Printable View
I was trying to keep my HT Higher than 1000 MHZ...
Reached a 1300 MHZ HT and that was great ...
But what about performance ????
WIN OR LOSE at those high HT ????
I set the FSB at 250 and multi X 8 = 2000 MHZ CPU like default 3200+ speed...
Left my GFX 5900 card on 400 GPU - 900 MEM.....
And here is what i get......
http://www.xtremesystems.org/forums/...chmentid=14759
Strange .....
AT X2 LTD = 2X250 = 500 HT i get the BEST 3D Scores but at X5 LTD = 2X250 HT i get the BEST PI Scores....
I dont know what to say....
If its possible make some of your BENCHES and post it right here to make a conclusion :D :D
very interesting. did you run '01 and '03 a couple of times to make sure it was not just benchie variances
3 times per 3D Bench ;)
Here is an other view...
Marked green apps Highest score...
http://www.xtremesystems.org/forums/...chmentid=14760
Very strange indeed.
Excellent finding Dag, now only some with an Nforce3 250 would do the same, hell I would but my board took a crap on me.
nice work. i hope your card is usually oced cause its killing 3d performance. maybe ocing it would change the 3d results at higher hts?
that'd be a good idea except i know for a fact that my board doesn't have LDT of 5xQuote:
Originally posted by morbidj
Excellent finding Dag, now only some with an Nforce3 250 would do the same, hell I would but my board took a crap on me.
Anyone care to offer a possible explanation of what's going on there with 3D?
This is a totally un-educated guess, but maybe it influences the real time clock slightly?Quote:
Originally posted by xgman
Anyone care to offer a possible explanation of what's going on there with 3D?
awesome, so really, for the sake of lowering to overclock...there really isnt that huge of a concern for lower performance. Accross the entire board...they are fairly close.
3D Benches repeated again and allmost get the same.....
I dont know WHY but maybe its a mobo issue......
I will repeat those benches with different multis to see if anything change....:rolleyes:
i think its a chipset issue not a mobo issue
but thats real weird how 3dmark like low HTT
someone plz bench on nf3 250, and well see whats up with this
i cant wait to try this mobo!!!
i think that with a higher oc you may see some performnce increases at higher HTs more than lower. just a thought
very interesting! will make it a sticky :)
the HT speed is the bandwidth from the CPU to the AGP slot if I remember right...
maybe the videocard responds better to bandwidth similar to AGP8X bandwidth...
man that sounds like a brain fart, cause I can't bring the idea together....
Today just arrived my new ABIT KV8PRO 1.1 Revision Board....
I will repeat all benches from begining to see if those strange results appear again..... ;)
Very interestng results indeed.
It is possible that the BIOS is disabling some tweaks when the higher LDT multis are used eg Extended Control Time is one of them.
Dagalidis,
I will package up a special version of A64 Tweaker which will display these settings. It would be very interesting to see what is being used. Just give me a day or so to do some clean up :)
EDIT: try this version: A64Tweaker 0.5 XT beta . Not all of the feature work yet, but it should display your current settings correctly.
yes, interesting stuff !!
We need some results from NF3 ppl :), to compare to the VIA chipset...
Thank you my friend....Quote:
Originally posted by CodeRed
Very interestng results indeed.
It is possible that the BIOS is disabling some tweaks when the higher LDT multis are used eg Extended Control Time is one of them.
Dagalidis,
I will package up a special version of A64 Tweaker which will display these settings. It would be very interesting to see what is being used. Just give me a day or so to do some clean up :)
A64 Tweaker is VERY VERY good tool.... :toast:
It time to clear up whats going on here ..... Your Special Version is Welcome .... ;)
I just edited my previous post with a link to a debug build.Quote:
Originally posted by Dagalidis
Thank you my friend....
A64 Tweaker is VERY VERY good tool.... :toast:
It time to clear up whats going on here ..... Your Special Version is Welcome .... ;)
Lets hope it works :D
I will try late in the night when my rig is up again.....
Thanks...
joke on/
Quote:
Originally posted by Dagalidis
I will try late in the night when my ring is up again.....
Thanks...
" ... my ring ... " AHAHAHAH
ur getting married?? :banana4: :banana4: :banana4:
/joke off
:)
Quote:
Originally posted by Blaster
joke on/
" ... my ring ... " AHAHAHAH
ur getting married?? :banana4: :banana4: :banana4:
/joke off
:)
:D :D :D
Nooooo i'm allready married since 1990 :(
Too BAD ehhh ? :D :D :D
With your FSB at 250, and memory 1:1, and a 2X HT multi, you're getting DDR500 and an HT of 500. Just a guess, but could there be a benefit to having the HT match the memory speeds?
