Printable View
I think only splmann can remove all doubt here, but from what I see, I can't be 100% sure that it's bugged. If latency was 31.x ns it would be a bit more clear, or we had a :banana::banana::banana::banana:load of similar scores in the database to compare it to - but atm I'm giving this one the benefit of the doubt.
I vote we just change the name of bench to maxcpu.
It scales alot more from cpu than ram anyway......
Only with SB though right?Quote:
Originally Posted by chew*
SB - Core freq = L3 freq
Not true with older architectures, on BF/GT it scales with uncore, right :cool:
I dunno, It scaled more with cpu clocks on amd to whether it be 754,939,AM2,AM3..........In fact we found that L3 was actually worse.......semprons at 4gig were kicking phenom II's in the nads at same clocks.....difference between the 2? L3..............Quote:
Originally Posted by miahallen
Not to sure about intel, never really ran it on intel till now because I felt the bench was sort of retarded.
1. I can run this bench far more unstable than any other bench out there......1m is harder to run and that doesn't even stress ram.......and this is supposed to be a memory benchmark.
Put simply if I can validate, I can maxmem.......
2. cpu speed can trump ram speed in this bench.
For instance SF3D ran 1140 ram verus my 1115........gained what 50 points over me?
Me running at 5.0 to 5.5 gained me over 200 points no changes to ram.
Anyone with a 5.7 chip can beat the current scores without to much effort. :shakes:
chew is about to hail a storm of :banana::banana::banana::banana: all over maxxmem W.R. soon boys dont waste time discussing it even heh
on a serious note it seems that latency can bug a bit every now and then as ive seen it myself but i havent got a CPU that can bench at his frequency to be able to confirm his score
petri is that what you are seeing, latency above 33 or what?
Heh I would not say that, If I do play with it again I will at least crank ram up more, just tossing cpu speed at it is a lame way to win :rolleyes:Quote:
Originally Posted by dinos22
Ohh and for the record this bench is still gay, just in case i haven't said it enough.........I wouldn't even claim WR if I did get one......
@splmann...Just run it again with the same settings and be done with it:)
Although you exaggerated the hell out of your last few posts, what's the alternative for benching memory?Quote:
Originally Posted by chew*
32m 01 and 3d 05 :DQuote:
Originally Posted by _mat_
Seriously don't have the answer for that, but I don't think this bench is the answer or ever will be unless it can be fixed which I doubt as most of the issues are likely us running on the edge.
Also, note I only run this bench when one of my team mates keeps poking me to.
I would much rather sit here tuning 32m on amd hours and hours on end and waiting 14+ minutes for a run to complete for water runs.......thats saying something....
aida64 :D lol nah i'd go with SuperPI 32m and everest bench..
Memory performance on our favored platforms will always be dependend on cpu frequency. But 32M involves too much cpu ...Quote:
Originally Posted by Johnny87au
Maybe the folks over at AIDA64 can make their results comparable (and free), if that's possible. As far as I see it, they calculate even more numbers, so that should be no problem at all. The bench seens also very stable ... or lets say more stable than MaxxMem. I know someone there ... I will call him up today.
sounds like a good alternative matQuote:
Originally Posted by _mat_
Problem solved.
http://www.hwbot.org/community/submi...m_2714.9_marks
Also now that we have run at that speed, I can safely make a determination.
splmann, your run is certainly bugged, I'd pull the score if I were you.
Nice ! I have higher Backup's but i dont upload this to much discussion . I hope we get a better memorybenchmark in the future !Quote:
Originally Posted by chew*
I see real record done by MSIMAX. Well done!
Splmann, you can not call those scores you have as a backup, when the one you called record was bugged. I hope you see what many people have said here in a friendly way.
Benchmark might bug sometimes, but it is us, who decide will those scores be published or not. You decided to show yours, but others didn't.
I think this discussion is over now, at least from my part.
Keep pushing the limits people!
haha this one is all chew and his veeeeesta tweak :rofl:
http://www.youtube.com/watch?v=ZwkSP...layer_embedded
booting in at those mem clocks damn chew :D
Very nice work and certainly a very good cpu! :toast:Quote:
Originally Posted by chew*
Another run from me today with a different cpu, higher ram.
As we can see the scaling is linear with cpu and or ram speed. One does not show massive gains without the other, nor do they show massive gains in general.
gawd damn sic memory man :up:
Amen to that! :yepp:Quote:
Originally Posted by chew*
hax chew hax, i mean buggy maxxmem :D
So; CHew... you maintain that the result by splmann was bugged? He has not revealed his subtimings, and we are talking ~½ns delay difference...
Oh and you forgot to reveal which ram you used :p:
want to know my subs?Quote:
Originally Posted by M.Beier
Compare our 32m runs, ASUS is not that slow = my subs are tighter ;)
I dropped from 7-7-5 in my 32m run to 7-7-6 but tightened other things up for maxxmem.
Also my speed is higher. From the scaling I see, you need 5.8+ and 2300+ to hit 32.5.
Ram is corsair gtx2. So no ram or a specific rams SPD is not making a diff for latency ;)
Here is more linear scaling:
http://sf3d.pp.fi/images/ASUS%20Maxi...1149%20ajo.png
Actually this test does not bug that often, if you know what to expect from it.
chew, compare our latest scores. Memory write and latency. I think your score is a bit off.. :D
It is getting ridiculous :rolleyes:
Diff board big guy ;)Quote:
Originally Posted by SF3D
Plays a very large role in this bench.