Can i do better

[NEOAethyr]

Here's to me willing to bet I get a 280 prime stable shot by the morning .

 

Here's what I got so far, I'm using my adata right now to prime 280x8.

The samsung tcc5 is not ready just yet to do this in dc, I need to find out specifically if tREF is an issue.

However now I'll have 2 more timings to worry about once I'm done testing the tcc5 by it's self, which I'll do after priming.

After modding tAC and tCK to death, no joke, went beyond the specs for both at one time or another.

tCK of 0 freaks the memory out though.

tAC of 0 is ok but slower, it's a very inportant timing to have just right, otherwise you'll be slowed down from it.

I got it perfect though, good old superpi, in single chan mode, it one heck of a program to bench with, I can see the diff off 0.01 ns with that program hehehe, if I avg out the time of multible runs and such by eye it's easy.

 

Anyways, here's my ref for 280 so far, so freaking close, priming now at 1024-4096k fft's.

 

 

;;;; 280x8 @ 1.5v VCORE

 

VMEM: 2.8v

Vdd: 1.7v

 

[tAC](TCC5): 0.65ns ?? (Needs more testing)

[tAC](ADATA): 0.21ns

[tCK](TCC5): 5ns ?? (Needs more testing)

[tCK](ADATA): 0.1ns

 

Burst Length: Page ( = 2m??, in PAE mode)

 

[tREF-PIC]: ?? (Needs more testing)

[tREF](TCC5): 3.9us ?? (Needs more testing to see if can go higher)

[tREF](ADATA): 62.5us

 

[tRAS]: 11

[tRCD-R]: 4

[tRCD-W]: 2

[tRP]: 3

 

[tCL]: 3

CMD: 2

 

[tCS]: 0

[tWE](TCC5): 0

[tWE](ADATA): 0

 

[tRC]: 9

[tRFC]: 14

 

[tDOE]: 0

[tRRD]: 0

[tWTP]: 2

[tWTR]: 3

[tREXT]: 3

[tRTP]: 0

[tRTW]: 3

 

??:

[tQHS]: 0 ?? (Needs more testing, but I don't know anything about this setting)

[tDQSQ-Max]: 0 ?? (May or may not make any diff since it's a ref only to go by for the board, not the actual timing, I would still like to know what it is)

 

;;

 

Notes:

Unsure if tREF is per stick.

If the highest is used the ADATA value will screw over the TCC5's stability.

If the lowest is used then so beit, but it'll slow down the ADATA.

 

Unsure if tCS is per stick, never messed with it

No way for me to find out either since it's as low as it gets by stock.

 

Unsure if tAC & tCK is per stick or not.

This would be pretty darn inportant.

 

CMD does'nt have a proper name, I need to find that out...

 

With the above timings in single channel mode, with the ADATA...

Passes Superpi 32M with ease.

Fails at the end of s&m's memory test, normal length, at test #4 with 2 erors mid way before I stoped it.

Likely passes prime95 normal sized fft's, of 1024k-4096k, will test.

 

;;

 

;;;;

 

 

Btw, anyoen think this is nuts?

I went way beyond the norm to get all of this setup, and for instance 0.1ns for tCK is'nt in the specs for ddr , it exceeds it's absolute min by 0.9ns ^^.

tCK is used for ref by all timings in some way, either it's added to them or it's multiplied by them.

Though I could'nt boot at 0, in single chan, at 280 or by holding insert to boot, I was able to se the memory in s&m fter holding insert with the tcc5 in there too.

Normally the memory is just not there if you go to low, I did it with tWE before my memory could handle it even at 275.

So I'm thinking it's not multiplied, however I think it's supposed to be multiplied though..., so I was able to run my memory with all 0 timings...., ohwell.

 

You guys got some of these new timings on the nf4's?

I'm wondering if you guys do for ref.

[kitfit1]

What i meant Neo, is that if you manufacture a chip and later convert the chip to something else, surely the cost of conversion must be higher than the cost of manufacture. Add the two together and that makes for an expensive cpu.

[NEOAethyr]

Yeah you're right.

 

The original name was for those odd mobiles.

Reading up in the central brain's readme pdf, it talks about it, but it pretty much was meant for the actual orignal name...

However it tells you to look at the version #'s, and apperenlty there were thorten's with 512k of l2, very strange though, I don't know the story on that.

 

Anyways my cpu is a ver 11.

