Super-Pi 32M > i965 v. i680 . . .

[hipro5]

Hello all.... :)

Many of you have asked me to do a comparison in Super-Pi 32M between the i965 chipset and the i680 chipset.....
There you go.... :)

I used the EXACT SAME Hard Disk with the EXACT SAME tweaks....I didn't touch the HD so to be equal..... ;)

I used OBP's timings as for the i680 chipset and the rams that where used in both occasions were the Team extreme 1200s.... :)


ASUS P5B-Deluxe i965 chipset with rams at 2:3 and 600MHz (1200MHz DDR).....


http://www.thelab.gr/images/Hipro5/P5B_Deluxe/png/Super_Pi_32M_13.03.078_3600_Ram_600.png


ABIT IN9 32X - MAX i680 chipset with rams at 600MHz (1200MHz DDR).....


http://www.thelab.gr/images/Hipro5/ABIT_IN9/Super_Pi_32M_13.25.344_3600_Ram_600_i680_1.png


Now let's talk a bit about these results..... :)

Firstly if you put on i680 chipset the tRC = 1 , you gain BULL compared to tRC = 8.....
If you put tRRD = 1 instead of 2, you gain BULL out of it.... ;) :)
SO DON'T tell us that the "secret" of the NVIDIA i680 to get lower scores in Super-Pi is the tRC and tWR bulls..... :p:

When I played the Rams at 618MHz on the i680 chipset instead of 600MHz with the above timings, I gained about 3sec in 32M ~13.22m.....

OVERALL: YES the i680 chipset IS FASTER than the i965 chipset as for Super-pi 32M BUT playing our CPU about 100 - 150MHz MORE ..... :p: :D

Any comments will be appreciated.......PLUS ANY of you who has BOTH platforms, try it YOURSELF TOO and find out how "tricky" can be the tRC etc, etc......compared to i965 chipset..... ;) :p:
Bench BOTH 965 and 680 with the same HD (without Intel inf or NVIDIA drivers) and report here too.... :)

FINAL: DON'T tell us bulls about how faster is the i680 compared to i965 chipset coz you know it's .....just....bulls.....and don't try to "blind us" by showing different tables about ram timings and such things coz these are just for "closing people's eyes" and .....Show us FACTS.....I have a FACT here.....
I hope some others will bench with both platforms too, to comfirm this or not.... ;)
I'm talking in general here and not for any particular person.... :)

[VictorWang]

lol....bro u can have 2~3sec faster with 965s and 10~15sec faster with 680i.
all the same timings.

[hipro5]

lol....bro u can have 2~3sec faster with 965s and 10~15sec faster with 680i.
all the same timings.

I know, but I don't show benches here.....I show comparisons between them with the same config...... :p: Plus this is my 3D HD and I tweaked it a bit for the Super-Pi runs..... :)

PLUS YOU confirm too by your words that i680 IS slower than i965.... ;)


EDIT: Another thing is that in Super-Pi 1M you can get (no matter what platform) the EXACT SAME score by working it in SINGE DDR mode than in DUAL DDR mode.....Also you might get even BETTER resaults coz you can tighter MORE one single stick of DDR..... Try it..... ;o)

[KiD0M4N]

Erm, is it just me, or do i notice the 680i running 360x10 on a quad core? If you want a straight on comparision... run it with exact same parameters. Just try putting the X6800 in the 680i board and setting 400x9.

Just my 2 cents,

[hipro5]

Erm, is it just me, or do i notice the 680i running 360x10 on a quad core? If you want a straight on comparision... run it with exact same parameters. Just try putting the X6800 in the 680i board and setting 400x9.

Just my 2 cents,

OK....Would you like 450MHz fsb and 1200MHz rams on i680 to be a bit "better"....?.... ;)

[Viss]

EDIT: Another thing is that in Super-Pi 1M you can get (no matter what platform) the EXACT SAME score by working it in SINGE DDR mode than in DUAL DDR mode.....Also you might get even BETTER resaults coz you can tighter MORE one single stick of DDR..... Try it..... ;o)

Tried that a while back. Dual was only 0.048 faster compared to single in 1M which is less then i would have thought.

[Gautam]

The quad being a B3 will actually be slightly faster in pi than the B2.

