100K 2001 done with 2 ATI 2900 Stock Cards [Archive]

Pedro Rocha

12-19-2007, 08:11 PM

This was one my personall challange in "my favorite game" , some people think can't be done, but I did

Setup
- QX9650 @ 5.350 cooled by mini-cascade (-85ºC)
- Asus Maximus Extreme (coold boot at -80ºC :( )
- Cellshock 1440 1800 7-6-6-18
- 2 ATI HD 2900XT iin CF- stock cards @ 880-990
- Silvertsone OP1000W
- Old school tweaking :)

Photo of the setup :

http://pedrorocha.planetaclix.pt/Cascade100K.jpg

As you can the space between the 2 16X PCie 2.0 slots is quite tight:( but the 2x16 helps a lot, specially, Nature.

And here is the big 100K with stock cards :yepp:

http://pedrorocha.planetaclix.pt/100K_2001_ATI.jpg

Compare URL: http://service.futuremark.com/compare?2k1=9304978

Well now I can rest, until I receive my new dual big rotary cascade, but even with the mini-cascade this QX9650 give a rasonable superpi benchable speed as you can see :

http://pedrorocha.planetaclix.pt/QX9650_5500.jpg

OK enought for today

100K stock cards done :rofl:

VoRtAn_MaDgE

12-19-2007, 08:18 PM

http://www.highenergyconstructs.com/hecla/wp-content/uploads/2007/09/shocked%20nazi.JPG

Congrats dear Pedro :up:
Time to get more cold for that big boy :up:

Great score as always ;)

FUGGER

12-19-2007, 08:20 PM

I see you are having fun benching up at 5.5Ghz. nice chip.

jabski

12-19-2007, 08:52 PM

nice work Pedro.
Congrats :)

AndreYang

12-19-2007, 10:19 PM

nice work:up:

Sampsa

12-19-2007, 10:36 PM

Good job Pedro, I'll also try to break 100k, it has been now over 5 years since my last 3DMark2001 SE run :)

metro.cl

12-19-2007, 10:51 PM

::woot: :woot:

I think im your biggest 3Dmark01 fan Pedro i just love that bench and always admired you results on it

:stick: :stick: mod the cards

Zannema

12-20-2007, 12:36 AM

Ohhh 100K with stock 2900:bows: :bows: :bows: :bows: :bows:

SIOUX

12-20-2007, 01:56 AM

Wow...amazing 3d01 my friend:toast: Expecialy love the nature score you got out of those ATI cards:)

saaya

12-20-2007, 02:14 AM

http://www.highenergyconstructs.com/hecla/wp-content/uploads/2007/09/shocked%20nazi.JPG

Congrats dear Pedro :up:
Time to get more cold for that big boy :up:

Great score as always ;)

hahah, nice picture :lol:
congrats pedro!!! :toast:

massman

12-20-2007, 03:23 AM

Zeus

12-20-2007, 04:09 AM

Amazing :eek:

It seems the you always manage to outperform the other benchers in 3DMark01 while (almost) every tweak is already known. What your trick, mate? :D (If you don't want to share, ignore this reply ;) )

Make no mistake, 01 is highly tweakable, because every run can be ran seperately one has the ability to change LOD, CPU speed and even memtimings for each test individually.

Also, old school benchers always have some extra tweaks up their sleeve.

@Pedro, well done> :clap:

massman

12-20-2007, 04:14 AM

Make no mistake, 01 is highly tweakable, because every run can be ran seperately one has the ability to change LOD, CPU speed and even memtimings for each test individually.:

You can't change cpu speed between two test runs ;). Not sure about the ram timings.

Zeus

12-20-2007, 05:35 AM

You can't change cpu speed between two test runs ;). Not sure about the ram timings.

Took the test, ran all tests at 9x400 then changed fsb with setfsb and ran nature at 9x405:

http://service.futuremark.com/compare?2k1=9305132

The proof of the pudding is in the eating. ;)

massman

12-20-2007, 05:44 AM

3DMark used to give a speed error mistake when changing the cpu speed!

Could you please test at 9x425?

"This project can not be published due to detected CPU speed variation."

Zeus

12-20-2007, 06:06 AM

3DMark used to give a speed error mistake when changing the cpu speed!

Could you please test at 9x425?

"This project can not be published due to detected CPU speed variation."

I will do that when i have the time, it's just a trick you can use when you're running your cpu on the edge of stability so you can back down a bit for running nature for example.

