MrIcee on a AMD Roll

[OPPAINTER]

Nice going there MrIcee,
Congrads my Friend and keep going with that AMD of yours :toast:

OPP

[DaGooch]

Excellent!
;) :thumbsup:

[Hiwayman]

Great result MrIcee, Congrats:)

[zakelwe]

Fantastic results for all your hard work. Everyone knows it's easy for those P4 guys to get massive scores by just turning the machine on ;)

I assume you are looking forward to the XP2600 or XP2700 when they are readily available. Onwards and upwards past 19 000 when those come out no doubt ?

I assume the 9700 scales as well as the GF4 ?

Regards

Andy

[SKATAN]

well 1000 more mrks will be in the bag with tbred at 2.8ghz , that one is for sure


Congrats mrIcee you make that kx7 fly

[Chong345]

Congrats MrIcee! :toast:

[DDTUNG]

ICEE you're DA MAN!

:toast: :toast: :toast: :banana: :banana: :banana:

DDTUNG:cool:

[Hardass]

Well done my friend. Another outstanding job.

[qwerty57]

wow!!! thats just amazing....very, very nice.. congrat's you must be happy? :p

[N8]

Very nice, MrIcee! You guys ROCK keeping AMDs up there w/ the 3+GHz Intels :)

[JCviggen]

Awesome score :)


I just rounded the 18K cape today by the way. Now that may not be spectacular anymore, but the radeon was clocked only 370/340. Then I think 18K is a nice score. Only 3490 MHz too. I used 166x4 on the memory and my lobby scores are through the roof for only 370/340 :) Lobby high i'm equal with OPP's 20.3K lobby high score...
Guess I need a better radeon... it wont go faster than this on stock cooling :(

[Marci]

Can't wait to see what u get from the second revision KD7 Boards Randi... good goin!

[docah]

Quite amazing Mr Icee, Now i'm definately looking foreward to those newer amd cpu's.

Side note: I'm going to have to start an old hardware museum from all the deprecated parts i've got! (or build up some pc's and sell em :) )

[PiLsY]

Congrats Randi :) Good job m8. Even though you did knock me out the top 10 you swine ;). Never mind - ill be chasing you again in a week or 2 ;).


The MSI KT4 Ultra may be worth a look in...

FSB up to 250mhz in 1mhz steps, vdimm to 2.8, vAGP to 1.8, Vcore to 1.85. Uses the HIP standard Voltage regulators so easy to mod. KT400 chipset too so a nice bandwidth increase. Pleanty of memory tweaking stuff in bios too from the limited look I got.

PiLsY.

[PiLsY]

JC - youre loosing a lot of points on that lobby test my friend. My 18081 score was at 407/340 with cpu at 3184, mem at 187mhz with tightest timings. You should be slaying my score there. More likely up in the 18500 range.

PiLsY.

[JCviggen]

Originally posted by PiLsY
JC - youre loosing a lot of points on that lobby test my friend. My 18081 score was at 407/340 with cpu at 3184, mem at 187mhz with tightest timings. You should be slaying my score there. More likely up in the 18500 range.

PiLsY.

hmmm... lobby??

http://users.pandora.be/JCviggen/lobby.JPG

since your core is at 400 and mine is 370, I think my scores are pretty far ahead of yours in the lobby...

I lose my points in the car test and the dragothic mainly... dragothic is pure GPU speed and 370 just isnt enough.
Mind you, I just got some C-clamps to put the MCW462 with 156W pelt on the GPU, that should help ;)
Although sadly I just noticed I dont have any hoses that fit the block... will be for tomorrow then

[Bulldog]

Congrats heck of a score for any cpu....OUTSTANDING for AMD great job:toast:

[PiLsY]

Lobby is memory bandwidth limited m8. You should be rocking way up higher than 250 with 169 x4 turbo m8. Id imagine more like 280fps...especially considering the cpu speed.

Compared to 325mhz default with the core at 407mhz I gain roughly 12fps at equal cpu/memory settings.

Maybe im totally wrong here, but I always saw much better fps in lobby when going up on cpu/memory rather than graphics.

