Help w/ 3.2B Multi on EP43-DS3LR

[RoscoP]

Good evening fellow oc'ers. I'm trying to run 333 x 10 on the 333 Mhz strap with 3.2B Multiplier. My GA-EP43-DS3LR doesn't want to boot with 3.2B, but it will post with 2.4B at this speed. My RAM I'm using is OCZ Fatal1ty PC2-8500 1066 Mhz @ 6-7-7-20 2.0V. I have manually set my timings and voltage for my RAM and when I tried 333 x 10 with the 333 Mhz strap and 3.2B multi, it won't boot. What could be causing my board from posting when using the 333 Mhz strap and 3.2B multi?

I was previously doing 266 x 12.5 on the 266 Mhz Strap with a multiplier of 4.00A. I wanted to see if trying out 333 x 10 on the 333 Mhz strap would yield more performance. I'm currently doing 333 x 10 on the 333 Mhz strap with a multi of 2.4B. My RAM is running stable at PC2-6400 800 Mhz @ 5-5-5-15 1.8V The only benchmark I have ran so far is Super Pi (probably isn't good to measure memory bandwidth), but it seems stable. I know every board varies as to what strap should be used, but does anyone have any suggestions?

[Chike]

Re: Help w/ 3.2B Multi on EP43-DS3LR

I had lots of issues overclocking at first, this post helped.

[RoscoP]

Re: Help w/ 3.2B Multi on EP43-DS3LR

Chike, thanks for the reply. I checked the post you gave me a link to. I changed my BIOS to have those recommended settings. I still can't boot with 3.2B multiplier. Also, I noticed that -

CPU Termination ---------- 1.200V* ----- : [Auto] <--- set to the value to the left (Auto over-volts this too much when overclocking) Auto is OK if not overclocking
CPU PLL ------------------ 1.500V* ----- : [Auto] <--- set to the value to the left (Auto over-volts this too much when overclocking) Auto is OK if not overclocking

What is a good CPU Termination and CPU PLL for the overclock I am doing ? I notice the asterisk next to the 1.200V and 1.500V... are those recommended or the Auto settings?

I was able to confirm that using the 333 Mhz strap does give real world performance gains over the 266 Mhz strap I was using. Crysis gives me a few more FPS in the Crysis Benchmark Tool 1.5 (cpu_benchmark & gpu_benchmark). I'm really wanting to run my RAM at 1066 Mhz via the 3.2B multiplier to see if I get any more gains.

Any suggestions or input will be greatly appreciated. Thanks.

/edit
I realize that I have not changed my MCH Vcore from the default yet. I am running a 2x2GB OCZ PC2-8500 kit. Maybe I need to up the MCH Core to 1.200V? I don't see why I would be stable with 333x10 on the 333 Mhz strap with 2.4B, and not be able to post with 333x10 on the 333 Mhz strap with 3.2B, regardless of MCH core....

[Chike]

Re: Help w/ 3.2B Multi on EP43-DS3LR

All I know is that untill I have set the voltages as recomended in that post to the values on the left, and the PCI frequency to 100, I wasn't able to boot at 333Mhz and the 3.2 multiplier as well (I have E7200).
I haven't tried to pin point wich of them did the trick, I changed them all as recomended.

I'd suggest you set everything like in that post suggest, and then if you manage to boot and have stability, you can start eperimenting which is needed or not. You may need to raise vcore for 333Mhz, or maybe leave this value on auto (works for me with CPU EIST Function enabled.)

From my personal eperience don't expect much performence gain (if at all) by raising memory speed. If memory bendwidth is your concern you will get it, but for normal use you won't notice a diference.

DDR 1066 does not add much compared to DDR 800 at 266Mhz bus frequecy, but with 333Mhz you may have an extra 1,5GB/s or even more, and a little lower latnecy.

[Psycho101]

Re: Help w/ 3.2B Multi on EP43-DS3LR

Different straps (MCH timings) do require different MCH voltages. It's just the same as RAM requiring 1.8V to do 5-5-5-18 but needing 2.1V to go 4-4-4-12. Increase MCH core to at least 1.2V with 4Gb of RAM.

To be honest, I'm not liking your RAM much. 6-7-7-20 for 1066.... makes me wonder if OCZ got the chips out of the reject bin. Sounds like it's very old RAM released when 1066 was new, and is probably average quality DDR800 with a timing and voltage increase. I'm not being disrespectful about your components, just warning you to bear in mind, thet it may be the cause of your issues. The primary reason I avoid OCZ RAM is because of the half-way ranges they supply with unconventional timings like yours and 5-6-6-18 etc. Just seems that they're using anything they can find and adjusting timings to suit the quality of the the RAM rather than specifying RAM timings they want and finding IC's that deliver that speed.

The only answer as to what CPU termination and PLL voltages you should use is "the ones that are stable". There is no formula, no way to guess or calculate what voltages you need unfortunately.

Don't feel that you need to stick to the "right strap for the FSB" either. I'm perfectly stable with a 411 FSB and 2.66C RAM multi.

[RoscoP]

Re: Help w/ 3.2B Multi on EP43-DS3LR

Different straps (MCH timings) do require different MCH voltages. It's just the same as RAM requiring 1.8V to do 5-5-5-18 but needing 2.1V to go 4-4-4-12. Increase MCH core to at least 1.2V with 4Gb of RAM.

To be honest, I'm not liking your RAM much. 6-7-7-20 for 1066.... makes me wonder if OCZ got the chips out of the reject bin. Sounds like it's very old RAM released when 1066 was new, and is probably average quality DDR800 with a timing and voltage increase. I'm not being disrespectful about your components, just warning you to bear in mind, thet it may be the cause of your issues. The primary reason I avoid OCZ RAM is because of the half-way ranges they supply with unconventional timings like yours and 5-6-6-18 etc. Just seems that they're using anything they can find and adjusting timings to suit the quality of the the RAM rather than specifying RAM timings they want and finding IC's that deliver that speed.

The only answer as to what CPU termination and PLL voltages you should use is "the ones that are stable". There is no formula, no way to guess or calculate what voltages you need unfortunately.

Don't feel that you need to stick to the "right strap for the FSB" either. I'm perfectly stable with a 411 FSB and 2.66C RAM multi.

Hey, don't feel bad about letting me know the truth on my RAM. I made the mistake when I ordered it :-( If I could order again I would order DDR2-800 that does 4-4-4-12. I got suckered into ordering this "Fatal1ty" OCZ RAM because it had a nice combo deal from newegg. Prior to my current setup I was stuck on a Socket A rig that I had since 2003. I used to have a mobile Barton 2500+ in an Abit NF7. I am glad to be learning how to OC newer hardware. Anyways, I'd like to not learn from mistakes, but I do learn "better" from mistakes.

I was able to get 3.2B multi whenever I upped the MCH Vcore to 1.2V. I have not seen any performance increases in games though. If I do not see any real world difference between my multi @ 3.2B (RAM @ 1066 6-7-7-20 2.0V) and 2.4B (RAM @ 800 5-5-515 1.8V) should I run the 2.4B multi and keep my RAM at 1.8V to give it a longer life?

Does anyone think my GA-EP43-DSL3LR could do the 400 Mhz strap with stock cooling on the Northbridge? I havn't applied AS5 to the Northbridge or anything. The board's specs says it can do 1600 OC... maybe its worth a shot? Can the E5200 even handle a 400 Mhz strap?

Thanks for the info Chike & Psycho :-)

[Psycho101]

Re: Help w/ 3.2B Multi on EP43-DS3LR

If you'd like to continue running at DDR-800, I'm sure the performance would be almost equal to 1066 due to the not so good timings you'd have to use to get 1066.

FSB strap and FSB speed are different. A FSB strap is simply a set of optimised timings, similar to RAM timings (wait states) applied to the northbridge to optimize performance for the FSB speed indicated in the straps name. An actual FSB increase is the main stay of overclocking, as you know, and it's this rather than strap settings that will overclock the system.

I'm sure your board would be more than happy at a 400MHz bus speed, but I'm not 100% sure your CPU would be OK there. You may be able to get it stable, but it could be difficult. I'd recommend using a 300MHz FSB and a multiplier of 12 to give 3.6GHz. Setting a 2.66 RAM multiplier will bring you up to DDR-800. I'd estimate fromwhat I've seen that you'd need to give the CPU about it's max rated voltage of 1.3625V to be stable.

My favourite e-tailer (UK based) sells E5200 systems pre overclocked to 3.75GHz at 1.3625V on a Gigabyte GA-EP43-S3L. It does have a good CPU cooler fitted, the Scythe Kama Angle, and according to the review in Custom PC magazine, runs at a near perfect 67c at load. Aslong as your heatsink is up to it, you should get similar results. Notrthbridge heat shouldn't be an issue. Use the least amount of northbridge voltage as possible to be stable, and as long as you have good case airflow, you'll be fine.

[RoscoP]

Re: Help w/ 3.2B Multi on EP43-DS3LR

Do you typically expect to see a gain in overall performance when using a higher strap? Sorry to ask such a simple question...

[profJim]

Re: Help w/ 3.2B Multi on EP43-DS3LR

The EP43 mobo will probably top out at about 420Mhz cpu host frequency.
At this speed you'll want to use 2.00D or 2.00B System memory multipliers to run your memory at slightly above 800MHz with tighter memory timings.
The 2.00B MCH strap will give you a bit more memory performance than the 2.00D strap.
You might find it easier to get your system stable if you first try the 2.00D strap.

We can help you maximize your memory's performance with some additional bios settings once your system is stable.

[RoscoP]

Re: Help w/ 3.2B Multi on EP43-DS3LR

The EP43 mobo will probably top out at about 420Mhz cpu host frequency.
At this speed you'll want to use 2.00D or 2.00B System memory multipliers to run your memory at slightly above 800MHz with tighter memory timings.
The 2.00B MCH strap will give you a bit more memory performance than the 2.00D strap.
You might find it easier to get your system stable if you first try the 2.00D strap.

We can help you maximize your memory's performance with some additional bios settings once your system is stable.

I tried to do 400 Mhz strap w/ 2.00D multi (400x8 = 3.2Ghz CPU, 400x2.00 = 800 Mhz RAM) to achieve the 1:1 FSB:DRAM ratio, but my board won't post. I also tried the 2.66D multi on the 400 Mhz strap (300 x 11 = 3333Mhz CPU, 300 x 2.66 = 800 Mhz RAM) and I was unable to get it to post.

I am afraid my E5200 just won't work on the 400 Mhz strap. I've heard about the E5XXX chips only being able to max out at about 330-340 FSB. Is this true? I'm sure my EP43 can handle the 400 Mhz strap, I think it's the E5200 stopping it from posting. Can anyone else confirm this? Is it possible that I need to up the MCH Vcore to 1.300V to achieve the 400 Mhz strap? I really don't want to put anymore vcore to my northbridge...

[Psycho101]

Re: Help w/ 3.2B Multi on EP43-DS3LR

TO be honest it's really not worth going through too much pain to get the 400 strrap going. put the strap setting on auto and try all the settings for system memory multipliar that get your RAM at or near the speed you want to run it at.

Start with a FSB of 333 and a multiplier of 10.5 to get 3.5GHz. Use the 2.4B multiplier to get DDR-800. If it's stable, up the FSB a little to 345. A 1:1 ratio is no different from using a higher multiplier. You won't get better performance from a 1:1 ratio, not on an Intel Core 2 platform.

You seem to be using "Strap" and "FSB" interchangably. They're totally different. If your motherboard has a 400 strap, then of course it's able to use it. It's the RAM and CPU that dictate it's stability, along with timings and voltages set.

E5200's that can run a 400 FSB are in the minority. Your CPU has a 12.5 multiplier.... use it, or at least use *10 or more. There's no need to try and go for a 400FSB. The only difference you'll likely see is in benchmarks. CPU speed is king.

[RoscoP]

Re: Help w/ 3.2B Multi on EP43-DS3LR

TO be honest it's really not worth going through too much pain to get the 400 strrap going. put the strap setting on auto and try all the settings for system memory multipliar that get your RAM at or near the speed you want to run it at.

Start with a FSB of 333 and a multiplier of 10.5 to get 3.5GHz. Use the 2.4B multiplier to get DDR-800. If it's stable, up the FSB a little to 345. A 1:1 ratio is no different from using a higher multiplier. You won't get better performance from a 1:1 ratio, not on an Intel Core 2 platform.

You seem to be using "Strap" and "FSB" interchangably. They're totally different. If your motherboard has a 400 strap, then of course it's able to use it. It's the RAM and CPU that dictate it's stability, along with timings and voltages set.

E5200's that can run a 400 FSB are in the minority. Your CPU has a 12.5 multiplier.... use it, or at least use *10 or more. There's no need to try and go for a 400FSB. The only difference you'll likely see is in benchmarks. CPU speed is king.

Psycho, I have good news. I traded my friend my OCZ Fatal1ty RAM for his OCZ Gold that does 1066 Mhz @ 5-6-6-18 2.1V (I know, not the best timings, but better...) My big question now is - Do certain straps only allow your FSB multi to get so high?

ex. Can a 200Mhz strap allow you to do 333 x 10 ?

I understand that certain straps will allow different latencies for your RAM...

I also heard that if you are wanting to do the 400 Mhz Strap you should set your FSB multi below or above 400 by a tad bit. Maybe I should give that a shot?

[Psycho101]

Re: Help w/ 3.2B Multi on EP43-DS3LR

Straps seem for me to either work or not. they don't do anything more, or at least I haven't gotten them to.

At the moment with the over clock in my sig, I can run with a multiplier of either 2.5A or 2.66C. most others that keep my RAM in a sensible range (leaving only both 1:1 multi's and 2.66D) fail to boot completely, no matter the settings.

From my testing I've concluded that on my combination of components (same was true on my Q8200, just different multi's were awkward) that strap selection is often hit and miss, and it's better to try all within your RAMs operating range for the best stability. This doesn't over ride my fundamental principle of setting an over clock I want then working on getting stable. It just means I'll try straps and multipliers first on something I believe should otherwise work.

[RoscoP]

Re: Help w/ 3.2B Multi on EP43-DS3LR

I was able to do 333 x 10.5 = 3500 Mhz, but it required much more vcore than my 1.24375 at 3333 Mhz. I would really like to do 3500+ Mhz, but my main problem is my load temperatures.

I am using an Arctic Cooling Freezer 7 Pro with Arctic Silver 5. Real Temp reports 65-68 C during Intel Burn Test [max 10 loops] @ 3500 Mhz. When I have it at 333 x 10 @ 1.24375v, I get around 60-62 C Max load temp via Real Temp.

Are these temperatures bad enough to lower my OC & vcore? For some strange reason my BIOS reports a lower CPU temperature than does Real Temp. I am more inclined to trust Real Temp, but I'm not for sure. I think I need to re apply AS5 on my heatsink in vertical lines under the heatpipes. I used the spreading technique and I've been told this is not as efficient as spreading thin lines. Is this true?

Also, why does Windows recognize the speed of my CPU wrong? CPU-Z recognizes the speed right, but most games / other apps recognize my CPU speed incorrect.

If anyone can answer any of my questions above I would greatly appreciate it.

[Psycho101]

Re: Help w/ 3.2B Multi on EP43-DS3LR

Re-applying your thermal paste will likely not reduce temps by much. If you used the right amount first time that is.

65-68c is acceptable at least to me, when running IBT. My overclock reaches ~71c using IBT, but only gets as high as 60c under prime load. I've run Realtemp on several occasions all day, while doing various stuff, including gaming and video transcoding. Temps never got over 63c, and averaged I's guess about 60c.

Your 3500Mhz overclock is more than viable IMO.

As for Temp readings, the BIOS uses a completely different temp sensor than Tealtemp. Realtemp accesses the CPU's built in temp diodes, imbedded into the cores themselves. The BIOS is reading a temp sensor located on the CPU socket. BIOS readings are often cooler than core readings. It's the core readings that are important, and it's these that Intel refers to when stating the max temperature range. They're not always accurate, as they're designed purely to initiate thermal throttling at the stated TJMax, and are more accurate at high temperatures than low ones. Your sensors sound as if they're in good working order though. Both of mine are stuck at low temps, but start working and matching each other at 40c and above.