Super Pi 1m scores

[Alec§taar]

WTF! Sorry... upsetting! Double-post...



(If you don't upload your photos on the init. posting? It treats edits as quotes!)

APK

[Alec§taar]

Editing & Reposting Current Results...

Apk

[giorgos th.]

the score you gave i think its the previous at 2826 right?

[Alec§taar]

Originally Posted by giorgos th.

the score you gave i think its the previous at 2826 right?

I think so... I have been doing SO many of these? I can't tell anymore, lol...



* It's been a "benchmark A.M." here by ALL means...

APK

P.S.=> 2840mhz now, using 258FSB x 11 mult. currently... I will post scores one more time with correct/current Super-Pi 1.5 mod in minutes... apk

[Alec§taar]

33: Noted your post on last page, almost missed it... I will reply shortly! apk

[Alec§taar]

Originally Posted by _33

That ram should be cranked much more. If you could play with your ram multiplier and just leave it at first to 3-3-3-8, but with ram going at 240mhz, it would give you under 30 second score. And basically you should get close to 1500 score on Sciencemark 2.0 with your dual core. And if you crank it even more say 275mhz ram, you'll get around 28 seconds or less.

I will take your thought, to heart here (and, also to mind & "work-in-practice" as well).

It seems my system has STILL more "potential" in it, via o/c & RAM-memory timing chains adjustments...

See BELOW (edited it out last page, put here, for YOUR reference) for my latest!

Thanks, as it still may give you MORE ideas on what I ought to tune up/tweak out, here.

Originally Posted by _33

BTW, I don't know about you but I have 2 GB of ram... It's not exactly SuperPI related, but I was wondering how you could work with a Windows XP system and 512MB ram?

First off: I don't use Windows XP - I use Windows Server 2003 (MSDN subscriber here in past), not that it matters really: I make up for memory usage LARGELY thru various .reg hacks &/or services + background process trimmings anyways, on ANY Win32 based OS.

I place my pagefile.sys onto a CENATEK "RocketDrive" 2gb Solid-State RamDisk board, on its first partition, & it ONLY, on said partition.

(Will get FASTER & BETTER, once the DDRDrive x1 PCI-Express Solid-State drive releases, because it uses a faster bustype (PCI-Express x1 slot (iirc, 1,000mb/sec, blows AWAY PCI 2.2 bus the CENATEK uses @ 133mb/sec) & also faster RAM type (DDR, vs. PC-133 SDRAM the CENATEK rocketdrive here uses))

Since the OS views ALL MEMORY, as "Virtual Memory"?

WELL, here, when my RAM in chips on the mobo gets flooded/exhausted?

I am paging to RAM yet again, albeit to an SSD (not a far slower HDD based pagefile.sys)...

It is like an "auxillary memory board" when used w/ a pagefile.sys placed onto it - albeit, the bustype (PCI 2.2/133mb-sec) hurts it on "BURST" speeds the most imo.

This is where the DDRDrive x1 PCI-e model will absolutely BLOW IT AWAY & make it faster than the far older (circa 2001-2002) CENATEK "rocketdrive" SSD can do currently!

(Then, the rocketdrive by CENATEK will go back into where it came from -> My 2nd rig, Intel P4 3.2ghz H/T CPU, 512mb DDR-400 RAM, & WD "Raptor" 36gb 8mb buffer 10k rpm disk (it is my SQLServer 2003/IIS 6.x lab practical rig here) & the DDRDrive will take its place in THIS system)

Originally Posted by _33

Maybe one day I'll get an Opty 165, but man 512MB ram that's kind of on the edge for windows XP! After booting up, I am left with 1650MB free, so with 512MB I'd have maybe 150MB free.... Just a tought.

Maybe I will as well, those Opterons are SERIOUS performers... but, that means buying ECC RAM too though (but I am sure my mobo WILL take either).

APK

[Alec§taar]

REPOSTING TO LAST PAGE, best I can do...

APK

[_33]

Originally Posted by Alec§taar

I will take your thought, to heart here (and, also to mind & "work-in-practice" as well).

It seems my system has STILL more "potential" in it, via o/c & RAM-memory timing chains adjustments...

See BELOW (edited it out last page, put here, for YOUR reference) for my latest!

Thanks, as it still may give you MORE ideas on what I ought to tune up/tweak out, here.



I place my pagefile.sys onto a CENATEK "RocketDrive" 2gb Solid-State RamDisk board, on its first partition, & it ONLY, on said partition.