PiLsY.

[PiLsY]

Gah I didnt mean lobby - I meant car chase low detail :p. Sorry m8.

The way it seems to break down is...

Car : CPU/Memory limited
Dragothic : Gfx limited
Lobby : Gfx core/system memory limited
Nature : All of the above but mainly gfx core.

It just goes to show on the lobby test how much of an advantage that rdram bandwidth gives. You hammered me by over 20fps low and 10fps high detail despite the much lower core clock. you get a radeon capable of 400/350 in there and youll be looking at 19k methinks.

PiLsY.

[JCviggen]

yeah I thought you meant that other test... and for some reason my car score DOES suck... havent figured out why yet.

I ran it normally and I got 220 FPS ... after the whole run I got my score and I though i"ll run the car low again... bam 240 FPS and 200 points more ... very strange

[OPPAINTER]

JC,

Are building somthing like this My Friend:D

OPP

[Chong345]

Hehe. thats kinda what mine looks like, well kinda.

[TheDude]

CONGRATS MrIcee!

Very happy to see an AMD in that slot!:D :toast:

[Creative]

there are 2 guys above me in the under 3100mhz class and thats MrIcee and his AMD rig :D and another guy here from Xtreme with rdram,but doesnt give his name...:)

CONGRATS MrICEE!!!!!! :toast:

[baker18]

Way to go Randi.............

[sysfailur]

Geebus! Way to go man. I've been mad busy with stuff lately--every time I check back your score is higher. :)

[MrIcee]

Again my friends..thank you so much...you guys are the best and the reason why I do this stuff with ya's:D

I want to let you know that last night I blew an 18,864 in the quest for 19,000 with my 2300Mhz AMD:) The only problem is that I uploaded the score fine..but I got a flag thrown at me from MadOnion that the score cannot be published. I can look at it just fine myself..and everything looks ok. Whats up with that crap???
I'm going to write them and find out whats going on....I hope this is not some other BS shenanigans again.

Randi:D

[TheDude]

That's amazing man! Hope MO gets their Sh*t together! I know you must have worked damn hard for a score like that!

:toast:

[Chong345]

MrIcee,
Slow down. You are going way to fast. LOL

GREAT JOB! :D

:toast:

Chong

[JCviggen]

maybe they figure a 2300 MHz Duron isnt possible... lol

Awesome score though m8, on to 19K it is ;)
What kind of clocks on the Radeon ?

[majormav]

congrats randi nice going m8 just got back from honeymoon will be benching again in a few days looks like you way out in front of me now ,,,, congrats again

[MrIcee]

Thank you guys:)

JC...I'm only pushing 474/730 right now...I still have more to go...taking it 1Mhz FSB and a few Mhz on the Radeon at a time to determine stability and what the limits are...so far so good. I have not taken the shim off the Radeon yet...that may be a weekend project, tho I'm making fairly decent contact with the GPU.

Randi:D

[JCviggen]

I've been trying to lap the shim down a bit, but its damn hard stuff ;) Very slow progress.

I'm still at 370/340 (if my card still does that, after last night's probs at 75 AGP my memory OC was down to 330 *sigh*

[PiLsY]

JC

I hate to be the bearer of bad news but I should point out that my memory OC dropped from 340 stable to around 325 stable just before it died. Over 3 days I lost the 15mhz memory speed. I could just about bench at 330 but on the car test the red car's panels were covered in flickering black squares. Didnt notice artifacts anywhere else except on dragothic where the dragons horns on its head "grew" every time it moved up and down. The top point stayed the same but the rest got bigger to maintain it. Doesnt make much sense to read, but if you see it im sure youll understand what I mean.

Heres a new order to try running the tests in (bear with me on this - I spent 4 hours last night testing with the GF4 to figure it out....), which should gain you around 50 - 120 points depending on how good a run you get. I figured it out by experimenting with each test to see how much of a score bias is based on each, compared to the FPS/Points gain or loss you get for running it earlier up or later down the order. Should be tweaked to perfection. Just so its absolutely clear go into custom and deselect any tests not listed for each pass and select the ones that are. The first time you click benchmark you should only run the pass 1 tests, second time only the pass 2 tests and third time only the pass 3.

Pass 1 :
Car High
Dragothic High
Nature

Pass 2 :
Lobby Low
Lobby High

Pass 3 :
Car Low
Dragothic Low

I lost around 0.4fps from running Nature as the third bench, so no huge loss there (about 10 points). I gained 7fps running car high first (125ish points gain) and 4 fps running Dragothic High second (about 60 points gain). Running Lobby tests on a second pass separately gained me absolutely nothing, so scores the same. Running Car low on the third separate pass lost me about 13 fps, losing 35 points overall, and dragothic low last lost me 7fps losing me 30 points.

Net overall gain of about 110 points.

PiLsY.

[mbrock]

I found some paper thin Copper foil on the back of a LCD laptop screen. I figure it will be easier to shim the GPU up a hair to get better contact.

[OPPAINTER]

Shim on my first card, got a .20mm copper sheet and the hardware store.

OPP

[JCviggen]

OPP,

I put my prometeia on my 9700 today :D
Freaky stuff, not very easy either. Had to use a cold plate. Didnt have great contact but I got ice on the back side though. Not sure how cold the core was.... it did only 400/355. Talk about a crappy card. Default voltage though. Anyway these clocks got me to 18.8K already. Its a start. With a decent card I should be able to get higher. Either my CPU sucks ass or my mainboard does... my bet is on the mainboard. CPU does 3D At 3550 MHz default voltage 1.52. I can barely run it any higher, same as I had with that 2.66 ... since it went higher on your IT7, I better check that out.

MrIcee : hope you can get that score published soon, some digital bloke is at 18.8 just above me. Before I kick his ass in a few days, why dont you do it first with an AMD :D

[OPPAINTER]

JC,

I'm trying to picture why you needed to use the cold plate, didn't the prometeia evap fit inside the Radeon Shim??

OPP

[JCviggen]

Originally posted by OPPAINTER
JC,

I'm trying to picture why you needed to use the cold plate, didn't the prometeia evap fit inside the Radeon Shim??

OPP

no it was just not high enough... the side of the microfreezer hit a cap and I couldnt find anyway around it to make contact Ã:banana:nd still allow my card to more or less fit in the AGP slot. A 3mm silver cold plate did the job though, so it was only a little bit extra hight that I needed...( the low cap at the right top of the core) too bad though, it would have gotten a better overclock without the plate

[OPPAINTER]

I didn't think about the Height, I was going to try it myself, I'll have to figuer it out.

OPP

[Creative]

Damn I wish I had read this sooner PiLsY cos I only need 89 points for 18K with my setup and I think your directions above would do it for me but I just cbf benching anymore after 10 days straight :( :D

I have sent you a PM bud...:)

Cheers

[Hiwayman]

Hey JC,

Nice bench with stock CPU sttings and an obviously lame 9700. :)

I'm just playing around with my *new* Asus P4T533 32 bit mobo again. Do you find that the vcore has to be kept real low for it to be stable or did you fix that somehow?

I've go to do something I can't keep using the old TH7-II for much longer.;)

[JCviggen]

Hiwayman,

the lower the Vcore, the more stable the board is yaeh... also the higher memory overclock you can get. I've done some extensive mods to this board involving extra caps and a Vrimm mod. In x3 mode I've done 180-190 stable at 1.7V+ on the core, but that was with a 2.53 not a 2.80. This 2.80 really doesnt seem to like voltage. But it has to be said, I tested a 2.66 before this and it was also terrible... only low voltage, and I couldnt run 3D at any high speed. I could boot as high as OPP could with that CPU but not run 3D ... so I'm going to test an IT7 and see if that miraculously fixes my OC or not. If not, i'm stuck at 3520 MHz but with fast memory... its a strange board I can tell you that much... lol

Actually the TH7-II is still a pretty good board... at the same speeds you can put it against any board and it'll do pretty damn good. Only its not very friendly on the CPU overclock. You lose some there

[Hiwayman]

JC,

Your experience exactly mirrors what I have found, like so many others.:(

This is the second board I've had, I rma'd the first one. You've done really to make all those mods to try and sort a buggy offering from Asus ( not my favourite manufacturer). But, I'm sorry to hear that you're sill having problems with it.

Oh weeell, one more try I guess then.... RMA time *again*:rolleyes:

BTW, I recon the TH7-II loses me 100-150MHz in oc.

Did you ever try the P4T-E + Tubo PLL route? That seems to work ok and does not have the voltage stability problems of the 32 bit mobo......:D

[JCviggen]

I dont have any T-PLL's around and i'm not that electrically skilled :( I can do some Vmods but thats about it i'm afraid.

Also, those boards dont really get sky-high x4 performance either, just very very good in 3x mode... my P4T533 will do 168x4, its just that I cant run 3Dmark very high ... I think (168x4) (3520 MHz) should be the equal of (174x3) (3650) any day...

[Hiwayman]

Yeah, I havn't got the confidence to try the "home-made Turbo PLL" either yet.

But hey, maybe if all this doesn't work out I'll be forced to try it.... :help:

[MrIcee]

Hi guys:)

As I said yesterday..I wrote MO to find out why my last .3dp can't be published...it's being fowarded to the appropriate person but heres the immediate answer I got:

"Hey Randi,

I'll forward your project & email to RistoJC who is our database guru! :) Hecan answer what the problem is. I'm not sure why you aren't able to publish it. There must be something wrong with the project, or then your CPU has fluxuated too much during the benchmark. Dunno.. But as said, I'll forward this to the right person!"

CPU flucuated too much?? Thats a new one to me:) Oh well..lets keep our fingers crossed they straighten this out...I'm still working towards 19,000. BTW...I do believe it might be a filter in place on "Duron" cpu's...JC may be right. I submitted a run of 16,060 at 1699Mhz no problem...and my last successful unlimited MHZ submission was at 2304Mhz...could be 2316mhz triggers a rejection filter as the Mhz might be bogus for a "Duron". I explained to them its a T-Bred and not a Duron...that 3DMark does not recognize it as such on my KX7.

Randi:D

[SKATAN]

MrIcee

I had the exact same problem with a score of mine :( and only changed the cpu 5mhz less on non game benchs

I know the limit is higher than 5mhz ( i think it´s 50mhz of variation)


RistoJC´s anwser


Hello

As said before we do not filter by CPU speed variance in our end.
Because your project CPU speed change has been detected
it has varied MORE that the 5Mhz that you reported (the actual
measured cpu speed is not always what it is set to be, it changes
depending on measurement conditions). Also the direction it changes to
has significanse.

Because we do not store the CPU speed measurement data it is not
possible to verify how much it has varied but since too much change
has been detected, your project is disabled.

The correct way to use our benchmark is that it is run on one frequency
from beginning to end so changing the cpu speed during testing
distorts the results. This is why it is not allowed to change
the CPU frequency. Because of differences in HW and measurement
conditions some variation is allowed.

So the question is: Is driving the benchmark to its variation limits
by artificial means legit? Currently it is but if the variance triggers
activate then benchmarker has gone too far and the project is disabled.

I only changed 5 mhz on non game benches GRRRR

3dmark2k1 is on hell of a bug


I know theres a limit for cpu variation and when the limit is exceded 3dmark writtes in the project and the orb doesn´t let you publish :(

I had to do the bench again


less then 200 for 19k AMD :)

with tbred B you will go 20k easy my friend

nice to see an AMD ahead of so many Peefours

good work :)

[macci]

Originally posted by PiLsY
Running Car low on the third separate pass lost me about 13 fps, losing 35 points overall, and dragothic low last lost me 7fps losing me 30 points.
Low detail test: 1FPS = 10Marks.. so 13FPS on Car Low = 130 points and 7fps on drago low is 70 marks

so you lost some 200 points there ;)

[PiLsY]

Macci :).

It doesnt work like that m8 (or so it seems). All tests seem not to have equal standing on your final score. I could be completely wrong here, but when I tested everything the car low seemd to affect things MUCH less than dragothic low.