Program HT x1 HT x 2 HT x 4
3D2k01 21.871 22.138 22.207
3D2k03 7098 7130 7140
PC2002
cpu 7131 7121 7135
mem. 8840 8939 8996
Aquamark N/A 48,564 48,884
PC2004 4286 4363 4396
well i have tested with epoX KDA3+ (nforce3250Gb) with 4.27 latest nforce drivers i think...
amd athlon 64 3200+ c0 clawhammer 2202Mhz
Corsair BH-6 @ 2-3-3-7 200.4 Mhz
cat. 4.8 by Igor Pavlov (best driver i ve tested)
MSI RX9800 @ 438/378
i cant agree with dagalidis but since i saw his thread i thought i must bench too :D seems to do not as much on the epoX board i wouldnt say that its with all nforce boards so... (more reviews :D plz)
least but not last i want to say i only benched the 3dmarks once and i think the aquamark score on HTx1 would be little bit less than with HTx2
seEn andL
Nice Post andL64.....
I will try again with my new ABIT mobo 1.1 Rev and post some new findings.... ;)
Quote:
Originally posted by Dagalidis
Nice Post andL64.....
I will try again with my new ABIT mobo 1.1 Rev and post some new findings.... ;)
so plz do so!
we need more reviews i want to know if its on alle nforce3 like that...and also k8t800(pro)
and dagalidis i wanted to do this since a few times but i didnt recognize some real difference so i didnt looked at it exatcly but since i saw ur thread i MUST do it :D
i ll bench other benches like sisoft i think memory get s also a little bit better but why it didnt use HTT or does it?
we ll see and also i ll bench with the new "hell on earth" board if i can get one :D
soon i ll also get a pre tested cpu then i ll give ya some more...
seEn andL
PS: lets see how high ur HTT could get !?! i ran 280x4 without problems
...and how the hell i could post a chart , list ?
Was able to begin booting at 350 x 6 @ LTD X 3 but Crassed on Windows Loading SCreen.....
Damn it was to close but my mobo need Aircondition because i have 39 C Case temp :( and dont help for high FSB OC..... :D
Dagalidis
Hmm.. can you maybe bench same resulting HT speed achieved by different methods? For example 5x200, 4x250, 3x333 :) I wonder if that will make any difference @ all.
Last night was trying to find my max 1:1 free of errors using MEMTEST 1.5 as mem diagnostic tool....
I was able to hit the amazing 262 FSB using ONLY 2.62 Vddr and I was really socked….
I the most of my OC I used 2.8 Vddr as default option because I thought this was the only way to get my system stable….
I was wrong… my ADATA VITESTA DDR500 using Hynix –D5 chips just love low vddr …
I conclude that CPU mem controller is heating or don’t like more that 263 MHZ because no matter what I do this was the max acceptable OC for 1:1 FSB/MEM….
Just check my steps at the XLS table and make your judgment……
MEMTEST = 10 loops of 5th test
3D = AQUAMARK03 Bench
SPI = Super PI 2MB
It took me a lot of time to have long Tests....
http://www.xtremesystems.org/forums/...chmentid=15801
I will Run Tweeker to see what changes make my ABIT when change LTD when return home.....
ADATA is a very good perfomer MEM and those Hynix D5 chips are clocking well and with low VDDR as you see at my EXCEL..... :slobber:
wow those are amazing results! extremely interesting!
Yes.... MEM Controller sufer from heat and water is needed for higher 1:1 clocking .....
But ADATA is JUST GREAT :D
So you use #18? I mean the FSB/mem config. Very nice with 1T at those speeds with the volts! Is it not possible to tighten to 8-3-3, at least?
eh...
nice nice nice i heard of it that the a data likes low voltages ver very much and thell ll do not even mor with xtreme or higher voltage so...
but i got bh-6 how far can i go ? i think 3,2V shouldnt be a problem... i ll only get little bit nervous cause of the mem controller in side the amd64 :D
bu no risk no fun!
soOn awaiting the dfi board :D
seEn andL
Not only you.....Quote:
Originally posted by andL64
soOn awaiting the dfi board :D
seEn andL
Believe me .... lol
Not sure how you ran the test as far as order. Meaning did you run LDT 2, then 3, then 4, then 5 and rebooted before each run. If that's the case then reversing the order may give different results
Dagalidis great findings.
btw I think you mean LDT. ;)
Correct ........ Sorry for Mistake....Quote:
Originally posted by iceman2g
Dagalidis great findings.
btw I think you mean LDT. ;)
No....Quote:
Not sure how you ran the test as far as order. Meaning did you run LDT 2, then 3, then 4, then 5 and rebooted before each run. If that's the case then reversing the order may give different results
I was starting from LDT 5 and when not passed MEMTEST or Windows Benches then searching the best BIOS option to make system stable again...