(Will get FASTER & BETTER, once the DDRDrive x1 PCI-Express Solid-State drive releases, because it uses a faster bustype (PCI-Express x1 slot (iirc, 1,000mb/sec, blows AWAY PCI 2.2 bus the CENATEK uses @ 133mb/sec) & also faster RAM type (DDR, vs. PC-133 SDRAM the CENATEK rocketdrive here uses))

Since the OS views ALL MEMORY, as "Virtual Memory"? When my RAM in chips on the mobo gets flooded/exhausted, I am paging to RAM yet again, albeit to an SSD (not a far slower HDD based pagefile.sys).



Maybe I will as well, those Opterons are SERIOUS performers... but, that means buying ECC RAM too though (but I am sure my mobo WILL take either).

APK

Hmmm, nice little thing you have there, the CENATEK module. I never heard of such thing for pagefile. Good idea! The only problem I see with this item is slowing down your PCI bus a chunk, ever so often when pagefile is accessed. That could be a hassle with sound card and other more crucial cards such as firewire or video capture (like I have - eg wintv).

Now I was wondering, where did you read that an Opty absolutely needs ECC ram? Link appreciated. Thanks. for now let's say I just don't think that's a required asset, unless proven otherwise.

[Alec§taar]

Originally Posted by _33

Hmmm, nice little thing you have there, the CENATEK module. I never heard of such thing for pagefile.

It works, & will work better/faster, using the DDRDrive x1 PCI-e model, by far on DDR400 RAM (vs. PC-133 SDRAM) & MOSTLY, on bustype imo @ least using PCI-Express x1 slot (vs. PCI 2.2)

Originally Posted by _33

Good idea!

Yea, I wrote some reviews (for CENATEK on their homepage, "An independent users review" in 2003) & articles on the use of Ramdisks, in general!

BOTH software AND hardware types...

(HARDWARE (for CENATEK review) & SOFTWARE - While doing paid contract work for EEC Systems when I wrote up some code for their SuperCache-II product & did reviews on "creative uses of ramdisks" as the article title & also writing up a competing ramdrive program based off MS-DDK template years before they had one out).

(The SuperSpeed.com/EEC Systems article USED to be on EEC/SuperSpeed.com website, & is now altered as to the page for it as well as the material (some) by their staff, doubtless accounting for changes in hardware since I wrote it, WAY back in 1996 iirc)

SuperCache-II is a better diskcache than the native NT-based OS one by far by the way, look into it.

That was all done back as far as 1996 & made it into Windows NT magazine (currently Windows.NET mag) w/ a REALLY strong/solid/good review no less!

(Yes - I am VERY proud of that one here!)

& the idea took off worldwide from there... hence, why companies like HyperOS & GigaByte (IRAM) are doing business vs. the CENATEK company now.

ON using SSD's:

It's NOT "original thought" though, because SSD usage has been around for servers & such or high-end workstations for a LONG time!

(E.G.-> quantum/maxtor iirc, used to make them too, but FAR more expensive ones, costing thousands, as far back as I can recall around 1995 or so)...

HOWEVER, how I apply SSD's (and to some extent, even SOFTWARE ramdisks can have a pagefile.sys on them, IF you have more than 4gb of RAM that is (this IS the 'catch), depending on how much RAM a PC has on its mobo in chips) to PC's?

THAT truly IS pretty much "my original thought" though...

I.E.-> NOW? Well, many others use this trick/tip/technique, & it works (for much more than the pagefile.sys, see my signature below in regard to the CENATEK "rocketdrive" & you will see what I mean & have done as "work-in-practice" using it to its FULL extent here, for how I use MY system personally).

Originally Posted by _33

Now I was wondering, where did you read that an Opty absolutely needs ECC ram? Link appreciated. Thanks.

Always was under that impression... not true, I take it?

(I have asked here more than once about that, but never got an answer!)



* GOOD if so, & NOT true... ECC ram costs more iirc!

APK

P.S.=> Thanks for correction on this note (ECC RAM & Opteron CPU use), if needed, & also for the tips on the RAM timings... it WOULD be NICE, to reach past say, Dj Dn (Opteron 165 user, & 1446 score vs. my 1441 (approx)). on the ScienceMark 2.0 tests especially IF that is doable as you state! I will try it... apk

[Alec§taar]

Originally Posted by _33

The only problem I see with this item is slowing down your PCI bus a chunk, ever so often when pagefile is accessed. That could be a hassle with sound card and other more crucial cards such as firewire or video capture (like I have - eg wintv).