Good test hipro, but I agree that it'd be nice to see 400x9 2:3 on the 680i, just to make things clearer to everyone. ;)

[hipro5]

The quad being a B3 will actually be slightly faster in pi than the B2.

;) :toast:


Good test hipro, but I agree that it'd be nice to see 400x9 2:3 on the 680i, just to make things clearer to everyone. ;)

I'm doing it now.... ;)

[Yoxxy]

I think to make good comparison, exact settings should be used. Use OPB speed and timings to compare.

[hipro5]

Update......

Now.....WHO can save me about 20 more seconds so to beat and overkill the i965 chipset?.... :D

http://www.thelab.gr/images/Hipro5/ABIT_IN9/Super_Pi_32M_13.16.735_3600_Ram_600_i680_1.png

http://www.thelab.gr/images/Hipro5/ABIT_IN9/Super_Pi_32M_13.16.735_3600_Ram_600_i680_2.png


Now we have an EXACT comparison between two of them.....No?...... :)

Save me now 13 + 7 more seconds MINIMUM to outperform the i965......Anyone?..... :p:

[gandalfone]

I will try with dfi dark and dfi 680LT at 400x9;) .....but....for me 965 is faster to 680...

Hipro what's volt mem to timing opb?? My team 1200 at 2.7v with dark go to 505Mhz 3.3.3.9....but that timing(opb timing with 680) not working:mad:

[hipro5]

I think to make good comparison, exact settings should be used. Use OPB speed and timings to compare.

First I - personaly - don't think OBP's 32M was done at 5420MHz...... :p: and please don't start...... I compare here the i965 v. the i680...... IF you want to "match" somehow OBP's score and convert it at 3600MHz, you have to show me (in general) a score at 3600MHz with the i680 chipset of about 12.55min and below......About his timings, yes they could be done..... ;)



I will try with dfi dark and dfi 680LT at 400x9;) .....but....for me 965 is faster to 680...

Hipro what's volt mem to timing opb?? My team 1200 at 2.7v with dark go to 505Mhz 3.3.3.9....but that timing(opb timing with 680) not working:mad:


I feed via a DDR Maximizer these Team Xtreme rams at 2.68Volts and I need about 2.71Volts for 618MHz with the above timmings...
tRC = 1 IS a joke.... tRC = 8 is more stable and performs EXACT THE SAME.... ;) It's like the bug - let's say - the ASUS P5B-Deluxe had with booting at Cas 4 and changing it to Cas 3 within Windows BUT the performance was like it was Cas 4..... Just a bug....and for "show off"..... ;) Also tRRD = 1 is the same.....Doing it = 2 is the same....no gain out of it.... Anyone out there could try that.... :)

[zbogorgon]

try virtualization, it shouldn't be much of a problem :banana:

[Yoxxy]

To be honest it is never going to be possible to prove any of this stuff.

Everyone is in a crappy position.

You could have booted CAS 5 from Bios, switched to CAS 4 on 680i (which would do nothing) and booted 965p direct. There is just going to be too many variables. The only way you are going to prove beyond a "shadow of a doubt" is by having everyone DV cam any WR run.

[zbogorgon]

To be honest it is never going to be possible to prove any of this stuff.

Everyone is in a crappy position.

You could have booted CAS 5 from Bios, switched to CAS 4 on 680i (which would do nothing) and booted 965p direct. There is just going to be too many variables. The only way you are going to prove beyond a "shadow of a doubt" is by having everyone DV cam any WR run.

it is quite possible to prove !!! :slapass:

[Enoc]

it is quite possible to prove !!! :slapass:

yeap..you should prove it...with your wr core 2 duo extreme...5,420mhz should be a peace of cake with your cpu + a i680 board with 1,200mhzrams...
remember to post a vid at those frequencies ...:rolleyes:


Many of you have asked me to do a comparison in Super-Pi 32M between the i965 chipset and the i680 chipset....

nice comparo Hipro5 :clap: , but there is no need to throw names..is this a comparo o a pursuit?
and the big thing about superpi32m is TWEAKS and BANDWITH...for you to compare 32M pi you have to have the same tweaks...or ask in the XTreme Pi forum after japan legend Newbettle shared some of his tweaks how many seconds a lot of guys have improved...let me say to you this..ALOT...and you forget OS, HD etc..it helps...
so..if you do a 5420mhz run, same os, same hd , same tweaks...then...you should compare...if not...then you can not compare my friend...;)

you guys should end this...just reunite, bench and learn from each other :) ...

this is a chase that has no end my friend...