If you know different m8 let me know :). Is that info on the low detail fps to points ratio from madonion? If so something odd is going on as I seem to be losing about 3 points per fps rather than 10.

Anyway, does that same scale include high detail tests (1fps = 10 points)? I should think under those rules itd be more like 1fps high = 20 points no?

I did try and figure out a formula as to how it gets its score based on my results, but didnt manage it :p.

PiLsY.

[PiLsY]

Hmm just totted up my results according to the orb and it seems to be (approximately) the following :

Low detail - 1fps = 10 points
High detail - 1fps = 20 points
Nature - 1fps = 20 points

However, the results dont exactly match - im getting a variance of +-30 points going through my scores. Some are within 2 or 3 points correct, others are up to 30 points out. Does 3dmark calculate scores on FPS to 2 decimal places, but only display up to 1? That would account for the variance.

However, it still doesnt explain my GF4 scores that I based my original test order on. Using the above method the score I got is over 25 points out? I re-ran it to be sure and got the same results again. The same score (give or take 5 points) and the same fps (give or take 0.2fps) and running in the order above still gives me better scores, despite the apparently lower overall score I should be getting. Im wondering if something else (probably obvious) that ive not noticed is coming into play.

PiLsY.

[Hiwayman]

Pilsy M8, :)

I thought that low tests give 10points / frame and Hi's + nature give 20 points/frame.

Erm..... you got there first,

[macci]

Pilsy,
"The 3DMark score is calculated using the following formula:

3DMark score = (Game1LowDetail + Game2LowDetail + Game3LowDetail) * 10 + (Game1HighDetail + Game2HighDetail + Game3HighDetail + Game4) * 20

Game1LowDetail means the average frame rate of the first game test in low detail etc."

source => G:\Program Files\MadOnion.com\3DMark2001 SE\Help\3DMark.htm ;)

[macci]

However, the results dont exactly match - im getting a variance of +-30 points going through my scores. Some are within 2 or 3 points correct, others are up to 30 points out. Does 3dmark calculate scores on FPS to 2 decimal places, but only display up to 1? That would account for the variance.
I have yet to find variance higher than 2points (your score on the front page is within 2points correct). And yes its fairly obvious that 3dmark uses more than 1 decimal for calculating the result hence the minor variance.

[PiLsY]

Yeah the 18091 score is pretty much bang on.

The last few runs with the GF4 on det 40.41 havent been though.

Point in case...

Car Low : 222.5 = 2225
Car high : 80.8 = 1616
Drago low : 251.8 = 2518
Drago High : 134.3 = 2686
Lobby Low : 213.2 = 2132
Lobby High : 96.9 = 1938
Nature : 87.5 = 1750

Total = 14865
Actual = 14857

Total variance should be less than +-3.5 points due to rounding the decimal place up or down. Actual variance is 8 points - more than double. Had to use this example as I didnt have the others saved. Its either taking something else into account or not too hot at maths :p. Either that or ive made some foolish mistake adding up.

Anyway, Im not doubting your word macci - just 3dmarks ability at addition :). Having checked through the majority of my results most of them tally within 2 to 3 points of the above maths, but 4 I have saved are WAY off. One of my voodoo3 scores is out by 34 points of average. Very odd.

PiLsY.

[MrIcee]

Skatan:)

Well heres the deal my friend...I did not nor do I ever change Mhz speeds running 3DMark...that run was done at 11.5x201Mhz (2316mhz) straight thru with no variances....nor did I alter videocard overclocks. This is why I don't understand the cpu fluctuation thing. There has to be something else wrong..and I suspect that the fact that possibly the 2316mhz Duron recognition of my T-bred could be the problem. I ran another bench at 2304Mhz and it accepted it fine.....just for trial and error. It will be interesting to see in the end what is wrong...in the meantime its a bit upsetting I can't take my place on the ORB:)

Randi:D

[JCviggen]

MrIcee,

Could you please get a few pts more ? :)