I don't have ANY other PCI cards in my system - it is the ONLY 1...

In fact, I thought about what you're noting, years ago, & constructed THIS pc to account for that, with this card being the ONLY PCI-bus expansion card present in THIS machine!



(I do have WinTV32 though, but external USB model!)

APK

[infrared]

Originally Posted by Ketxxx

lol i just cuda swore i saw you touting some duo clocks a while ago

Nah, was a Pentium D i borrowed to play with. Only a crappy 805 though, so no records broken

[_33]

Originally Posted by Alec§taar

I don't have ANY other PCI cards in my system - it is the ONLY 1...

In fact, I thought about what you're noting, years ago, & constructed THIS pc to account for that, with this card being the ONLY PCI-bus expansion card present in THIS machine!



(I do have WinTV32 though, but external USB model!)

APK

Still off topic, but I was wondering; does CENATEK also do the 3 1/2 bay version of that board in IDE? I tought I read such thing not so long ago, but maybe it's not CENATEK. Anyway this thing must cost a fortune! Solid State disk example

[Alec§taar]

Originally Posted by _33

Still off topic, but I was wondering; does CENATEK also do the 3 1/2 bay version of that board in IDE? I tought I read such thing not so long ago, but maybe it's not CENATEK.

That's GigaByte's IRAM GC SATA II model iirc & iirc, also the HyperOS III unit too...

Here on these forums?

We had a HUGE discussion on them here, & this is HOW I got wind of the DDRDrive:

DDRDrive's company owner (Chris George) wrote me here in "PM" in fact, while we were discussing the Gigabyte unit.

He stated he remembered my work around SSD's in the past (and software ramdisks, I wrote one up YEARS ago in the mid 90's ("APK Ramdisk") based off the MS-DDK template for them) & articles, etc. regarding HOW to apply them.

He let me know, to "hold off" until his unit releases, regarding getting a newer/better SSD based on current technologies!

Yes - it will be the "SUPERIOR WARRIOR" imo, vs. Gigabyte/HyperOS III (all out there currently) & mainly imo, because of the PCI-e bus usage, more than DDR/DDR2 use would & mainly, for burst "read/write" I/O.

APK

P.S.=>

Originally Posted by _33

Anyway this thing must cost a fortune!

Not really: I got a 1/3rd pricebreak for the review, because once I did it & submitted it to them, noting I did REALLY like it & THAT I WANTED TO BUY ONE?

CENATEK liked what I wrote & I asked if I could get a break on the cost of it (they did so)... ended up costing me about what a high-end vidcard does nowadays... & I can use it in any system w/ a PCI 2.2 bus slot! apk

[Alec§taar]

33, scan this, offer your thoughts please in regards to your suggestions above/earlier regarding RAM timings:

DDR-400 RAM user here & using dividers (DDR333):

CPU HTT Frequency = 4x

DRAM CONFIGURATION:

CAS# Latency Tcl = 2
Min RAS# Active Time Tras = 6T (down from 8T, currently for this Super-PI 1.5 mod test)
RAS# to CAS# Delay Trcl = 3T
Row Precharge Time Trp = 3T
Row Cycle Time Trc = 9T (down from 16t (down from 12T (down from 10T))
Row Refresh Cycle Time Trk = 14T
Read to Write Trwt = 2T
Write Recovery Time = 2T
1t/2t = 1t

JUMPER FREE CONFIGURATION DATA:

O/C Profile = Manual
CPU Frequency = 257mhz
PCI Clock = 101mhz (upped from 100mhz default)
DDR Voltage = AUTO (cannot do over 2.6v on my RAM, tried it already 2x)
Chipset Voltage = 1.6V (max)
HT Voltage = 1.25V (max)
CPU Mult = 11x
CPU Volt = 1.5V (max)

-------------------------

33, what do you recommend changing (that is my BIOS data, currently, above)... you too giorgos.th: OR anyone that thinks I can eke more outta my rig!

APK

P.S.=> Still learning the "fine art" of o/c'ing... & could use your assist here 33, or giorgos.th, because you have gotten me gains already on THIS test (giorgos.th) & sounds like I can get more on ScienceMark 2.0 per 33...

Yes, this IS largely WHY I chose this forums in fact, over others out there: TO LEARN in an area I let go "way slack" the last 7 years or so - hardware-know-how (concentrated largely on coding Win32 mostly in THAT timeframe, & time to 'catch up' now).

