ok i will wait for vdimm mode to see what can my hyperx do and then decide what to do. bu your score increase is very impressing
Printable View
ok i will wait for vdimm mode to see what can my hyperx do and then decide what to do. bu your score increase is very impressing
Thanks macci........ :toast:Quote:
Originally posted by macci
OUTSTANDING stuff there Hipro :toast:!
You really killed the PCMark mem competition :D
but as far as ASUS vs ABit goes.. yes Abit is faster than ASUS in 5:4 mode (not much thou. the hexus score I got at 4333 was with slow asus bios). But in 1:1 mode ASUS is atleast as fast Abit ;)
And heres an unbuffered:
its at 262x12 thou :D
http://www.solidhardware.com/macci/p...62_unbuf_s.gif
I think that BOTH of those Mobos are damn fast........ ;)
Here's another one at 261MHz a bit tweaked...... ;)
http://www.pctech.gr/hipro5/photos/P...b_3956_png.png
Thanks DJ....... :)Quote:
Originally posted by SpicyHuevos
Nice job Hipro its all you now as i have rescently lost a little faith in pcmark as a performance indicator for memory.
So Enjoy have fun good luck etc:)
Thanks shido.......and for the VGA Card 9800 pro...... ;) :DQuote:
Originally posted by shidokanjohn
excellante work once again you crazy *&*^*&^ i mean Hipro5
:)
cracked the 13k finally (well actually it was not that hard after all)
256x16 1:1 2-2-2-5 on P4C800:
13044 with R9800Pro (no overclock)
oh and btw how about a new challenge...3DMark2001SE with non-overclocked video card? :D
here's mine:
http://www.solidhardware.com/macci/3...tock9800_2.gif
full screenshot
16x250 1:1
R9800Pro
Any time my friend:toast:Quote:
Originally posted by hipro5
Thanks shido.......and for the VGA Card 9800 pro...... ;) :D
That's cool....... :DQuote:
Originally posted by macci
oh and btw how about a new challenge...3DMark2001SE with non-overclocked video card? :D
here's mine:
http://www.solidhardware.com/macci/3...tock9800_2.gif
full screenshot
16x250 1:1
R9800Pro
Non-overclocked cards???.......I haven't done it yet I think.!!!! :)
Macci after all , running 5:4 with aggressive timings, is BETER or equal than running 1:1 with lose timings.......I had a 9800pro yesterday(thanks shido) and I benched again with 1:1 and with 5:4.......5:4 IS a bit better than 1:1........So our theory is right about those new mobos and nowadays Rams........I have done this before with my MX440 and I had seen "small" differencies.So I've done it yesterday to be 100% sure with the 9800pro........5:4 and aggressive timings IS BETTER than 1:1 and lose timings........
This is 1:1 with the GEILs at 2.5-7-4-4......
http://www.pctech.gr/hipro5/photos/P..._24530_png.png
.....and this is 5:4 with Kingstons at 2-5-2-2........
http://www.pctech.gr/hipro5/photos/P..._24594_png.png
after doing some benchmarking myself with 2 different kinds of ram tonight and only Memtest of PCMark2002 I think I have found a few patterns here...
having very high 1:1 at like 280 at cas 3-4-4-8 doesn't matter one bit.
cas 2-2-2-5 will always beat it.
at cas 2-2-2-5 at 232 I can get just over 6400 on memory.
on cas 3-4-4-8 at 281 1:1 I can get just over 6440, oh yeah, forgot to mention, this is the bandwidth test of Sisoftsandra I am talking about here.
but the results of cas 3-4-4-8 ram on PCMark2002 at 281 1:1 is only about 9860 with my Ti4200.
but running cas 2-2-2-5 at 232 both 1:1 and 5:4 always wins by over a thousand points with this same ti4200.
so now it seems like the Memtest of PCMark2002 measures more speed, and bandwidth less.
cause these two sets of ram can produce nearly identical bandwidth numbers, but thier PCMark2002 ratings are very different, so having high cas is more important then having high ram ratios it seems.
anyone see the same thing?
PCMark2002 Memory bench IS counting down the speed of P4 or AMD internal cache(L2)......It's a synthetic bench......So the higher frequency you bench,the higher the resaults are.......(don't read this macci).... :D ;)
your gonna have to translate that for me, cause although I understand what your saying, I don't see the relation to ram...
but then again, it is 4 am, been benchmarking for 4 hours, and don't have a single benchmark worth showing after tonight, although I managed to sqeeze another 15 mhz outta my card, I couldn't get one benchmark for 3Dmark tonight :(
just a few tests on sisoft and PCmark...
so I am basically a vegetable right now ;)
Yup I've come to the same conculsion when I tested the Adata PC4000. It was slightly slower than 5:4 2-2-2-5 all the time.Quote:
Macci after all , running 5:4 with aggressive timings, is BETER or equal than running 1:1 with lose timings.......
btw, how does the Geil do in sandra unbuffered?
PCMark mem is not really the best membench cos its way too CPU dependant (basically all the tests that are sub 512k will fit to CPU L2 cache..and thats pure CPU MHz) and does not always show the real world performance. Also certain test scores seem to jump up and down quite a bit with no obvious reason.
Here an example comparison:
http://www.solidhardware.com/macci/pcmark.gif
my 256 1:1 2-2-2-5 vs. Tom Holck's 264 1:1 2.5-3-4-5
Now the first two tests (raw block read 3072KB/536KB) seem to be like they should - my 256 with aggressive timings gives better performance than 264 with slow timings. Then the rest of the raw block read tests again seem to be ok and Tom has the higher score there because his L2 cache is running at higher speed (=higher CPU speed).