Apologies for the topicstarter for off-topic talk.

k|ngp|n

12-20-2007, 06:17 AM

::woot: :woot:

I think im your biggest 3Dmark01 fan Pedro i just love that bench and always admired you results on it

:stick: :stick: mod the cards

I second that all the way :clap: :clap:
Congratz on making it to the 100k club with the ATI setup 2001 Legend!
I love the the angle of the cards :D

Pedro Rocha

12-20-2007, 06:38 AM

http://images.etsy.com/all_images/1/1c3/814/il_fullxfull.8716744.jpg

Even with stock, I think this score can be improved with better tweaking and more CPU speed (full run at 5.4Ghz)

And I have a "bad" QX9650 comming from US, that was tested by 2 friends, let's see how "bad" it is on the cascade :rofl:

Yes I change a few things between tests (LOD, performance level and even mem timmings), for some tests may be necessary to decrease CPU speed.

You can just change CPU speed just a bit (no more than 5% up or down) or you have a "CPU speed variation" error.

Normaly I start with lobbies with higher CPU speed and tight timmings and lower performance level possible (6 or 7), if it passes LH and LL pass everything.

Every CPU is diferent, with E6850 I need to change CPU between tests, wthis this wonderfull QX9650 is not necessary.

The 45nm Intel CPU's are really great :D

andre X_X

12-20-2007, 06:47 AM

http://www.funny-pictures.biz/face-surprise.jpg

OMG... you've done it again... congratz.
I'd love to see those cards on subzero, what would be the score? 110k?

N3RO

12-20-2007, 07:10 AM

Congratz Pedro ;)

Amazing Nature score for a pair of ATI graphic cards :D

Will be nice to see a 2001 run with 8800U SLI and 680i if you still have them ;)

[]s,

Turrican

12-20-2007, 08:21 AM

great result pedro :toast:

SF3D

12-20-2007, 09:20 AM

Great score!

Hmmm... How much I will lose in my own bench session next week? 100K would be nice.

You got PM :)

Pedro Rocha

12-20-2007, 05:46 PM

I second that all the way :clap: :clap:
Congratz on making it to the 100k club with the ATI setup 2001 Legend!
I love the the angle of the cards :D

Thank you Vince, coming from the current 3D King it that is quite an eulogy..I really admire your benching skills, persitence and the fact that you can bench with 3 LN2 pots all alone :shocked:

Eheh the angle of the ATI cards its quite strange in fact, but they work like that in 16x + 16x and at least there is some air flow :)

Congratz Pedro ;)

Amazing Nature score for a pair of ATI graphic cards :D

Will be nice to see a 2001 run with 8800U SLI and 680i if you still have them ;)

[]s,

Thank you N3RO :)

I still have a 680i SLI board but unfortunately they dont OC with the Yorkfields :(

780i OC work, but unfortunately seems not to be the best OC mobo due to big problems with high vcore...:rolleyes:

Anyway my current goal in 2001 is achived - prove that, with right teaking, was possible to achieve 100K with Stock Cards :D

If I find a CPU that can do 3D at 5.6Ghz with -105 (hope new double rotary cascade can hold -105 with QX9650 at 1.9v) maybe 1st place on 2001 on the ORB is possible ...I have some hope in a certain "bad" YF that is comming .;)

Amazing :eek:

It seems the you always manage to outperform the other benchers in 3DMark01 while (almost) every tweak is already known. What your trick, mate? :D (If you don't want to share, ignore this reply ;) )

No, no. I always share my tweaks (like I did with the ACPI tweak for AMD's). For me XS is all about knowledge sharing and fair play should be the 1st rule of benching.

Here it goes :
- PCIe 16x + 16x at 145Mhz (you need an ATA HD for that)
- Enougt voltage on SB
- Performance level = 6 or less
- Go to ATI drivers and put all details to performance and Catalyst AI to low level but enable
- LOD depends, but I sugest values from 3 to 4

And that's it, belive me the tweaks real change from one setup to another setup. Test your best multi / memory timmings and lod levels.

You also need to test the best runnig order, in ATI normaly I start with Lobbies, then CH, then Dragos, CL and finally Nature (in ATI, oposite to NvidiaSLI, you dont gain a lot in runnig Nature on 1st place)

But then again testing and discovering the best performance possible it's the must fun part of benching :D

That's why I call 3DMark2001 "my favorite game" - all optimizations counts, hope Futuremark never shut down 2001 Orb :)

andre X_X

12-20-2007, 07:44 PM

thank you very much for sharing your tweaks pedro....

if its a game for you, then you're giving us the cheats :D

Nordling

12-21-2007, 12:08 AM

Pedro Rocha
Use water cooling on VGA and do 105K! :up: :up:
P.M. CPU - real good!

Lightman

12-21-2007, 12:13 AM

OMFG :eek:

This is purely awesome!

Congrats Pedro Rocha! :bows:

BTW are you planning to do same thing with HD3870??