IMO, your (folks here) hardware know-how excels over others I have seen online + folks here are generally a LOT cooler than some other 'hardware-fiends' forums imo @ least... thanks! apk

[giorgos th.]

Alec use this and take a screenshot with the settings when you run the fastest sp 1m.

[intel igent]

@ alec have you tried HTT 3x?

[Alec§taar]

Originally Posted by giorgos th.

Alec use this and take a screenshot with the settings when you run the fastest sp 1m.

I used it in the "My overclock" thread (when Tatty One, POGE, & Ketxxx first taught me about o/c'ing modern rigs here @ these forums around when I first arrived here 4-5 months back (MAN! Time flies)).

LOL, on o/c'ing:

I was doing it as far back as 486 days circa 1991-1992, but it was done WAY differently then, than it is now (replacing clockcrystal/oscillator/strobe used for mobo timings)... that, or throwing jumpers or switches on the mobo on later machines still (circa 486 VLB era-> Pentium I/II times).

Things are a LOT different now, especially on AMD stuff (this rig's my first AMD since 1998, when I had a K6-III @ 450mhz, Super-7 ASUS mobo type)) vs. Intel stuff I was most likely to use on a PC (all intel's since 1992, until this one, & the one I note above which was older AMD).

Things ARE a LOT diff. now, especially regarding memory timings. I am learning though, this IS the main point... you guys guide me, I go off on my own after that, & then you turn me onto more, I try it... & so on!

Hands-on tests: Nothing like it in the WHOLE world - "The University of Life"... & yes, sometimes, the 'school of hardknocks' too!

Still?

I would prefer to approach it now from the BIOS level, omitting softwares & their subdrivers used (another layer of filtering, CPU-use, & complexity in the way)... iirc, A64 Overclocker tool uses Frank Delattre's CPU-z drivers though (some trivia I learnt from experimentation w/ it in the thread I mention above).



* Anyhow/anyways - take a peek @ the above BIOS data (yourself, or 33 better yet, because he sounds VERY confident in what he stated above)... I could use it, & so could others IF we "stumble" onto a BETTER "ram timing chain" than I am using above currently!

APK

[Alec§taar]

Originally Posted by intel igent

@ alec have you tried HTT 3x?

Nope... didn't want to "step down" to DDR266 levels is why, but, from what Ketxxx & others basically told me?

Doesn't REALLY matter - you mostly have to be concerned w/ data CPU-z 1.36 gives you back in its RAM page (DDR400 = 200mhz, so you want to be MOST concerned with how much you get back of that (or exceed it as I have thusfar @ 201.9mhz) 200mhz DDR400 rating, when using "dividers" methods (stepping back on HTT multiplier & DDR type))

Am I wrong on that account?

If so??

Heck -> "SCHOOL ME!"...

APK

[giorgos th.]

i dont know anything about your ram but give a shot at this.
in the mem clock frequency dont put it in 166 mhz but use yours.
and give a little more voltage if you like f.e. 2.7V

[Alec§taar]

Originally Posted by giorgos th.

i dont know anything about your ram but give a shot at this.
in the mem clock frequency dont put it in 166 mhz but use yours.

Understood, & will do, replacing (reposting again for YOUR & 33's reference) what I have in my BIOS currently in these settings:

DDR-400 RAM user here & using dividers (DDR333):

CPU HTT Frequency = 4x

DRAM CONFIGURATION:

CAS# Latency Tcl = 2
Min RAS# Active Time Tras = 6T (down from 8T, currently for this Super-PI 1.5 mod test)
RAS# to CAS# Delay Trcl = 3T
Row Precharge Time Trp = 3T
Row Cycle Time Trc = 9T (down from 16t (down from 12T (down from 10T))
Row Refresh Cycle Time Trk = 14T
Read to Write Trwt = 2T
Write Recovery Time = 2T
1t/2t = 1t

JUMPER FREE CONFIGURATION DATA:

O/C Profile = Manual
CPU Frequency = 257mhz
PCI Clock = 101mhz (upped from 100mhz default)
DDR Voltage = AUTO (cannot do over 2.6v on my RAM, tried it already 2x)
Chipset Voltage = 1.6V (max)
HT Voltage = 1.25V (max)
CPU Mult = 11x
CPU Volt = 1.5V (max)

-------------------------

The GOOD part of all of this imo @ least? Is 2 things:

1.)That IF we find a better/faster rate here?? I can apply it potentially @ least to ScienceMark 2.0 as well (and my daily use of my system here as well, albeit that is usually below a 258FSB, lol, for stability in ALL tests & activities))

&

2.) Others w/ AMD Athlon64 x2 DualCore rigs will be able to hopefully/potentially as well... especially if they are 'victims' (lol, as I am, but thru my own purchases & volition) of DDR-400 memory that's NOT "overclocking ready"...



