What not to do + Mobile 2400+ IQYHA 0402 SPMW 35 W (bought this week @newegg) results

[c627627]

What not to do + Mobile 2400+ IQYHA 0402 SPMW 35 W (bought this week @newegg) results

Mobile Barton AMD Athlon XP 2400+ AXMD 2400 FJQ4C 35 watts

IQYHA 0402 SPMW Purchase date: Monday June 14, 2004 from newegg.com

I'm using Thermalright SLK-947U heatsink with Thermaltake Smart Fan 2 and three $14.99 K-Byte (SpecTek) PC2700s at 6 3 3 2

Epox 8RDA+ nForce2 with a wire trick to enable high range 13 and up multipliers.


Booted at 2 GHz just to see what the default voltage is, 1.5 volts, could that be right for 35 watters?? 35 watters are 1.35 volt chips at default, what's the 1.5 about?


13.5 x 185 = 2500 MHz failed Prime95 Torture test at 1.9 volts after 1 hour. Peak temps at 1.9 volts 53 C. (RAM was previously tested stable at low 190s FSB.)


What now, reverse course and head back to find a stable 1.8 or 1.85 V point or go above 1.9 volts?


I guess I should have gone for the $10 cheaper 45 watter, it would probably perform similar. Or maybe $1 more for 2500+ 45 watt mobile?

If 1.35 is indeed the default voltage, going over 1.9 is what troubles me.

But.. you know people do 2 volts with their mobiles and if I failed 2.5 GHz after 1 hour at 1.9, I guess 2 Volts may get me over 11 hours of Prime95?


By the way, learn from my mistakes:

I'm using a fixed mobo now, but when it broke down, I bought a used one to use while I fixed the old one:

...................................HUH!?

========================



.................Hm, ah, what a heck.

========================






.................................BOING!


========================




....Do not install heatsinks on mobos with broken parts.

 

[dead_man311]

Ouch that chips lil black. Thats got to suck. 2 Volts wow i can only give mine 1.81 thats only just the bios

What heatsink where you using on the cracked socket ? That SLK-947U Bolts on

 

[c627627]

Actually, that was done with a clip on Thermalright SLK-900A.
That chip, believe it or not, is running the system from my mrs. It's a 2100+ B overclocked to 2.2, still running without any apparent problems (!)

On another note, I just stopped Prime95 Torture Test after 14 hours, successfully testing the
35 watt mobile 2400+ IQYHA 0402 SPMW at [185] x 13 = 2405 MHz at 1.8 Volts.

Hm... since it failed Priem95 at 2500 MHz and 1.9 volts after just 52 minutes earlier,


What would you all do, leave it at 2405 and 1.8 volts?

Considering I bought that chip earlier this week, 35 watt 2400+s are not doing as well as they did, (I would have hoped for 2.5 at 1.9V,) what would have a 45 watt 2400+ mobile done or 2500+ 45 watt, do you think, if you had a chance to read about people recently buying them at newegg?

 

[The Coolest]

As you can see from my sig, my mobile can do 2.5GHz @ 1.93v (its a 45w chip), anything lower and it failed prime. At this high voltage I wouldn't let the chip go above 50C full load.
Right now I'm running at 2.4GHz @ 1.8v.
I recommend you run Prime95 for 24hours on the smallest FFTs, if it can handle that, it'll handle anything.
Also you can get ZIF socket for a couple of bucks, so you can repair that mobo.
About why you're getting 1.5v instead of 1.35v. That happens because the mobile and desktop CPUs have different meanings of the L11 bridges. a certain setting on a desktop will be interpreted as a different one on a mobile, and vice versa.
Basically 1.35v on a mobile will be interpreted as 1.525v on a desktop, if that doesn't exist the BIOS will probably just round it to the closest value.
the 1.45v chips will have 1.575v on desktop mobo's

 

[c627627]

Thanks for that voltage explanation, my default V was 1.525 V indeed.