Pedro Rocha

12-21-2007, 04:12 AM

OMFG :eek:

This is purely awesome!

Congrats Pedro Rocha! :bows:

BTW are you planning to do same thing with HD3870??

Thanks :)

Yes I will, my next challange will be 2 stock HD3870's :D

With 2 stock HD3870's I have done this with CPU at just 4.550mhz air cooled of course:

http://pedrorocha.planetaclix.pt/ATI_3870_85698_Stock.jpg

Let's see how far will go with CPU in the mini-cascade ;)

Lightman

12-21-2007, 09:14 AM

I noticed that HD3870 Nature score is lower than on HD2900 at the same CPU clock. Memory bandwidth plays a big role here, but your cards are clocking way better on mem than mine :)

I haven't seen any vMem MOD yet in place, so if you can do that then 100k should be easy :D

Can't wait for more news :up:

PS. Try new Cat.12 drivers, they helping a lot in most CF scenarios! Didn't tried 3DM01 yet, but I will ;) .

EXPANDER

12-21-2007, 01:46 PM

3dmark2001 100k good.:up:

massman

12-22-2007, 01:54 AM

No, no. I always share my tweaks (like I did with the ACPI tweak for AMD's). For me XS is all about knowledge sharing and fair play should be the 1st rule of benching.

Here it goes :
- PCIe 16x + 16x at 145Mhz (you need an ATA HD for that)
- Enougt voltage on SB
- Performance level = 6 or less
- Go to ATI drivers and put all details to performance and Catalyst AI to low level but enable
- LOD depends, but I sugest values from 3 to 4

And that's it, belive me the tweaks real change from one setup to another setup. Test your best multi / memory timmings and lod levels.

You also need to test the best runnig order, in ATI normaly I start with Lobbies, then CH, then Dragos, CL and finally Nature (in ATI, oposite to NvidiaSLI, you dont gain a lot in runnig Nature on 1st place)

But then again testing and discovering the best performance possible it's the must fun part of benching :D

That's why I call 3DMark2001 "my favorite game" - all optimizations counts, hope Futuremark never shut down 2001 Orb :)

Thank you, Pedro :yepp:.

Are you running a FAT32 ATA disk? I guess yes, so what cluster size are you using?

Pedro Rocha

12-22-2007, 07:51 AM

Thank you, Pedro :yepp:.

Are you running a FAT32 ATA disk? I guess yes, so what cluster size are you using?

FAT32 ATA with default cluster size, the gain with lower cluster size is almost none.

G H Z

12-22-2007, 09:01 AM

Nicely done Pedro :toast:

I agree '01 is great fun, some people just don't have the patience for it ;)

Sampsa

12-22-2007, 04:52 PM

My first 3DMark2001 SE test session since 2002, Radeon HD 2900 XT CrossFire with stock cooling:

http://plaza.fi/s/f/editor/images/3dmark2001se_98133.png

http://service.futuremark.com/compare?2k1=9306496

I think it is a good start and 100k should be possible :)

SIOUX

12-22-2007, 05:20 PM

Thats a pretty good start Sampsa :toast:

Pedro Rocha

12-23-2007, 07:10 AM

My first 3DMark2001 SE test session since 2002, Radeon HD 2900 XT CrossFire with stock cooling:

I think it is a good start and 100k should be possible :)

That's very nice Sampsa, 100k should be easy :up:

Does your bord support 16x + 16x Pcie?? - that makes a big diference in Nature

How your cascade hold the temperatures of your QX9770 - idle and load temps on evap??

Pedro Rocha

12-23-2007, 09:36 AM

A very small improvement, more CPU but higher multiplier (12x) so less FSB and memory bandwith.

Here is it :

http://pedrorocha.planetaclix.pt/100359_2001_2900XTStock.jpg

Need better cooling on CPU and something more than ATI stock coolers for more..

Compare URL » http://service.futuremark.com/compare?2k1=9306892

Anyway I'm not in bad position in the ORB of My Favorite Game considering I'm using stock video cards :ROTF:

http://pedrorocha.planetaclix.pt/4Orb2001.jpg

My friend Ricky was so excited with this result that offer his triple stange and his good old double evap cascade in order we reach again the first place ..maybe after Christmas we will try ;)

nachthymnen

12-23-2007, 09:56 AM

A very small improvement, more CPU but higher multiplier (12x) so less FSB and memory bandwith.

Here is it :

Need better cooling on CPU and something more than ATI stock coolers for more..