* BRB, going to test!

APK

[_33]

Originally Posted by Alec§taar

Understood, & will do, replacing (reposting again for YOUR & 33's reference) what I have in my BIOS currently in these settings:

DDR-400 RAM user here & using dividers (DDR333):

CPU HTT Frequency = 4x

DRAM CONFIGURATION:

CAS# Latency Tcl = 2
Min RAS# Active Time Tras = 6T (down from 8T, currently for this Super-PI 1.5 mod test)
RAS# to CAS# Delay Trcl = 3T
Row Precharge Time Trp = 3T
Row Cycle Time Trc = 9T (down from 16t (down from 12T (down from 10T))
Row Refresh Cycle Time Trk = 14T
Read to Write Trwt = 2T
Write Recovery Time = 2T
1t/2t = 1t

JUMPER FREE CONFIGURATION DATA:

O/C Profile = Manual
CPU Frequency = 257mhz
PCI Clock = 101mhz (upped from 100mhz default)
DDR Voltage = AUTO (cannot do over 2.6v on my RAM, tried it already 2x)
Chipset Voltage = 1.6V (max)
HT Voltage = 1.25V (max)
CPU Mult = 11x
CPU Volt = 1.5V (max)

-------------------------

The GOOD part of all of this imo @ least? Is 2 things:

1.)That IF we find a better/faster rate here?? I can apply it potentially @ least to ScienceMark 2.0 as well (and my daily use of my system here as well, albeit that is usually below a 258FSB, lol, for stability in ALL tests & activities))

&

2.) Others w/ AMD Athlon64 x2 DualCore rigs will be able to hopefully/potentially as well... especially if they are 'victims' (lol, as I am, but thru my own purchases & volition) of DDR-400 memory that's NOT "overclocking ready"...



* BRB, going to test!

APK

Actually the element that I'm missing to try to guide you is the memory divider. Let's say your CPU is running at 2800 even. If you had a memory divider of 10 (for example), then your memory would be running at 280mhz, thus ddr560. But it doesn't exactly work like that. The "very intelligent" bios writers decided to screw our brains into thinking there are no CPU to MEMORY divider and just throw out a bunch of MEMORY frequency, that, but *pure hazard* are directly influenced by CPU speed, thus LDT speed. So in essence you have to find the meaning of your memory divider and hoàw to set it up and get proper clock rates. I found that the latest version of cpuz is quite useful and seems just in giving the memory speed (they added HTT and proper voltage display).

I suggest just dropping those tight latencies in favour of a solid memory overclock. When you reach maximum overclock with slack latencies, maybe then try to tighten a bit and squeeze maximum out of them. If not satisfied, revert back. But I suspect you should get much better memory speeds. Try testing with sisoft sandra memory benchmark and Everest when manipulating your memory parms. See whaét you get as speed gain and latency tests.

[Alec§taar]

That was "too stiff" of a tightening giorgos.th...

E.G. #1 -> FIRST: I couldn't boot, got POST test onscreen video corruptions, etc. & could not pass POST...

E.G. #2 -> SECOND: had to lessen it again in areas - specifically:

Min RAS Active Time (Tras)

Row Cycle Time (Trc)

It also appears that Row Refresh Cycle Time (Trw) HUGELY affects me also (13 would not do it, not @ 258FSB @ least), so 14T again.

(All reset in BIOS by having to lessen those specific memory timings again to what is below now):

DDR-400 RAM user here & using dividers (DDR333):

CPU HTT Frequency = 4x

DRAM CONFIGURATION:

CAS# Latency Tcl = 2
Min RAS# Active Time Tras = 7T (down from 8T, currently for this Super-PI 1.5 mod test)
RAS# to CAS# Delay Trcl = 3T
Row Precharge Time Trp = 3T
Row Cycle Time Trc = 8T (down from 16t (down from 12T (down from 10T))
Row Refresh Cycle Time Trk = 14T
Read to Write Trwt = 2T
Write Recovery Time = 2T
1t/2t = 1t

JUMPER FREE CONFIGURATION DATA:

O/C Profile = Manual
CPU Frequency = 258mhz FSB
PCI Clock = 101mhz (upped from 100mhz default)
DDR Voltage = AUTO (cannot do over 2.6v on my RAM, tried it already 2x)
Chipset Voltage = 1.6V (max)
HT Voltage = 1.25V (max)
CPU Mult = 11x
CPU Volt = 1.5V (max)

-------------------------



* I can now get into Windows, & will test Super-PI as well... this IS tighter by some margin than what I had before though by all means... in the areas noted above.