The name however, that:

Athlon XP SFF Proccessor

 

It's just very strange, meant to be 3 diff cpu's...

At the begining either a thorton or this mid range (wierd one) mobile.

Both really, but I don't know what was determained 1st...

 

Then at the end, rebadged as a barton core and a higher end mobile at that.

 

Everything else though just gives off the model id's, the old 6a0(mobile xp) and 7a0(mobile mp).

The 7a0 being form the bios reading, because I modded the name.

 

I wonder if it's because of that...

It could be.

 

In any case the orignal name was the SFF or whatever, diff mobile, then it was changed to something else afterwards.

Like I said, probaly the pcb, the chip was likely tested before placing it on the pcb.

The id is very likely changed by the default voltage and the mobile bridges.

 

 

Strange stuff though.

I will try it one day, removing my cpu name mod and see what happens to version #.

 

Btw, 6hrs and 40mins priming so far at 280, not a single error .

I think I may of got that 280.

[kitfit1]

Defo an odd thing though Neo, perhaps they were just trying to rake up Mobiles from whatever they had that could do the job.

Anyway, glad the 280 is nearly there Don't forget to post us up a screenie when you get there

Still stuck at 270 myself, But i'm not gonna let this bugger stay there if it's the last thing it does

[NEOAethyr]

, 280, well it failed , joking lmao...

 

I'll attach a screeny.

Later on I gotta get my tcc5 to do the same :.

Then figuer out how to get them both to doit at the same time.

Now if both can doit by them selves, it's not hard to doit dualchan, however since I've been messing with the spd's, and if one of these timings happen to get set for both, one that is diff on both sticks, well, there goes stability...

Still, it's only a matter of time I think.

 

Anyways gotta attach that screen .

Same timings as the previous post of course, it's not like I could reset and change them while priming.

 

 

 

280, hahaha...

With the modified spd too, it's freaking fast as all heck.

Faster then dual chan at 275, gaming wise and general windows wise anyways.

M

It's more then 5mhz off bandwith wise in benchs.

Even with all the stuff I modded, I have'nt checked for lower timings int he bios because I know I shoud'nt try until I get dual chan working right.

So there's no point in making single chan any faster from the bios settings.

 

I wonder if I can go higher eventually ^^.

For now I'll take 280 though, and I'll stick with it until everything is 24/7 stable, at higher cpu speeds and etc.

It's allready light out, sunny, my room is'nt hot but it's not cold either.

So I dn if I can continue more testing, maybe priming and superpi, but probably not s&m.

I'll try anyways in a few though.

[kitfit1]

:drool: :drool: :drool: Yaheeeeeeeee, Looks like a new NF2 record :drool:

Well done Neo, well pleased for you man, the doggies danglies or what :nod:

[NEOAethyr]

I'm very pleased myself, however it's not quite there for everyday use.

 

Shoot, if I did'nt know about s&m, I'd probably prime it another 8hrs, if that passed, I'd doit again but at high ambient temps(even if I had to turn on the furnace, which I've done before at 266).

 

Then I would of called it good.

However, liek 266, it may of failed s&m still.

For 266 I needed more vmem.

Otherwise after a few hours of farcry I'd get a video hitch(a temp hardlock).

 

Maybe after a bit of hammering it'll come around.

But for now, this stick is all good, good enough for now.

I have to move on to the tcc5 though.

Get that stable.

 

From there, more breakin, and perhaps another bios.

 

If I knew what those 2 other timings were about, plus there's a few more I don't have timing names for, I may beable to doit straight up without hassle.

 

 

[tQHS]

[tDQSQ-Max] (tDQSQ is what I need info on anyways)

Got them both at 00h in the spd.

ADATA defaults.

 

Here's the other offsets and timings in the spd that I need info on.

I need real names, and info, these are hat I got set for them anyways:

 

F: 01 ;SDRAM Device Attributes – Minimum Clock Delay, Back-to-Back Random Column Access

 

20: 60 ;Address and Command Setup Time Before Clock

 

21: 60 ;Address and Command Hold Time After Clock

 

22: 40 ;Data Input Setup Time Before Clock

 

23: 40 ;Data Input Hold Time After Clock

 

 

The rest of the stuff I need to finish modding are min and max values.

Not really a big deal, because I've allready got them int he bios or I've set the actual timings them selves int eh spd allready, like tAC and tCK, got both min and max for those.