Compare URL » http://service.futuremark.com/compare?2k1=9306892

Anyway I'm not in bad position in the ORB of My Favorite Game considering I'm using stock video cards :ROTF:

http://pedrorocha.planetaclix.pt/4Orb2001.jpg

My friend Ricky was so excited with this result that offer his triple stange and his good old double evap cascade in order we reach again the first place ..maybe after Christmas we will try ;)

Amazing score like always!!:up:
I can Imagine your score with 3 stage cascade:eek:

massman

12-23-2007, 09:59 AM

Ricky & Pedro will bench together again? :D

Pedro Rocha

12-23-2007, 10:09 AM

Ricky & Pedro will bench together again? :D

http://nc.startribune.com/blogs/randball/wp-content/uploads/00000000001yes.jpg

:rofl: :rofl:

MTP04

12-23-2007, 11:01 AM

Well done Pedro, looking forward to more results. :clap:

Are you planning on doing any runs with the cards under subzero?

George_o/c

12-23-2007, 12:52 PM

My God, you for sure know how to tweak ... ;) I can't even imagine the score with all the setup on liquid nitrogen ... :shocked: :eek: :D

Oliver

12-23-2007, 02:51 PM

Amazing score Pedro.

I really looking forward to see some more action with 3870s and bench session/s with Ricky.

Keep pushing it.

LardArse

12-23-2007, 03:02 PM

Pedro the 01 Gooorooo... :up:
Congrats my firend on a great job!

Pedro Rocha

12-23-2007, 04:36 PM

Well done Pedro, looking forward to more results. :clap:

Are you planning on doing any runs with the cards under subzero?

Thank you all :yepp:

Yes we are planning put the 2 cards with sub-zero, near -100ºC in Ricky's dual evap cascade - let's hope the ATI 2900's are not coold buged like 1900XT :confused:

Here us a photo of the setup used to break the 2001 40K barrier in the year 2004 (FX-53, ATI 800XT), was really entertaining to break that WR, and the 1st place last for a few months until the FX-55's came out ..

http://pedrorocha.planetaclix.pt/RickyDualCascade1.jpg

The cascade on right side now have dual evaps, one evap is very narrow should fit in Maximus 2 PCie 2.0 slots. This graphics cascade have 3 compressors - 1 1st, 2nd stages - let's hope still works after so many time stopped ;)

PPInDaHouse

12-23-2007, 08:01 PM

Last time me and Ricky used the GFX cascade (3 months ago) it was at -90c working on a 88GTS if I remember correctly. Hope you guys can break the actual 3dmark01 WR.

btw, Pedro your 3DMark01 knowledge is huge ;)

Kispo

12-24-2007, 06:21 AM

Go Pedro & Ricky! good luck for 1st place on 3d2001 :D
Merry Christmas!!! :)

Pedro Rocha

12-29-2007, 05:07 PM

OK, here is an improvement only due to better tweaking - I used the exact same CPU and GPU's speeds of previous run :)

http://pedrorocha.planetaclix.pt/100642_2001Stock.jpg

Compare URL - http://service.futuremark.com/compare?2k1=9310202

And here is a 3DMark 2005 run

http://pedrorocha.planetaclix.pt/35894_3DMark2005_Stock.jpg

Compare URL: http://service.futuremark.com/compare?3dm05=3845155

Near 36 not to bad for stock cards ;)

And finally here is a Aquamark run :

http://pedrorocha.planetaclix.pt/312620_Aquamark.jpg

Let's see how far this CPU will go with the mini-cascade big sister (2 big rotarys) that Ricky is almost finish for me - if it holds -100ºC 3D bench at 5.500Mhz may be possible, I hope :yepp:

massman

12-29-2007, 11:18 PM

Naughty, you broke 100k before you switched to subzero cooling on GPU's - again. I can't wait to see the results with Ricky :)

hipro5

12-29-2007, 11:24 PM

Yes......Pedro hits again.....!...... :D :up:

Sampsa

12-30-2007, 03:44 AM

Pedro, that CPU is very good or maybe it's Maximus Extreme, I'm not able to run my QX9650 or QX9770 3DMark2001SE stable even at 5,4 GHz (evaporator -102C) with Radeon HD 2900 XT CF.

Pedro Rocha

12-30-2007, 06:10 AM

Pedro, that CPU is very good or maybe it's Maximus Extreme, I'm not able to run my QX9650 or QX9770 3DMark2001SE stable even at 5,4 GHz (evaporator -102C) with Radeon HD 2900 XT CF.

You should try the Maximus Extreme, I gain 50-70Mhz 3D benchable from the Blitz. I suspect the gain is due to better voltages stability - try to measure non-moded Blitz with multimeter and you see quite a big drops in full load.

Let's hope with new dual rotary cascade can do near 5.5GHz benchable. The mini-cascade holds well the heat with 2 cores enabel (drops to -85ºc on evap in 2005 GT1) but not soo good with 4 cores (drops to -70 evap in 2006 CPU test). I will ask Ricky to calibrate the new cascade with 4 cores enable.