[jmax_oc]

Thank you Hipro for all these explanations.
I didn't use 680i chipset and this is an other reason to confirm my choice.

For this 680i, I thing that the only way to really improve the RAM benefit is not is tighten timings but in sending more datas before raw or lines changes.
However, I don't know if it's possible to do this in nf680i (burst mode at maximum lenght)

[elmor]

Running 400x9 with 680i just doesn't cut it. Try 420x9 on both platforms (clockgen from 400 on both ofc), because bandwidth scales alot with fsb on 680i.

[K404]

tRRD might be a "dud" timing on 965 and 650/680I...i`m not noticing any effect.

Async Latnecy is also a dead timing on 650/680I, at least on Asus- I dont think they bothered setting up a 100% relevant and unique BIOS for the chipset...too busy adding more boards to their product lineup.

[manu]

what ram do you use?

[macci]

Running 400x9 with 680i just doesn't cut it. Try 420x9 on both platforms (clockgen from 400 on both ofc), because bandwidth scales alot with fsb on 680i.
...and it doesnt with the 965 then? Are you saying that at 420 the 680 would somehow magically be faster? That by boosting the FSB by another 5% things would dramatically change?
By "Bandwidth" what kind of bandwidth are we talking about? SAndra?

[SewerSide]

so..if you do a 5420mhz run, same os, same hd , same tweaks...then...you should compare...if not...then you can not compare my friend...;)



this is exactly what hipro has done, taken 2 mobo's, using the same ram, hd, os install and run the same tests, and has posted his results.:stick:

[eva2000]

maybe tRC = 1 versus higher value depends on motherboard ? on my Asus 680i striker extreme tRC 1 was definitely better than 8.

my last 3600mhz 32M on Striker extreme was

with Team 6400C3

http://fileshosts.com/intel/Asus/680i_striker/results/X6800_1/team/6400c3/0402/9x/2_3divider/400-600_5333_212-7-4-24_2t_1.525-1.35-2.6-1.75-1.6-1.35/superpi-32m_single_13m15s812ms.png

with OCZ PC2-8000 Titanium Alpha VX2

http://fileshosts.com/intel/Asus/680i_striker/results/X6800_1/ocz/titaniumalphavx2/8000/0402/9x/unlinked/2t/2_3divider/400-600-5333-2-1-2-7-2-22_1.475-2.525-1.35-1.75-1.55-1.35/superpi-32m_single_13m13s859ms.png

playing with P5K Deluxe now :D

At certain ram dividers 680i MAY beat 965P since it has unlinked mode though. But anyone knows 965P tweaked at same clock speeds will beat 680i chipset.

[Enoc]

I need to learn how to read! :stick:

he cannot compare his results with OPBs cause he needs to run same cpu frequency 5420mhz, same os, same hd , same tweaks as OPB , THEN he can compare...it's all i'm saying...this is not superpi1m...:rolleyes:

[dinos22]

hipro you are wasting your time bother :D...............can't beat P965 with 680i :D

LOL elmor what kinda comment is that ... you even woke up macci to give you a slapper ROFL

[erwinz]

ah.. happy place.. :D

[hipro5]

SO......Till NOW ALL of us agree that i965 CAN'T be beaten by i680 as for Super-pi 32M right?.... :)

@ eva2000: Your time is good BUT NOT 12.55min and below.....Try again with the i680.... ;)

[dinos22]

SO......Till NOW ALL of us agree that i965 CAN'T be beaten by i680 as for Super-pi 32M right?.... :)

@ eva2000: Your time is good BUT NOT 12.55min and below.....Try again with the i680.... ;)

i agree

1M is not an issue but 32M no chance :D:D:D:D:D

are you trying to prove your point from a different angle you sneaky bugger :p: :D

[hipro5]

i agree

1M is not an issue but 32M no chance :D:D:D:D:D

are you trying to prove your point from a different angle you sneaky bugger :p: :D

:D ;)

[kiwi]

Next is to compare P35 :)

[elmor]

...and it doesnt with the 965 then? Are you saying that at 420 the 680 would somehow magically be faster? That by boosting the FSB by another 5% things would dramatically change?
By "Bandwidth" what kind of bandwidth are we talking about? SAndra?