APK

P.S.=> Off to test now - brb, w/ results... apk

[Alec§taar]

BEST I COULD DO BELOW GUYS:

CPU-z 1.36 CPU-mhz data (258FSB (upper limit that I know of previously)):



CPU-z 1.36 RAM-mhz & timings data (using TRAS = 7 & TRC = 7):



Super-Pi 1.5 mod Score:





* BEST I COULD DO, USING THAT CPU-z 1.36 data noted above, & THIS BIOS DATA:

DDR-400 RAM user here & using dividers (DDR333):

CPU HTT Frequency = 4x

DRAM CONFIGURATION:

CAS# Latency Tcl = 2
Min RAS# Active Time Tras = 7T (down from 8T, (6T-8T is VERY 'touchy' here) & usually go 8T)
RAS# to CAS# Delay Trcl = 3T
Row Precharge Time Trp = 3T
Row Cycle Time Trc = 8T (down from 16t (down from 12T (down from 10T)) 7 vs. 8 here touchy
Row Refresh Cycle Time Trk = 14T (will NOT do lower using 258FSBx11mult)
Read to Write Trwt = 2T
Write Recovery Time = 2T
1t/2t = 1t

JUMPER FREE CONFIGURATION DATA:

O/C Profile = Manual
CPU Frequency = 258mhz FSB
PCI Clock = 100mhz (backed off to default from 101mhz)
DDR Voltage = AUTO (cannot do over 2.6v on my RAM, tried it already 2x)
Chipset Voltage = 1.6V (max)
HT Voltage = 1.25V (max)
CPU Mult = 11x
CPU Volt = 1.5V (max)

APK

[alpha0ne]

As per sig with 1.50Vcore though not exactly 100% stable

1M

[Alec§taar]

Originally Posted by _33

Actually the element that I'm missing to try to guide you is the memory divider. Let's say your CPU is running at 2800 even. If you had a memory divider of 10 (for example), then your memory would be running at 280mhz, thus ddr560. But it doesn't exactly work like that. The "very intelligent" bios writers decided to screw our brains into thinking there are no CPU to MEMORY divider and just throw out a bunch of MEMORY frequency, that, but *pure hazard* are directly influenced by CPU speed, thus LDT speed.

LOL! Sounds like my garage mechanic (guy who does work I can't on my ride, when I don't have tools, lifts, or just plain know-how experience) does, regarding engineers who design cars!

Sure, they get better/more efficient, but... harder to work on year after year (and I agree), harder packed/more tightly, & more cross members to remove & bolts, etc. & worse.

Originally Posted by _33

So in essence you have to find the meaning of your memory divider and hoàw to set it up and get proper clock rates. I found that the latest version of cpuz is quite useful and seems just in giving the memory speed (they added HTT and proper voltage display).

HTT first entry I believe, is LDT stuff you mean ->

CPU HTT Frequency = 4x

(DDR400 -> DDR333 & multiplier on the unified memory & system bus controller on this new AMD stuff - very unlike Intel stuff I am used to w/ NorthBridge-Southbridge layout).

BUT, I am learning, or trying to!

Originally Posted by _33

I suggest just dropping those tight latencies in favour of a solid memory overclock. When you reach maximum overclock with slack latencies, maybe then try to tighten a bit and squeeze maximum out of them. If not satisfied, revert back. But I suspect you should get much better memory speeds. Try testing with sisoft sandra memory benchmark and Everest when manipulating your memory parms. See whaét you get as speed gain and latency tests.

That's what Ketxxx, POGE, & Tatty One also say, & I am inclined to agree - look for a GOOD solid balance with stability.

For this test, today? I was willing to mess around is all... worth it, in the long haul, imo @ least!



* I am done w/ this test, happy w/ my results thusfar, so... time to concentrate on ScienceMark 2.0 again (which typically means backing off my 258FSB, & also doing SOME "loosening" of those RAM timings once more too)!

APK

P.S.=> Still, it appears from my results @ least? That TIGHTER RAM TIMINGS are your buddy in Super-PI testing as I had heard on the grapevine out here online... apk