Don't know if my wife gonna like the change, the mini-cascade produce low noise, I suspect dual rotary cascade will not be and can wake up hear with near heartquake noise :rolleyes:

Sampsa

12-30-2007, 06:25 AM

You should try the Maximus Extreme, I gain 50-70Mhz 3D benchable from the Blitz. I suspect the gain is due to better voltages stability - try to measure non-moded Blitz with multimeter and you see quite a big drops in full load.

Let's hope with new dual rotary cascade can do near 5.5GHz benchable. The mini-cascade holds well the heat with 2 cores enabel (drops to -85ºc on evap in 2005 GT1) but not soo good with 4 cores (drops to -70 evap in 2006 CPU test). I will ask Ricky to calibrate the new cascade with 4 cores enable.

Don't know if my wife gonna like the change, the mini-cascade produce low noise, I suspect dual rotary cascade will not be and can wake up hear with near heartquake noise :rolleyes:

Pedro, I'm using P5E3 WS Professional but I'll give Maximus Extreme a try... too bad it has a poor PCIe x16 layout design.

My rotary cascade's noise level is 65 dB while running.. it is not too bad for the performance.

Pedro Rocha

12-30-2007, 06:28 AM

Pedro, I'm using P5E3 WS Professional but I'll give Extreme a try... too bad it has a poor PCIe x16 layout design.

Yes try it, you need to use my technique of bend the graphic cards (see 1st pic on thread) lets call GPU's in V formation :rofl:

VoRtAn_MaDgE

12-30-2007, 10:11 AM

Don't know if my wife gonna like the change, the mini-cascade produce low noise, I suspect dual rotary cascade will not be and can wake up hear with near heartquake noise :rolleyes:

http://www.cartoonstock.com/lowres/iba0114l.jpg

:D

blind_ripper

01-01-2008, 01:55 AM

pedro if been breaking my head about ure nature score !
if tested min @
QX9650 @ 10x405
Abit IX38 PCI-E 140mhz
2900XT cf @ 900core 980mem
and still im getting 872fps nature : thats a good 50fps lower then ure's .
i dont know but if tested LOD's, old scool ATI tweak's .
tough if tested on NTFS hdd that cant explain the nature :(
yes AI is on low , nature on 3.3lods gives a good 3fps more avrage .
lower cpu clocks cant be the reasson , pci-e can go higher but no need for that atm . board Cross-fire support could be poor or something like that .

if u know what im doing wrong or whats wrong feel free to pm me or replay .

Pedro Rocha

01-01-2008, 07:22 AM

Blind reapper

With ATI's and YF CPU speed really does count even in Nature - don't have many doubts that the 1.400 Mhz CPU diference can explain the 50fps diference.

That is quite diferent from Nvidia, where Nature is almost non cpu speed dependent

I used 148Mhz PCIe in the 16x PCIe 2.0 slots of the Maximus Extreme, that means a lot of bandwith.

When I changed from Blitz Extreme to Maximus Extreme I have a big jump also inn Nature, only 2x16x PCIe can explain that.

The rest is ATI driver tweaking, I'm also used Lod 3 in Nature..

Now put your CPU at -100ºC and get 100K - Happy New Year :D

blind_ripper

01-01-2008, 07:55 AM

Pedro Rocha

01-01-2008, 08:05 AM

will need a new cpu first :p: , only 4.99ghz with cascade and high voltage (wurst yf ever! ) LN2 only 5.25ghz :( .
cards pretty good so far : 910core 999mem in cf on box air :D , il look for maximus extreme or the DDR2 version .

when i think off it it could be the board that hass poor CF support , though its 16x16x pci-e 2.0

Yes, could be the board - already read reports on bad PCie 2.0 CF support with other motherboards (Gigabyte for example).

If you have DDR3 sticks, you should really try the Maximus Extreme - that is one of the best and most performance mobos I ever tested, in 2007 I tested dozens of mobos :D

The only bad thing is the single PCIe 2.0 space, but you can always use the ATI cards in V shape like I do ;)

For P35 / DDR2 the best mobo tested in 2007 was the Foxcomm Mars, when Shamino and his crew have the new X38 / DDR3 on the market that board gonna be a real blast, no doubt about it :D

blind_ripper

01-01-2008, 08:15 AM

Yes, could be the board - already read reports on bad PCie 2.0 CF support with other motherboards (Gigabyte for example).