But then comes the Raw Block Write (3072/1536K) which again measures mem speed and obviously 256 2-2-2-5 SHOULD score higher here. Same goes to the Raw Block modify ones (3072/536K). Now its beyond me how those slow timings can give such good results on these. Also these tests jump up and down by more than 300MB/s on my system depending on how lucky I am. :P
Even Sandra unbuffered seems to give far more realistic bandwidht numbers than PCMark mem in some cases. For instance I'm quite sure that 256MHz 2-2-2-5 kills 264 2.5-3-4-5 on that one even thou the pcmark mem score would give the idea that those speeds are near equal.
As far as real world mem speed testing goes I haven't found better test than 3DMark2001SE Lobby hi detail (my record is 158.2FPS on that one btw).
My KHX2700 runnin at 178Mhz...5-2-2 1T CAS2....only bench's at a lonely 3500!! HAHAH
S-Sandra unbuffered is lower with GEIL at 285MHz and 2.5-7-4-4 than the Kingston(BH-6) and 2-5-2-2 ..... About 150 to 200MB/s lower....... ;)
Looks like this bench jumped to completely new level with 3.2EE.
It scores 14411 at default clock (thanks to 2.5MB cache)
http://www.solidhardware.com/macci/bench-pcmark.gif
pic from x86secrets article
i hate new cpus new money to spend :(
Shi(f)t.....That's a god dumn GOOD CPU for 3DMark macci........ ;) :D
I sense 16-17k ridiculous scores coming soon from some here ;)
x86secret webmaster just got this (P4 EE at ~3.5GHz):
15768
As you can see the CPU with 2.5MB cache totally owns the 512k version in Raw Block Read - 536KB, Raw Block Write - 1536KB and Raw Block Modify - 536KB. This is because the EE can use on die cache for those and the non-EE (with 512k) must use RAM. So basically PCMark2002 mem has nowadays very little to do with actual realworld RAM/bandwidht performance.. its all about CPU and cache speed. :/
Only the 3072KB tests (3 out of 21 mem-tests!) will measure RAM bandwidth.
i still think pcmark is a rather useless bench... the memory score thing is sort of unrealistic as well ... theres more to life than a lot of cache, memory performance doesnt increase 40 percent with some more cache...
BTW these EE chips arent selling for quite some time and supposedly not retail until early 2004 so i am not sure its relevant to compare too many things now
new age for pcmark mem benchies
new world order :D lol
Indeed. Like I said its basically CPU cache tester nowadays.. only 3 out of 21 memtests have to do with RAM speedQuote:
Originally posted by JCviggen
i still think pcmark is a rather useless bench... the memory score thing is sort of unrealistic as well ... theres more to life than a lot of cache, memory performance doesnt increase 40 percent with some more cache...
So basically if you want to test system bandwidth use Mark2001 Lobby hi :D
I think I got a goddamn Faker on my hand. Look at this:
http://www.overclockers.at/attachmen...779&fullpage=1
Claims to have 17578 points @PCmark02 MEM.
With this stats:
260MHz FSB
1:1 divider
Timings 3-4-4-8
I think this guy is seriously bull§§§§ting. :bs:
If he is running a EE chip then 17k is very possible
Here is mine @ 18k and 19k+ is definitly possible
http://service.futuremark.com/compare?pcm=1592120
I found the compare
http://service.futuremark.com/compare?pcm=1637175
Looks good to me
Heres my latest PC mark, just did a few moments ago.
FX-51 at 255x12= 3062 Mhz
OPP
Nice hard drive score too, running raptors in raid?
IBM IDE drives, promise PCI raid controller.Quote:
Originally posted by FUGGER
Nice hard drive score too, running raptors in raid?
I have never been able to get my Raptors into windows past 224FSB on any controller yet. Bummer,, those puppies put out a good 2400 score in PCMark.
OPP
wow.... 2400 hdd score :slobber:
Dual Raptors Raid-0Quote:
Originally posted by Evil_Spork
wow.... 2400 hdd score :slobber:
OPP
wow thats good stuff there. opp you have teh hdd tach program to test actuall MB/s? that would be an interesting view ;)
Use to, can't find it.Quote:
Originally posted by Evil_Spork
wow thats good stuff there. opp you have teh hdd tach program to test actuall MB/s? that would be an interesting view ;)
OPP
quite awhile ago when the raptors were new, I was asked to build a PC for a friend of mine-he wanted the fastest AMD PC he could get.
so at the time, I built him a Asus A7N8X-XP3000, R9800 pro, 1 gig of corsair XMS ram...
with 4 raptor drives, and no overclocking.
this is my friends PC and he doesn't need it.
the raptors were in raid zero, and I just dropped my jaw to the floor when I saw the benchmarks from it.
they were putting out scores around 3500.
my IBM deskstars in Raid Zero with 10 gig partion get 1850, which is pretty good for IDE. but damn, them raptors still left me amazed...
heres my only proof of a 1850 score with hard drives.
I wish I had the raptors to show you, but I did all the testing on his PC...
all I can still say is they still just amaze me ;)
http://service.futuremark.com/compare?pcm=1503654
man you guys are all breaking 17-18k and I still haven't broke 13,000 yet-I feel left behind :(