Alot of general timings are the same way, there's min and max values incase the bios wants to provide options for them, ie auto by default stuff.

Not really inportant though at all.

 

tDQSQ-Max

Another one not that inportant, probably.

But I dont' have this in the bios, nor is there a setting for it in the spd, I think.

Unless one of those unknowns is it.

 

That's about it for the adata.

 

For the tcc5, I need to finish tREF.

Then tAC, then tCK.

 

tWE, tAC needed breakin btw.

I use lower fsb's and 1m superpi to break them in to get to higher speeds.

then I did the same at 280 a few times, then ran 32m, worked out just fine.

I had to bench(used superpi) tAC though a few times to figuer out the optimnal though, because lower did not = better.

 

tCK, it was fussy(not allways booting 280 before I did a 32m superpi).

However it was'nt much of an issue.

I really did'nt have that much of a hard time booting lower values.

Unliek the others that I did have probs with, those would refuse to boot at certain speeds no matter how I tried booting them, not until a bit of breakin.

 

The burst length, enabling page, that was another tough one.

Had to break it in.

Page could equal the page size setup by the cpu, ie:

x86 standered: 4k

PSE: 4m

PSE, compatibility mode: 4k

PAE: 2m

 

Or it could be 16 bytes, I dn which.

Or more..., lol, I dn, I think it means the page size of the cpu though.

Probably explains why it's hard to use.

 

, man tons of thigns to do.

Tonight is likely when I'll do the tcc5.

I'll try doing a bit of it though in a little bit before I header off to bed.

If I can find out what boots and what does'nt, I can go from there later when I get some rest and when it cools down.

 

I'm pretty happy how my cpu is being cooled now though after the redo and modding the pins again.

Since I spread them further, and apperently did a good job appling the gel, it's been doing extremly well, better then ever before.

I'm very happy with it.

Perhaps I won't need to redo the gel once a month anymore, man that would be a god send :.

 

 

Oh and btw dude.

The 1st thing I would check fo timing probs on that samsung is tREXT.

At higher speeds, they hate it.

You may need to go higher.

At 275 with my old @ss tcc5, I needed 3, which sucked big time in single chan mode.

Not a biggy at all in dual chan though.

May be a good idea to try single chan.

 

Find out the min of all the timings of each stick, that's how I got 275x8 the 1st time I did it.

To tight though and the memory will actually start asking for alot of vmem and vdd.

Totally wierd.

Only in dual chan though, and both of my sticks do the same exact timings in single chan with normal voltage levels, but in dual chan, 1.8v vdd and 3.0 or 2.9v vmem, tight timings at 266.

Sorta tight, by this type of memory's standereds anyways.

[kitfit1]

Been away for a few days so i hav'nt done anything. Yes i will have alook at tREXT Neo. I was also thinking about giving it a go in single channel as well, because in DC i'm not that far behind you.

[Clay]

280 fell!!!!LOL I knew it would! Congrat's "Neo" .

Sorry I havent been around lately been takin' care of mom and overcluckin' my Ahh ahhh aaaaashhhussshhhhhhhh!!!Scuse me :tooth: It looks like 242 fsb is about where she quite's w/ that I still gotta' memtest those sticks on my nf4 but I think they are good. They memtested for 8hrs perfect(All test) on the other brand nf2 @240X10 but I still have to test on my other rigg to be sure. It needs a bigger psu fer' sure and a bunch of other stuff. Hehe well I gotta get to werk. Best to ya's "Clay"

[Sharptooth+]

I get this with an AN7 and modded bios (Enduracell/Tictac 18de81t). I'm in the same boat as you except that instead of 255fsb, anything above 235fsb makes 3D gaming an issue. I tried many things including changing agp and pci latencies, voltages and other timings. Nothing helped resolve the 3D issue until I dropped the fsb to 235.