If you have DDR3 sticks, you should really try the Maximus Extreme - that is one of the best and most performance mobos I ever tested, in 2007 I tested dozens of mobos :D

The only bad thing is the single PCIe 2.0 space, but you can always use the ATI cards in V shape like I do ;)

For P35 / DDR2 the best mobo tested in 2007 was the Foxcomm Mars, when Shamino and his crew have the new X38 / DDR3 on the market that board gonna be a real blast, no doubt about it :D
yeah il be waiting to on the X38 from foxconn , yeah but how in hell's name are u gone get a container between to cards ? :p:

il have a extreme in a week or so i hope ;)

jimmyz

01-01-2008, 08:36 AM

yeah il be waiting to on the X38 from foxconn , yeah but how in hell's name are u gone get a container between to cards ? :p:

il have a extreme in a week or so i hope ;)

I was thinking with one of these. http://www.plinkusa.net/webR1PCI-E1-1R.htm

Maybe he has a different solution though.

Zeus

01-01-2008, 08:38 AM

Wow, 148MHz PCI-e speed, don't try this at home kids. :D

I take it you're not using SATA disks or you're using a PCI sata controllercard?

blind_ripper

01-01-2008, 08:57 AM

I was thinking with one of these. http://www.plinkusa.net/webR1PCI-E1-1R.htm

Maybe he has a different solution though.
that looks idd nice :p .

Wow, 148MHz PCI-e speed, don't try this at home kids. :D

I take it you're not using SATA disks or you're using a PCI sata controllercard?

ata disk hu , i tested up to 145mhz so far here :) .

Pedro Rocha

01-01-2008, 10:15 AM

Wow, 148MHz PCI-e speed, don't try this at home kids. :D

I take it you're not using SATA disks or you're using a PCI sata controllercard?

PATA HD of course ;)

SATA HD's connected to onboard controlers barely passes 130Mhz or less :rolleyes:

blind_ripper

01-01-2008, 11:32 AM

haha found me a gpu tube for single slot cooling :p:, itsold but it will fit nice without the V card mod :p:

Pedro Rocha

01-01-2008, 02:07 PM

haha found me a gpu tube for single slot cooling :p:, itsold but it will fit nice without the V card mod :p:

Nice, nice ;) - Pictures please :)

andre X_X

01-01-2008, 08:46 PM

pedro have you tried 3dmark2k3?? i tested my CF on default, cpu 3.6ghz. mostly i got 65k but some times i got 70k... dont know what the problem here...

andre X_X

01-02-2008, 12:39 AM

tested already, my card able to bench on 901/999 1.25v gpu all 3dmarks but on 3dmark03 i have to run 847-860 on core to have 70k, if i run higher on core, i got 65k...

massman

01-02-2008, 12:44 AM

What desktop resolution are you using?

blind_ripper

01-02-2008, 02:50 AM

tested already, my card able to bench on 901/999 1.25v gpu all 3dmarks but on 3dmark03 i have to run 847-860 on core to have 70k, if i run higher on core, i got 65k...

with cards @ 890/980 i got 76.6k 03 :) cpu 4ghz though .

pedro il post pics later on , its one of the first tubes made here for SLI by jort .
its a perfect fit :p: , hope its can handel some load :D.

andre X_X

01-02-2008, 03:56 AM

What desktop resolution are you using?

1680x1050, does it really influence score? i'll give it a try on 1024x768... thx

with cards @ 890/980 i got 76.6k 03 :) cpu 4ghz though .

pedro il post pics later on , its one of the first tubes made here for SLI by jort .
its a perfect fit :p: , hope its can handel some load :D.
wow... i try 901/999, 901/980 cpu 3.7ghz only 66k, lowering GPU clocks only 70.8k... is it 4ghz yorkfield??

massman

01-02-2008, 04:19 AM

1680x1050, does it really influence score? i'll give it a try on 1024x768... thx

Not at first, but when you're rerunning the bench, your score will drop fast. That explains why you topped at 70k at first, but lost a lot of points afterwards.

All 3DMarks are best to be run on a desktop resolution which is equal to the standard benchmark resolution of the bench, especially 3DMark03 in which you can lose a lot of points.

blind_ripper

01-02-2008, 04:26 AM

yeah andre its a yorks
http://service.futuremark.com/compare?2k3=5666765
the 76.6k is on other disk witch i cant get to right now :( , but this is at 845core 999mem with the abit X38 board :p: . i know u have that board 2 use bios 12.B02 ;)

Pedro Rocha

01-02-2008, 05:51 AM

@ andre X_X

Sorry I never run 3dmark 2003 and dont plan to run it, because really dont like that bench - it's only GPU speed dependent and there is no art (tweaks) involved ..

