1.3vccio as that what the board autos to @ 4400.

Heres proof on 1.2z bios. I included voltages. If I even set vdmim manually to 1.35 this will not make it to bios. I would not call the bug voltage related I would say its manual settings related as literally I set speed primary timings that can actually run @ the 1.35vdimm limitation and set PPD 0. If I set rtl training no post. if I set turnaround training no post. if I set vdimm no post if I set vccsa no post if I set vccio no post. If I set subtimings no post. I should rephrase that. it fails OC it posts but fails to boot with desired settings. This is your 4300 limitation bug. Unfortunately my 32gig set can't seem to hack it over 4600 with 1.35 vdimm limitation or the vccsa is getting more retarded than it already is @ auto.....