I believe I'll leave it at 2.4 GHz and 1.8 Volts.
2.5 GHz may require close to 2 volts + I noticed an uncomfortable temperature jump between 1.8 and 1.9 volts.

Hey I thought Prime95 'Blend' was recommended, not the 'smallest FFTs' for the Torture Test?


Given my results, the question remains what the current recommended mobile would be, the 2400+ 45 watter or the 2500+ that costs the same as 2400+ 35 watter?

Or maybe it's all just luck no matter which of the three you go for?

 

[CandymanCan]

Umm that cpu in the pics looks fried, its all black and stuff.

Btw all this what u said with the pics speically the last cracked me up lol

 

[The Coolest]

Well Blend is recommended for overall system stability test I believe. Small FFTs stresses your CPU the most, and heats it up pretty well too. I personally use Prime95 for only CPU stability testing. for memory I use Memtest and for overall system stability, UT2004 for a couple of hours seems sufficient

 

[c627627]

Why not 3D Mark and what happens when these video tests fail, (like when Prime95 displays errors...)

 

[The Coolest]

I dunno, I usually don't use 3dmark as it makes my pc unusable. while you can see use the pc while running prime95, and 12 hours of memtest is usually enough, so you can just run it overnight

 

[RedDawg41]

just run the stress test on the cpu... that would be the recommended thing to do... your ram isn't up to par (never has) and if you want over all system stablility then fall back to known video drivers (certs), recommended timings for your ddr ram(spd) then work your proc up onthe multi/vcore with 3dmark2001se. find a range were it doesn't lockup. that might give you an idea or so. GL.

 

[lucas81]

We can post in the thread about that 2400+:
http://www.ocforums.com/showthread.php?t=305940
so as not to hijack this guy's thread but my conclusion since making that purchase is that most newly purchased 2400+ mobiles do less than 2500+ and 2600+ mobniles now.

They use to do better but there's a growing trend indicating most of them were now labeled as 2400+ because of their performance and not just marketing reasons like when they first came out.


The asnwer to your FSB question is:

Generic average overclocks:

PC2100 form 133 FSB to 140 to low 150s FSB, 100% insync.
PC2700 from 166 to 180 to low 190s FSB, 100% insync.
PC3200 way into 200s, depending on brand.


I have PC2700 RAM capable of low 190s at the most. I dropped the FSB down a little because it's summer and the temperatures in Kansas are very hot, I'll probably go back up when it gets cold here.

Also I need the wire trick to go above 12.5 multiplier so I would have to go through the hassle of getting the wire out of the socket to do the same MHz I have now with
185 x 13 = 2405
vs.
192 x 12.5 = 2400

and 192 is on the edge of what most generic PC2700 RAM sticks are capable of.

That would be why.

I quoted you from the other thread, so we can discuss here.
So you are at 185 fsb now, but that's because your memory can't handle higer speed. Have you tried to run the mem a-sync? Just to see what the limits are for your cpu?
The thing is, i ordered the exact same cpu from excaliberpc.com, and some OCZ 3500EB two weeks ago and still haven't received it yet. I was hoping for a something in the 2.5 - 2.6 range with my wc-setup. I also ordered a Sparkle 550 psu to make sure it's not the psu holding my oc back, like now.
But as I read more about the "new" 2400+, it seems like I picked the wrong cpu....
And could you tell me what your temps are?

 

[Docta_Z]

Stability is a minimum of 12 hours on memtest86 to test ram at higher fsb and 12 hours of prime95 on small fft's as this is the hardest test (blend tests ram as well - which you've already tested with memtest86)

Testing with 3dmark can be helpful - but also stresses your video card, so improper video card settings could lead to restarts , lockups or errors and fault you into thinking you have a cpu problem rather then a video one.

Don't bother running your memory async - it's a useless bottleneck that will create stability problems.