Ricky said "2003 is a bench for ladies" :rofl:

Anyway 2003 now is for Tri-SLI :D

blind_ripper

01-02-2008, 05:55 AM

:( just got a call , have to return cpu and card's tomorrow :(.
at the end of the month il have a other cpu with maximus :)

Gautam

01-02-2008, 08:48 AM

How much do you guys gain in nature from PCI-E speed in total?? The last time I tested was on i975X, and it was really only about 30 fps going from 100MHz to 135MHz in CrossFire.

blind_ripper

01-02-2008, 09:04 AM

yeah gautam thats was 8x8x now its 16x16x :D , dont know really what the gain is now .

andre X_X

01-04-2008, 12:36 PM

from 120 to 135 i got 5fps on Nature... its on IX38 abit.

ZoLKoRn

01-04-2008, 12:42 PM

from 120 to 135 i got 5fps on Nature... its on IX38 abit.

135MHz what HDD you used IDE or SATA ? I can max 125MHz on SATA if over 125 mbo don't detect hdd :(

andre X_X

01-04-2008, 12:47 PM

IDE of course :D... i can bench on 145mhz too

blind_ripper

01-04-2008, 12:50 PM

IDE of course :D... i can bench on 145mhz too

love it or hate it got ide to can boot up to 160mhz but can only bench to 140 :p

ZoLKoRn

01-04-2008, 12:52 PM

wow 145MHz on bench... i'll try hehehe :D

Thanks buddy ;)

Gautam

01-04-2008, 02:36 PM

from 120 to 135 i got 5fps on Nature... its on IX38 abit.

That's about what I'd expect...yet for some reason I'm 150 fps too low? :confused:

Pedro Rocha

01-04-2008, 04:48 PM

from 120 to 135 i got 5fps on Nature... its on IX38 abit.

I have strong feeling that PCIe 2.0 implementation is better in Maximus Extreme than any other X38 motherboard.. dunno why but seems that way..

Never experience such high jump in Nature, as I do when I changed to Maximus Extreme, to bad it have only single space between the 2 16X PCIe's :cool:

Other terrible thing is coold boot, my retail QX9650 dont have any coold boot a near -95ºC with any P35 board I tested, but coold boot at around -75ºC in Maximus Extreme :rolleyes:

Hope Asus engineers can solve the coold boot problem in Maximus Extreme with some Bios change :confused:

andre X_X

01-04-2008, 10:28 PM

Pedro have you experienced on 2900 CF you got lower score if you GPU clock too high?? i mean, i can bench 901 core on every 3dmark but the best clock is still 847 and 860, but 847 still the best, especially on 3dmark06 and 03. on 3dmark06 i got 19.2k on 875 - 901mhz but surprisingly got 19.8k at only 847mhz (tested few times)... memory doesnt problem, 981mhz or 999mhz is the best for me...

blind_ripper

01-04-2008, 11:10 PM

andre is the multie fixed with bios 12.08 ?

@pedro yeah i vind it rater strange : becaus on the stock bios with the IX38 i get only 750fps cf and with a little update to a newer one i got 860fps , clearly its the bios of the boards not the boards them self i think .

so my first demand to manufactuers is get the cf support fixed :D

Lightman

01-05-2008, 02:36 AM

I don't know which HD2900 BIOSes you're using guys, but from my experience 2900Pro ones was faster in benches at the same clocks and was giving less shuttering in 3DMarks :)
(for the record I had 300points difference in 3D06 and I think it was due to better PCIe handling, besides on Pro BIOS I was able to bench @160MHzPCIe and XT BIOS only 135MHz on my mobo)

Pedro Rocha

01-05-2008, 08:37 AM

Pedro have you experienced on 2900 CF you got lower score if you GPU clock too high??

No, I did not experience any of this. :confused:

880-990 is more or less the max of my 2 ATI's 2900XT with stock - so I upgrade to 880-990 Bios and normaly dont touch GPU's speed during bench session ..

andre X_X

01-06-2008, 12:22 AM

this is my run with my setup:

http://aycu12.webshots.com/image/40411/2004612636006380446_rs.jpg

PCIE was on 120mhz since Dragothics hangs on 145mhz... last time 135mhz still ok for drago's, will try later.. the CPU is max, GPU still have some room, Tweaks only PCIe and LOD 3.5 all run, Catalyst on performance, Cat AI on Low all bench. only Car Low and Drag Low run at 4.9ghz, rest of them on 4.86ghz...