DJ is at 18872 ... the more Xtreme ppl above him the better ;)

[MrIcee]

JC:)

Working on it M8;)

But I'm afraid no matter what I get at 2316Mhz or above I will not be allowed to publish it. I'm getting maxed right now and flakey trying to attain higher with the current CPU. My fingers are crossed that the 2600-2700+ cpu's arrive soon. I think if 2700-2800Mhz(maybe 2900Mhz) can be attained with these Rev B T-breds...from the numbers and gains I've seen in my own testing with 600Mhz increase in cpu speed...its not entirely inconcievable that Holicho can be challenged with an AMD in a Prometetia:)

From 1700Mhz to 2300Mhz speed increase equated to a 2800 point gain(16,060 vs. 18,864 with same video clocks)...I have not maxed my videocard yet either...stay tuned for the Rev B benching sessions:)

Randi:D

[JCviggen]

well you wont get linear increasing performance I think because you start running out of bandwidth a bit ... but it could still be a 2000 point gain ... interesting to say the least ;)

I broke 19K today. And while I was at it, I decided to pass it by a good 400 points :D

[MrIcee]

Welp...official word is my score will not be allowed due to "CPU variation" measured during tests. I'm not sure how this happens when you run at 1 speed consistently thru the bench...but I'm screwed now with my best score. I'll keep after it but afraid any further benching above 2300Mhz with my AMD will not be allowed due to this "CPU variation".

Randi

[SKATAN]

I told you :(

3dbugmark2k1 is fantastic

I think if you try again it will probably let you publish

or have you tryed over 2.3ghz and cant publish ???

[JCviggen]

Randi,

maybe run a bench again at 2316 and see if it gives you the same crap? You never know that it might have been a glitch in just that run...

[Creative]

Your just too good for the system MrIcee :D
hehehe

[MrIcee]

I have benched again at 2316Mhz...and it will not allow me to publish. Apparantly I've stretched my overclock so far on the 2200, or so it seems, that the cpu tho stable enough to bench the complete bench, is showing some sort of variation in the bench. Benching at 2304Mhz publishes just fine...but nothing over that. This really pisses me off as I believe from testing indications this chip will bench at 2350 or just under. Needless to say...this really sucks..another thorn in the side of AMD.

Lets just hope that MO accepts benches on highly overclocked 2600+'s. I'm going to ask if this anomilie has occured with Intel.

Randi

[Chong345]

So there is nothing MO can do?

Chong

[PiLsY]

I cant help thinking of some sort of strange sandwich when I see this thread title....

Anyway, good luck on increasing that score Randi :). I want to see you over 19k without having to use the 2600+ :p.

PiLsY.

[Hiwayman]

Good luck Randi! :)

You are upholding the *tarnished* honour of AMD singlehanded.:hehe:

And Pilsy :lol: @ sandwich/roll:D

BTW Is anyone gonna wish me luck? I'm trying out Asus P4T533 -32 #3 in the morning! :help:

[mbrock]

Is it the method of unlocking the CPU that makes it show as a Duron? I only used the loop method on the first L1 and MO shows my 2200+ as unknown. If yours showed as Unknown do you think they would accept 2300+mhz? Just a thought.

[MrIcee]

Marty and Chong:)

They've recently dealt with this CPU fluctuation thing most recently on a couple other cases..and their decision ultimately is to not allow publication.

As far as the CPU detection as a Duron Marty...they said it has no bearing on the decision....this CPU flucuation is detected between or during tests. To me it would seem a variation would be changing cpu speeds IE: FSB...but that is not the case here...it was straight thru at 2316Mhz as I've always done in all benches.

My feeling is there may be an internal flucuation or some type of anomilie with the Athlon itself when its pushed in overclock as far as I've taken it in benching. I am no engineer so I do not understand the technical stuff. Mind you...I've screenshot this puppy as high as 2496Mhz so far....I'm trying to break 2500 without LN2:)

Randi:D

[mbrock]

Good luck on 2500mhz.:toast: I guess I am going to have to get off of my butt and purchase a better cooling setup this H2O just isn't cutting it anymore.