 

[lucas81]

Yeah I know running the memory async, but I mean just to see what the cpu is capable of and to see if the cpu is really performing that bad (compared to the better oc-ing ones.)

 

[c627627]

So you run 3DMark just to see if it would lock up? How many times should you loop it?

Have you tried to run the mem a-sync? Just to see what the limits are for your cpu?

And could you tell me what your temps are?

Like Docta_Z said, running Socket A CPUs async is not a good idea no matter what benchmarks say, and I did some testing back in the day:
http://www.ocforums.com/showthread.php?s=&threadid=215500&perpage=50&pagenumber=2


My temperatures now are at 50 C because my room is very summery hot and the case temperature is 41 C despite a decent large Antec case with 3 fans in, 2 fans out. (I do like to keep it closed though.)

This CPU cannot do 2.6.

2.5 GHz at high voltages and extreme cooling, and in winter months: yes it could do 2.5, but not 2.6.


A good overclock for me is when you can step back and ease up on the noise and have the case closed.... I loved those 2100+ B's that can do 2.3 GHz relaxed...



If you're shooting for 2.6 GHz, I would recommend looking into a return and exchange for a 2500+ or a 2600+ mobile since you haven't even received your 2400+.


Note that you may get lucky but odds are not in your favor if your goal is 2.6 and above... good luck man.

 

[c627627]

By the way lucas81, I can lower the FSB and increase the multiplier to see what CPU is capable of easily (because it is multiplier unlocked.)

In other words, you don't need to increase the FSB to test the CPU capabilities if you have old unlocked CPUs or unlocked mobiles.

 

[lucas81]

Euhm...I think you misunderstood me about the async part. What I mean was, if you can push your fsb to say 200+ and it's prime stable, then if you upgrade your memory then you can run it at the 200+ fsb insync right? that's what i meant...
sorry for being unclear, but I'm not native english.
Yeah after reading all this I really would like to return the cpu, but they already shipped it. I have a tracking number but it hasn't been updated for a week now, so I don't even know where my goodies are. The only thing I know is it already has left the borders of the US. And since I live in Europe, I think just order a second mobile cpu... I hope not ofcourse...

 

[lucas81]

By the way lucas81, I can lower the FSB and increase the multiplier to see what CPU is capable of easily (because it is multiplier unlocked.)

In other words, you don't need to increase the FSB to test the CPU capabilities if you have old unlocked CPUs or unlocked mobiles.

So are you saying that 200*12 = 240*10 ? (yeah ofcourse that is in mathematics ) I mean, but if say 2400 mhz is the max of the cpu, the combination of fsb and multi don't have any influence on the max?
I mean, isn't it that mostly the fsb determines the limit of a cpu? and that after determining the highest fsb, you tweak the multi's?

I hope you understand what I mean, I can hardly express my thoughts in my own language...

 

[c627627]

MHz speed = [FSB] x multiplier

New recently purchased 2400+ mobiles average 2.4 GHz according to many reports here.

That means 2.4 GHz with either
low [FSB] x high multiplier
or
high [FSB] x low multiplier

They average 2.4 GHz.


Of course it's better to have high [FSB] but what kind of an FSB and multiplier combination depends on whether you have a motherboard capable of going way into 200s FSB and more importantly if you have a quality RAM stick(s) capable of going way into 200s FSB.


It used to be that some old CPUs were not capable of doing high FSBs but they are now rare. To answer your question:

all new mobile Bartons will hit high FSBs provided you have motherboard and RAM that is capable of running at high FSB speeds.

 

[c627627]

MHz speed = [FSB] x multiplier

MHz depends on CPU
FSB depends on RAM, motherboard (mostly)
Multiplier depends if you have a multiplier unlocked CPU or not as well as if you have a motherboard capable of booting with all multipliers

 

[lucas81]

Thank you very much for explaining to me You have learned me something new today
Ok so I just cross my fingers that I get a good one...(IF i get it. )
Don't know what's taking so long...