Nono, I'm not saying it will be faster. I'm saying that 680i memory bandwidth (as shown by sandra/everest or any other memory benchmarking software) scales very well and it would be interesting to see if 965 scales as good or if there is a gap. I'm basing this on the fact that 680 shows higher memory bandwidth than 965 (in benchmarks) and that the fsb might become a limiting factor for 680i.


hipro you are wasting your time bother :D...............can't beat P965 with 680i :D

LOL elmor what kinda comment is that ... you even woke up macci to give you a slapper ROFL

The end justifies the means :D

[Gautam]

I'll try for 12:55 in a bit, can 680i beat me there? :p.

[eva2000]

SO......Till NOW ALL of us agree that i965 CAN'T be beaten by i680 as for Super-pi 32M right?.... :) depends try doing compare of 9x266FSB 32M with 965P vs 680i with max mem frequency/tight timings that the memory can handle..

on 965P 266fsb 2:3 or 1:2 is best you can hope for 533mhz ram clocks.. 680i what dividers are open for ya :)




@ eva2000: Your time is good BUT NOT 12.55min and below.....Try again with the i680.... ;)
i can't get 12m 55 even on 965P lol

[Gautam]

If you make that 12x or 13x 266 I'll take you up on it.

[dinos22]

i can't get 12m 55 even on 965P lol

and neither can George errr i mean hipro5 George George :D:D:D:D

[gandalfone]

Gautam you have a very good O.S!!;)

[hipro5]

and neither can George errr i mean hipro5 George George :D:D:D:D

I don't need to do 12.xxmin with the i965......I need to do below 12.55min on an i680...... Can someone do it.....except...well......you know..... :D

@ Gautam: Do you have a i680 mobo to test it too M8....? :)

[DStealth]

I don't need to do 12.xxmin with the i965......I need to do below 12.55min on an i680...... Can someone do it.....except...well......you know..... :D



Neighbour,
The man has invented "Perpetuum mobile"
http://xtremesystems.org/forums/showpost.php?p=2172417&postcount=9
And you are judging him for such a small thing like 3dmarks and slower chipsets being faster :stick:

:toast: Goog work George. Мarvellous tweaking Guatam :toast:

[Gautam]

Thanks. :)


@ Gautam: Do you have a i680 mobo to test it too M8....? :)
No sorry, not atm...leaving the country for a couple of weeks...

[Misko]

I don't know what are u exactly want to improve, but let take some numbers sec x mhz are first number and in () are cpu speed and fsb/ram speed u can calculate dividers by your self:

3029671 k|ngp|n (5261 404/607)
2799531 (3603 400/600)


3039140 kyosen (5210 521/521)

2796161 NewBeetle (3590 399/598)

2984537 Victor Wang (5000 417/625)
2815879 (3750 417/625)


2821112 My team (3600 514/514)
3076056 (5400 540/540)


and finally WR done by OPB:


3002537 OPB (5425 417/624)

If we take a look on subzero result we can see that Victor Wang has a better perf number than OPB so i680 isn't so good like 965 on 32M (because Victor was using XP if I'm correct and it is slower that 2k3), but it isn't i680 what bring OPB WR, it is pi value output, here is picture my team vs. OPB:

http://img204.imageshack.us/img204/3006/untitled1yf8.jpg

Every single ocer in the world except OPB has pi value have the time of last loop about 80-84% at the average time of 1-20 loops. Even OPB himself took 82% on his 32M Run (ex WR) with i975X Mobo. Now his pi value output is lowered to 66% :slobber: , so i680 is still slower but he has some great tweak or i680 is capable to do some pi value output that other boards can't. I'm intensively working to find out that tweak but nothing still :(. Maybe it could be done if u put entire os in ramdisk (not virtual ramdisk just like Gigabyte or OCZ have that PCI cards u put DDR ram in it and it acts like hdd), but OPB have told me that he is using regular hdd. So maybe i680 is slower than 965 but he has some tweaks that other ocer's don't have.

@hipro5 did u try tRFC 19 on i680 what's the gain?