GLENBOY

01-13-2008, 10:57 AM

Pedro Rocha

01-13-2008, 03:29 PM

hiya pedro

thanks for sharing a couple of questions what would you say the performance difference is between the 3870s to the 2900xts , i see by most 3dmarks that the 2900xts still seem to outperform the 3800 series and whats your thoughts re tri crossfire in the future as i see you have the maximus extreme with 3 card slots, thanks pedro

In run with identical cpu speed (around 5.4Ghz) 2 stock 3870's with the upgrade bios and 900-1340 (against the 880-990 of 2 stock 2900XT's), and the results :

- In my favorite game : 97.6K agaisnt 100K - mostly due to much less Nature (820 on 3870's / 920 on 2900's)

- In 2005 : 34.750 - 3870's around less 1K than 2900's;

- In 2006 : 24.5K : 3870's around less 800points than 2900's

I would love to play tri crossfire with 3870's - the cards are cheap and very easy to OC even with stock coolers.

Question: does X38 boards with 3 PCIe gonna really support Tri Crossfire ??

If yes, maybe is a solution for my favorite game, since Tri-sli are most oriented to 2003 and I really dont like that bench -:(

Did Vince, Shamino, Fugger and rest of XS people run 2001 on Tri-Sli on CES ??

GLENBOY

01-13-2008, 04:31 PM

systemviper

01-13-2008, 04:53 PM

No, no. I always share my tweaks (like I did with the ACPI tweak for AMD's). For me XS is all about knowledge sharing and fair play should be the 1st rule of benching.

Now that sums it up, that is why I am here..:up: great job Extreme Superstazzzzzzz

best spot on the web......:yepp:

Pedro Rocha

01-13-2008, 05:05 PM

cheers pedro

it'll be interesting to see how tri crossfire pans out it may be more effective in 3d06,05 and 01 than the tri sli (but may not) , thanks for getting back , appreciated, were you using a modded bios with 1.35 volts on your 2900xts , i have it on mine and can bench at 900 on core (but with a few artifacts on a few benches ) or 893 no artifacts rock solid ,surprisingly on the stock volts(1.15 on mine) its still good to about 855 on core

I used the 2900's mod bios that Shamino put of Foxconn site, the 880-990 bios it is 1.25v, not 1.35v .

I do not advise anyone to use 1.35v core on 2900's with stock coolers, even at 100% fan speed (I cant stand that terrible noise) stock coolers does not hold 1.35v

hipro5

01-15-2008, 11:19 PM

Pedro my friend......You NEED about 1000MHz Core on the 3870s to "match" the 880/990 on 2900s....They can do it by moding their VCore with their DEFAULT heatsinks..... ;)

This is with the 3870s with their default heatsinks ON AIR....of caurse VCore moded.... :)

http://www.hipro-tech.com/images/hipro5/1_ASUS_P5E3_WS_Pro/RETAIL_York/2k5_37341_QX9650_5756MHz_ATI_3870_CF_999_1296.png

COMPARE 2k5 URL.... (http://service.futuremark.com/compare?3dm05=3899157)

At around 5500MHz CPU, I get a difference ~200 marks less on the 3870s than the 2900s......IF they could do ~1300MHz+ Core on LN2, I'm sure that they could outperform the 2900s on LN2.....I'll give it a try.... :)

PPInDaHouse

01-16-2008, 08:51 AM

soon we will start thinking who will be the first that breaks 40k on 3dmark05? :D

George_o/c

01-16-2008, 02:01 PM

Wow ! Nice result George ;) :up: Can't wait for the LN2 one ... :)

Ace123

01-16-2008, 02:40 PM

Good run.
Makes me want to go out and pickup a 2900 pro and see what I can do.

Pedro Rocha

01-16-2008, 04:26 PM

At around 5500MHz CPU, I get a difference ~200 marks less on the 3870s than the 2900s......IF they could do ~1300MHz+ Core on LN2, I'm sure that they could outperform the 2900s on LN2.....I'll give it a try.... :)

Very 2005 numbers my friend :up:

Please try to run my favorite game (2001) maybe with more GPU core speed Nature will be not so depressing low..

What PCIe you used ??
In Maximus Extreme can use 145Mhz or more with 2900XT but can not pass 130Mhz with 3870's :(

Pedro Rocha

02-11-2008, 11:43 AM

Here is some updates - QX9650 now on the dual rotary cascade but the 2ATI's 2900 still with stock coolers and no mods..

36K 2005

http://pedrorocha.planetaclix.pt/36006_3DMark2005_Abit5500_2900XTstock.jpg

Compare link: http://service.futuremark.com/compare?3dm05=3999561

101.706 in my favorite game:

http://pedrorocha.planetaclix.pt/101706_5580_Abit_2900XTstock.jpg

Compare link: http://service.futuremark.com/compare?2k1=9333107

Not fully tweaked yet - used LOD 3 for all tests

With better tweaking / lods, 102.xxx may be possible with stock cards in this DDR2 mobo ;)

In fact this was with a DDR2 motheboard (Abit IX38), now I'm testing a new DDR3 X48 motherboard that can give significative gains in system tests ..:yepp: