E4300 (B2 & L2) and E4400 Overclocking

jaffa

Distinguished
Sep 25, 2006
64
3
18,635
Long time reader of the forum looking for some advice!

I am aware that the topic of E4300's and E4400's overclocking has been covered time and time again however I am looking for information specific to more recent processor batches. I believe i am right in thinking that Intel has been more selective with the cores that are now put into the E4300/E4400's so i think that this topic is relevant to my objectives.

It is obviously worthwhile to give you all a run through of my other hardware choices as this will reflect on the final overclocking potential of any given CPU. So the specifications are as follows.

Case: Antec Nine Hundred
PSU: Enermax Liberty 620W Modular
Motherboard: abit Fatal1ty FP-IN9
RAM: GeIL 2GB (2x1GB) PC6400C4 800MHz CAS 4-4-4-12
HSF: Zalman CNPS 9700 NT
Sound: Creative X-Fi Xtreme Gamer Fatal1ty

Furthermore, I am in no desperate rush to decide on a graphics card at the moment though my candidates are the 8800GTS or 2900XT. The ongoing 8800GTS vs 2900XT will become a clearer picture and I will make a decision as and when more up to date drivers for each card are released.

What i am looking to achieve is a stable overclock of an E4300/E4400 to approximately the 3Ghz mark or ideally more using air cooling. I will be upgrading the CPU in about a year's time so i am not too concerned with the life of it.

Now my question's are as follows, given the selection of parts I have listed would i enjoy more success with an E4300 or an E4400?

It is my understanding that the B2 stepping over the L2 stepping on the E4300 would give better overclocking results but what is the availability of such cores?

Finally for the people who have been purchasing these CPU's more recently, what kind of overclocking results have you achieved?
 

shargrath

Distinguished
May 13, 2007
237
0
18,680
I have an E4300 at 3.2 (405x8). Vcore is 1.395, temps under 60c. Not sure what my stepping is but I bought it OEM 2 weeks ago.

Why don't you take a screenie of cpuz it should state the stepping and what not.
 

jonny_ftm

Distinguished
May 27, 2007
135
0
18,680
All E4300 are L2, Batch code is more important, even if not a rule of thumb
It is written also on CPU IHS, something like Q6...
 

jaffa

Distinguished
Sep 25, 2006
64
3
18,635
All E4300 are L2

Yeah, i've done a bit more reading and came to realise that now as well. Any suggestions where i could order CPUs from specific batches or will i just have to try my luck?
 

nvalhalla

Distinguished
Mar 14, 2006
1,076
0
19,360
I've never seen a place that guarantees a batch. Someone should make a store than bins that kind of stuff. You could charge a nice premium. If you want my batch on the CPU you're out of luck, I'm not about to mess up my nice OC results by taking my HS off. :)
 

jaffa

Distinguished
Sep 25, 2006
64
3
18,635
I've never seen a place that guarantees a batch. Someone should make a store than bins that kind of stuff. You could charge a nice premium. If you want my batch on the CPU you're out of luck, I'm not about to mess up my nice OC results by taking my HS off. :)

I wouldn't want to do that to anyone! But if you get bored this weekend... :D

I'm going to try Ebuyer/Overclockers/Scan, i have a feeling there may be some older stock lying about!
 

mastabog

Distinguished
Aug 21, 2004
13
0
18,510
I just ordered an E4300, Asus P5N-E SLI and 2GB of DDR2-667 this weekend from eBuyer so it will arrive very soon. I will post back here to let you know what stepping I received from eBuyer.

I intend to o/c the hell out of these components. I have (ordered) the following separate components to be used on the mobo:

- Termalright HR-05 heatsink to replace to uber-crap heatsink ASUS has on the northbridge
- Thermaltake CL-C0025 (vga ram heatsinks) to use them to cool down the voltage regulator mosfets which ASUS chose to leave in open air
- Additional thermal tape for the above (the bastards are heavy and sometimes they fall off)
- Zalman ZM-NB47J heatsink for the southbridge
- Scythe Ninja Plus Rev B for the CPU - this is somehting I already have
- Chrome plated copper heatsinks (from ebay) for the DDR2 sticks (*)
- 3 x 120mm Noiseblocker SX2Pro fans (one intake, one exhaust and one internal blowing in the northbordge, ram and partly the mosfets)

As you can probably tell, I'm pretty serious about pushing the limits with the rig - but keep it silent as well. If you're not in a hurry, you may want to wait for my o/c results which I will post with and without all the above o/c extras mounted shortly after receiving the components this week.

Will keep you informed. Cheers

(*) the sticks are cheap Kingston value ram i got for 37 quid (with google checkout) but I may be lucky like many others to get sticks with Micron D9xxxx chips that were reported to reach 900 MHz!!; I'd rather get 4GB of these than spend a fortune on "quality" ram, especially if they overclock reasonably well.
 

mastabog

Distinguished
Aug 21, 2004
13
0
18,510
Right, mobo and ram arrived today. CPU is on it's way.

Mobo is rev 1.01G and it looks like it has been opened before (lots of sticky areas with fingerprints on the back! -- this is the 2nd time ebuyer sends me something already used, but thi sis my 15th or so item I buy from ebuyer)

Kingston RAM chipsets are not Micron but ELPIDA model E5108AGBG-6E-E. This very stick is proved to overclock extremely well: 900+ MHz at 4-4-4-12!, also go up to 2.2 ... and all this without a heat spreader: http://forums.vr-zone.com/showpost.php?p=3727759&postcount=147

I can't wait to push these chips with a heat spreader mounted. If they oc well then I will instantly buy another pair (38 quid is just too good of a price!)

For the impatient:

I removed the northbridge heatsink and installed the HR-05. A note to anyone removing the stock heatsink: the thermal paste that sits between the heatsink and NB is actually a form of solder. You have to scrap it off with a rigid piece of plastic until you see the nvidia markings clearly and end up with a super shiny surface. At first I was afraid I took off the top nvidia layer when I pulled the stock heatsink coz it was looking like scratched aluminum ... you have to apply some force when scrapping that crap off.

The HR-05 makes very good contact (tested with AS5). Don't forget the cushion pad to avoid crushing the die.

I'll keep you posted.
 

mastabog

Distinguished
Aug 21, 2004
13
0
18,510
I just had time to test it and quickly o/c it .. first results:

cpu went to 3.0 GHz on *stock voltage* with multiplier x8 and fsb 375!
ram flies at 800 mhz at 4-4-4-12 at 1.92v (minimum available manually on p5n-e) and booted into windows at 900 4-4-4-12 at 2.08v (haven't yet run orthos)

remember i have put some heatspreaders on the ram and a hr-05 on the northbridge ... i'm sure I can go higher

amazing results for me, espcially since it was the first try. I'll test more tomorrow and post back.
 

jonny_ftm

Distinguished
May 27, 2007
135
0
18,680
Please post these details:
- stable orthos overclocks
- TAT or CoreTemp 0.95 temperatures
- vcore

Many thanks
Also, you'd better go on 9x multiplyer, easier for your RAM and not worst than high FSB, less heat and less voltage from NB and finally less stress on your CPU FSB limit (yes, many new E4300 are FSB limited)
 

mastabog

Distinguished
Aug 21, 2004
13
0
18,510
I think my E4300 is both fsb and clock limited :(

Stable orthos is 3.0 GHZ, 900 MHZ ram 4-4-4-12 @ fsb 375x8, 1.4 vcore, 1.2 vnb, 1.92 vdimm

Temps while orthosing: 47C CPU (52C core1/2 temp)

Noob question: Does using a higher FSB but the same clock (e.g. 375x8 = 333.3x9) have any benefits performance-wise? Is the memory communication between the cpu and ram or vga and ram any faster? Because I set the memory to run unlinked @ ~800 MHz regardless of the fsb and I know the sata and video card run on the independent pci-e clock, so ...

I remember than everest showed a bit higher memory benchmarks and lower latencies when using the 375 fsb. In the motherboard section it was also listing the FSb as having 12 GB/s bandwidth as opposed to 10.6 GB/s for 333 fsb. I'll check again though, I might be mistaken after all these benchamrks I ran the past 24 hours.

Other notes of interest:

While orthos bumps out with an error, I could run my system just fine at 3.12 GHz (fsb 390x8) @ 1.45 vcore, 1.56 vnb ... i used it for a few hours to play games (e.g. CC3), browse, copy files from/to external USB drives, used a few vmware machines ... no halts or errors. I prefer to keep it at 3.0 GHz with 800 MHz ram though; I never liked to be at the very limit of oc'ing.

It seems the cpu is limited at a bit over 390-395 fsb or 3.16 GHz clock. Regardless the FSB and multiplier I choose, if the end clock is over 3.2 GHz, it won't even post. I have raised voltages, left on auto as some reported it's better, used vcore offset, etc. It won't post if clock is greater than 3.2 GHz.

Similarly for the FSB: regardless of the multiplier used, it won't post if fsb is greater than ~395. Like above, I have tried everything.

Unless I'm missing something major, I believe my CPU is capped but I am happy with a stable 3.0 GHz.

The ram i got, kingston value ram sticks with elpida chips are awsome! orthosed at 910 MHz 4-4-4-12 at 2.08 vdimm ... also worked at 950 MHz but didn't orthos yet. I'm happy with 800 MHz @ 1.92 vdimm though.

I am looking to get the P5K (not deluxe) instead of this P5N-E as I will return this one to ebuyer (they accidentally shipped me a used mobo .. works ok though but I prefer being sure this isn't a refurbished/repaired model). I hear the P5K is an even more stable and o/c friendly mobo which also runs cooler.

Edit: yep, I remembered correctly, there was a memory benchmark difference between 333x9 and 375x8 (both with unlinked ddr @ 800 MHz)



Interesting though how the 375x8 one had slightly slower L1 and L2 results ...
 

jonny_ftm

Distinguished
May 27, 2007
135
0
18,680
Increasing the FSB while lowering the multiplier, will just let you run to the limit of RAM. With C2D, CPU frequency is more important than a small RAM bandwidth increase.

I'll let the multiplier locked (9), lower the FSB, lower your NB voltage (that's what makes your MB hot). Up to near 350-360MHz, you shouldn't increase NB voltage. In any case, after your overclock is stable (Orthos, min 5h, ideal >12h), lower your NB/FSB voltages to the minimal, and check stability.

Stable orthos is 3.0 GHZ, 900 MHZ ram 4-4-4-12 @ fsb 375x8, 1.4 vcore, 1.2 vnb, 1.92 vdimm
Your DDR is limiting you, release your timings first before overclocking:5-5-5-15, and set vDIMM to 2.0v. This really important to check your CPU.

Set your vcore at 1.45/1.47 (some think it's too high, see my signature), and now get the most of your CPU without beeing limited by memory. When done, tighten your memory timings or try a higher multiplier (unsynchronous).

Remember, a stable high overclock needs some days to be set, and you can't use your PC when Orthos is unstable, you'll expierience bugs/data loss
 

mastabog

Distinguished
Aug 21, 2004
13
0
18,510
Thanks a lot for the reply.

The nb voltage is already at the lowest setting, 1.2v :)

I already did what you suggested and it won't post when surpasses 3.16 GHz, regardless of the fsb/multiplier/voltages combination ... believe me, I tried all voltages possible, even the highest with 100mv voffset (ram was locked at stock settings, 667mhz 5-5-5-15 @ 2.1v to make sure i'm not limited by it because i know some mobos drop vdimm a bit when overclocking the cpu/fsb heavily).

I see yours is set to 3.115 GHz @ 1.58v .. that is a very high voltage for this cpu in my opinion. i can probably reach that too since I was almost stable at 3.15-3.16 (only orthos bumped out after some time), but I'm ok with 3.0GHz at 1.4v with max cpu temp at 47C (52C cores) ... will try 1.375v. Idle cpu temp is 34C (23C core). The room is pretty cool though (about 22C ambient).

I didn't know a stable overclock needs a few days to set in. Thanks for the tip! (are you sure about that though? I haven't heard about this and i've been reading quite a lot).
 

jonny_ftm

Distinguished
May 27, 2007
135
0
18,680
To get your max overclock I mean. That is: increase the FSB some MHz, test Orthos, if no bug at some mn, than run Orthos long hours. If fail: increase vcore, test again...
I personally use 16h orthos tests, so, yes, it takes long time, even at recommended 8h

I didn't see much temp difference from 3GHz and 3.15 on my chip, that's why I let it at that vcore

I'll see when it fails :)
 

mastabog

Distinguished
Aug 21, 2004
13
0
18,510
Yep, i know that. I thought you were saying that an o/c becomes stable in time, even if at first orthos fails, which sounded peculiar.

I'm not that desperate to get an extra 115 Mhz over 3000 Mhz. I don't like that super high voltage you got (1.58v) :); I intend to keep mine working for a long(er) while :)

I orthosed it at 1.375v vcore at 3.0 GHz for 10 hours so i'm quite happy.

Thanks for your replies. I hope this helped jaffa as well.

Cheers