[hipro5]

@ Misko: Look.....Here we don't talk about OPB and his "tweaks"..... Here we want to prove that the i965 is DEFINATELLY faster than the i680 chipset.... ;)

Gain with 19 tRFC is about ~1sec or less.... It couldn't give me another -20sec..... :)

Gigabyte iRAM is going slower than a regular HD.... ;)

[Misko]

@ Misko: Look.....Here we don't talk about OPB and his "tweaks"..... Here we want to prove that the i965 is DEFINATELLY faster than the i680 chipset.... ;)

Gain with 19 tRFC is about ~1sec or less.... It couldn't give me another -20sec..... :)

Gigabyte iRAM is going slower than a regular HD.... ;)

10x m8 for iRAM info ;), :toast:

[Gautam]

Heh, yes i-RAM is a bit slower than a normal HD, but it depends on how you use it. ;)

You can't use VictorWang's performance product as a comparison as his multiplier is lower.

You might as well use this to compare :p:

http://www-personal.umich.edu/~gautamb/2330_1858.PNG

(2652858.1336) :lol:

Instead:

Compare k|ngp|n- 5261.7 x 575.797s = 3029671.0749

hipro - 5403.3 x 564.172s = 3048390.5676

esdee - 5135.1 x 598.406 = 3072874.6506

joe cool - 5006.0 x 614.734 = 3077358.404

These are all results on the 13x multiplier with the 2:3 ratio...the only way you can make a valid efficiency comparison.

[Misko]

You can't use VictorWang's performance product as a comparison as his multiplier is lower.



I agree but on the other hand Vic is using XP, so it is not bad ti see his perf number also.

[Gautam]

Windows XP and Server 2003 are practically identical in 32M. In fact in some cases XP is faster.

[ryba]

Hipro5: Did you used a CW in your test?

[andre X_X]

no way 680i can beat 965 @ spi, specially 32m.... in 3dmark 680i might win, but not in spi, even 965 could beat 975x. just googling and find out review about 680i vs 965, none of the result said that 680i faster than 965 in spi32m.....
Thats just NonSense.... its just like NetBurst vs Intel Core, theres no chance netburst can win...

[Hazzan]

Just To Compare i965 straap 1066 ( ABIT AB9 Quad GT ) and Straap 1333 ( ASUS P5B deluxe ) :) :) use all same hardware OS Driver..........:) :) :)

SISOFTSANDRA 2007 on Memory Bandwitch AB9 QUAD GT :

http://img.techpowerup.org/070515/AB9 QUAD GT Sisoft.jpg

SISOFTSANDRA 2007 on Memory Bandwitch ASUS P5B Deluxe :

http://img.techpowerup.org/070515/ASUS P5B DLX 500 SISOFT.jpg

Spi 4M AB9 Quad GT :

http://img.techpowerup.org/070515/AB9 QUAD GT Spi 4M.jpg

Spi 4M ASUS P5B dlx :

http://img.techpowerup.org/070515/ASUS P5B DLX 500 Spi 4M.jpg

SPi 32M AB9 Quad GT :

http://img.techpowerup.org/070515/AB9 QUAD GT Spi 32M.jpg

SPi 32M ASUS P5B dlx :

http://img.techpowerup.org/070515/ASUS P5B DLX 500 Spi 32M.jpg

:) :) :)

[andre X_X]

pak hazzan, please compare you abit i680 with either 965 board you have...
btw, those abit surely fast with 1066 strap :toast:

[Misko]

pak hazzan, please compare you abit i680 with either 965 board you have...
btw, those abit surely fast with 1066 strap :toast:

But he doesn't have i680 abit, quad GT is 965 c2 intel

[VoRtAn_MaDgE]

975 - 965 - 680i period!

I can even see difference in 1M (personal experience)...talking about the 32m is really a walk in the park to beat 680i with same cpu clocks on a 975/965 platform.

[shaolin95]

This makes me rethink my upgrade mindset....I had already figured the 680i was the better mobo but it seems I was wrong... Decisions...decisions....

[andre X_X]

But he doesn't have i680 abit, quad GT is 965 c2 intel

well misko, he has it ;D
he's my friend here

[Misko]

well misko, he has it ;D
he's my friend here

Sorry didn't know for that one ;)