So if needed LDT decreased back to X4 - X3 as you see at the EXCEL....
CodeRed Sorry for this late reply…..
I have screens of A64Tweeker of LDT X 5 & X4 but nothing changed…
I’ve also tested with LDT X3 – X2 – X1 but nothing new appears on Settings…
This is LDT X 5 Screen…
http://www.xtremesystems.org/forums/...chmentid=15939
This is LDT X 4 Screen....
and as you see ... nothing changed.....
http://www.xtremesystems.org/forums/...chmentid=15940
Here is Unbuffered Score of SANDRA MEM @ 260 1:1 @ LDT X 4 @ 2.6 VDDR :D
http://www.xtremesystems.org/forums/...chmentid=15941
Oh well it was worth a try :)Quote:
Originally posted by Dagalidis
CodeRed Sorry for this late reply…..
I have screens of A64Tweeker of LDT X 5 & X4 but nothing changed…
I’ve also tested with LDT X3 – X2 – X1 but nothing new appears on Settings…
This is LDT X 5 Screen…
There are a few more LDT registers, maybe I need to make another version ;) Or easier to just dump the values with WPCREDIT (bus 0, device 24, function 0).
Thanks for doing the testing.
Maybe I missed something I was refering to the 3d runs at the start of the threadQuote:
Originally posted by Dagalidis
No....
I was starting from LDT 5 and when not passed MEMTEST or Windows Benches then searching the best BIOS option to make system stable again...
So if needed LDT decreased back to X4 - X3 as you see at the EXCEL....
Just a small note.
Ive played with many Adata sticks in the past, and found that they have hated any more than 2.7v.
Try using 2.7v as your top limit; might help you reach 270-280 ;)
Hmmm...Quote:
Originally posted by Kinsy
Just a small note.
Ive played with many Adata sticks in the past, and found that they have hated any more than 2.7v.
Try using 2.7v as your top limit; might help you reach 270-280 ;)
Yes i know what you mean...
I found that 2.6 VDDR @ 260 1:1 is the best combination right now....
BTW i hope VTT mod will help to get more than 262 1:1 when watercooling be ready and push VDDR up to 3.2 Volts...
Nothing new about ABIT @ Via Chipset to POST....
But this weekend i will try my new DFI board and run same TESTs again to compare with old (Via Chipset)....
Wish me luck with DFI..... :D
dam it i think have DT43 chips and i need 3.2v to get to 270 1T.
that cant be stable? i only get 2200/2270 unbuffered :(
Dagalidis, about the abit, what is your opinion ? and NF3 vs via ? :p
:D
ABIT KV8 Pro is a good mobo but was really hard to make it work perfect...Quote:
Originally Posted by AndreAPL
Was a good performer with low cost and many overclocking options but had hardware problems (VTT Track) which make MEM overclocking limmited to 262 FSB 1:1 using ADATA Vitesta DDR500 with Hynix -D5 chips...
On the other hand DFI until now is the best mobo which i had, and the most easy to overclock....
DFI increased my overclocking potential up to 295 1:1 FSB using the same Vmem compared with ABIT.... :slobber:
Also my CPU reached stable 2500 MHZ using 1.76 Vcore VS 1.82Vcore used on ABIT :D
An other great + on DFI is the fact i can USE HIGH FSB like 357 x 7 = 2500 MHZ @ 277 MEM 1T rock stable...
So YES .... DFI is a great mobo and worth your attention...
here is SS Sandra benches with HTT 4x
Was a reason ever found why some people get better 3D scores with X2 HTT than with X4 HTT?
Bump could anybody answer the last question.
The reason could be that the Hypertransport link becomes unstable at those speeds, yet manages to keep from totally crashing because HT is a packet protocol and can resend packets (a bit like TCP). This causes congestion though, and lowers the 3D scores.
I would try increasing the LDT voltage at speeds > 1000MHz.
HT1.0 doesn't support retry. It doesn't resend errored packets like PCI-E does. Most times when an error occurs on the HT bus the system hangs right away, but not always.
Quote:
Originally Posted by Gogar
i can run 3dmark or any other bench under windows a few times WITH THE SAME SETTINGS and the results are always OFF , 1 or 2 percent plus/minus. Maybe even more.
Therefore i really dont care anymore about a few points up down in any bench under windows....except you do a whole lot of test runs (like 10 or 20) and then average them.
I am always amazed how people give attention to 3dmark/everst etc. scores when they see it's only like 50 or so pts higher w/ a different driver or setting etc.....this means literally NOTHING since there's always the margin up/down where it might differ from run to run anyway.
MAYBE one of the more reliable tests might be rightmark memory analyzer....but then again it also fluctuates SOMEWHAT.