 

--Fernando

[Clay]

@Sharptooth: I was expectin' it to do abit better. the T1 modded bios uses "DFI" Rom strips. But I'm not sure which ones he put in.("Trats") I'm still lookin' for the site that had archieved all the modded bios for all diffrent mokes and models.boards.(Found it once) I think it's a french site. I've read at "XS" that they have sumptin' like that but I haven't found it yet. It just passed memtest on the ahh aaahhhh ahhhhssssshhhhussshhhh!!! Scuse me :tooth: LOL a7n8x-e deluxe @242x10(8hrs) I'm sure I can get 240X10.5 but the vcore only goes to 1.825v. And @245X(??)+ fsb it has problems. I think 240X10.5 will fly but 242X10.5 is reaching my processor really wants juice much past 2.5ghz. I don't think its the memory but as I havent tested them on my nf4 yet.So I'm not sure. That site had so many modded bios for nf2's3's-4s of all makes it really a shame I can't find it now.They had more for abit's than asshhuusshhhhss too. merlin,trats,hellfire,tictac and 4-5 other modders that now escape me . Because now I think that one would do better with a merlin bios. But who knows?? I do kinda' think my fsb wall maybe a bios thing (Psu has sumptin' to do with it too)

But who knows I'm still lookin' for the money for a lp-b or ui. And a nuther "DFI". Right now I'm kinda' tied up with off topic issues. That is not really most important. (Kinda' is though) Nf2 overclocking....))

@"Neo" Is there anyway I could get you to look at a cpu-z dump and maybe tell me which dfi rom-strips trats loaded into this bios?(It's not a "DFI" though.) It seems kinda' slack but maybe not. 3650 everst reads there abouts @240x10. Haven't tested @242.(I wanted 246X10!! LOL)

@Kitfit congrat's on the 270.(Nice bios "Neo")

 

Best to ya's all. I gotta' get back to werk!LOL"Clay"

Edit: I think it may do 250 single stick but I like DC better (LOL)

[NEOAethyr]

Been hammering my cpu alot today.

Made a custom batch using hexus pifast...

 

Anyways I got, just today, gaming stable at 275x9 @ 1.625v.

Well, for qiute a while anyways, after some time it'll exit the game or what not but still plays for a good half hour.

 

Anyways got another instruction on my cpu to appear suddenly .

 

Running 275x8 @ 1.45v for now.

 

SiSoftware Sandra

 

 

Power Rating(s)

CPU Core Power : 75W (estimated)

CPU Cooling System Thermal Resistance : 0.03°C/W (estimated)

 

Sensors

CPU Temperature : 34.0°C / 93.2°F

Auto Fan Speed Control : No

CPU Fan Speed : 5720rpm

CPU Voltage : 1.47V

 

Features

FPU - Co-Processor Built-in : Yes

VME - Virtual Mode Extensions : Yes

DE - Debugging Extension : Yes

PSE - Page Size Extension : Yes

TSC - Time Stamp Counter : Yes

MSR - Model Specific Registers : Yes

PAE - Physical Address Extension : Yes

MCE - Machine Check Exception : Yes

CX8 - Compare & Exchange Instruction : Yes

APIC - Local APIC Built-in : Yes

SEP - Fast System Call : Yes

MTRR - Memory Type Range Registers : Yes

PGE - Page Global Enable : Yes

MCA - Machine Check Architecture : Yes

PAT - Page Attribute Table : Yes

PSE36 - 36-bit Page Size Extension : Yes

PSN - Unique Serial Number : No

CLF - Cache Line Flush Support : No

DS - Debug Trace & EMON Store : No

ACPI - Software Clock Control : No

(W)MMX Technology : Yes

FXSR - Fast Float Save & Restore : Yes

SSE Technology : Yes

SSE2 Technology : No

SS - Self Snoop : No

HTT - Hyper-Threading Technology : No

TM - Thermal Monitor : No

PBE - Pending Break Enable : No

SSE3 Technology : No

MON - Monitor/MWait : No

DSCPL - CPL qualified Debug Store : No

EST - Enhanced SpeedStep Technology : No

TM2 - Thermal Monitor 2 : No

DAZ - Denormals Are Zero : No

 

Extended Features

SYCR - Extended Fast System Call : Yes

EMMX - Extended MMX Technology : Yes

FXSR - Fast Float Save & Restore : Yes

3DNow! Technology : Yes

Extended 3DNow! Technology : Yes

SMP - MP Capability : Yes

 

Power Management Features

VID - Voltage Control : Yes

FID - Frequency Control : Yes

TS - Thermal Sensor Built-in : Yes

 

 

It's the:

SYCR - Extended Fast System Call : Yes

I'm pretty sure I did'nt have that one before.

I know I did'nt have it this morning.

I wish I had my sandra dump from when I got smp enabled...

Then I'd know for sure what I had.