Jump to content
Corsair Community

P5W DH Deluxe/E6700 & PC6400Pro


Recommended Posts

I have this working like a charm but it took some tweaking. Please read

 

CPU Freq 300 <--this is a mild OC

Dram Freq 900 <--this is not 1:1 timiing

PCI Express Freq 100

PCI Clock Sync 33.33

Mem Voltage 2.20V

CPU Voltage 1.3875V

ICH Voltage 1.20V

MCH Voltage 1.65V

Performance Mode Standard

FSB Termination Voltage 1.40V

Timing 4-4-4-12

Dram ECC Mode 2 Clocks

Hyper Path Enabled

Dram Throttle Disabled

 

After 5 months this works best with the PC6400Pro using a P5W DH Deluxe with a E6700. You can tweak if you want but I wouldn't change FSB term. Voltage, ICH Voltage nor MCH Voltage, leave those the same. You can use Memtest to tweak your mem but you have to know what you are doing. I am surprise these options were only available in 486 MB bios and not the bios we use today. When using Memtest you have to make changes one at a time. That means once you make changes to any 1 value hit the "apply" button and move on to the next value. But if you are using dual channel (twin sticks) make sure you do them in pairs first if you have 2 sticks. In other words when you change values on channel a and b hit the apply button and move on.

 

PC8500C5D on page 2

Link to comment
Share on other sites

Thank you for your insight. Very informative. :D:

Opps I almost forgot: Below is a pic of PC6400Pro at 3-3-3-10 @ 300 FSB

http://i11.photobucket.com/albums/a154/eastcoasthandle/E6700/3_3_3_10_ram-1.jpg

 

I have to admit in order to keep it at 3-3-3-10 I have to put a 80mm fan on it. Once they start to heat up I get errors. Voltage at 2.20

 

But thats with v1.3

I hear that the newer versions use Promos

Link to comment
Share on other sites

I have found that the move from 1:1 to any greater ratio whether it be a raise of FSB or DRAM (that is non-concurrent) to be far less beneficial in testing. For example:

 

1:1

 

http://i41.photobucket.com/albums/e269/Ropey123/1to1.jpg

 

 

5:4

 

http://i41.photobucket.com/albums/e269/Ropey123/5to4.jpg

 

Now with benchmark testing, there is a minimal difference and some show 1:1 to be more beneficial in cache and ECC action where some show benefit in 5:4where write, etc. show some benefit.

 

When I stopwatch BF2 and Oblivion on load, there is no difference. When I test in FRAPS there is no difference. For some reason, Quake4 shows no difference in load times but FRAPS shows an ~6FPS increase. That makes no difference however when the FPS are so high on a decent video card anyway though.

 

This leads me to believe that a greater DRAM to FSB ratio than 1:1 is not nearly as beneficial to CORE2 vs AMD where the benefit is much easier to find with the digital stopwatch. The thermal heat output of greater than 1:1 ratios does not seem to be a real benefit. The reasoning for such higher bandwidth ability is for either tightening up the timings or overclocking a high FSB and keeping a 1:1 ratio with the CPU:DRAM.

Link to comment
Share on other sites

I have found that the move from 1:1 to any greater ratio whether it be a raise of FSB or DRAM (that is non-concurrent) to be far less beneficial in testing. For example:

 

1:1

 

http://i41.photobucket.com/albums/e269/Ropey123/1to1.jpg

 

 

5:4

 

http://i41.photobucket.com/albums/e269/Ropey123/5to4.jpg

 

Now with benchmark testing, there is a minimal difference and some show 1:1 to be more beneficial in cache and ECC action where some show benefit in 5:4where write, etc. show some benefit.

 

When I stopwatch BF2 and Oblivion on load, there is no difference. When I test in FRAPS there is no difference. For some reason, Quake4 shows no difference in load times but FRAPS shows an ~6FPS increase. That makes no difference however when the FPS are so high on a decent video card anyway though.

 

This leads me to believe that a greater DRAM to FSB ratio than 1:1 is not nearly as beneficial to CORE2 vs AMD where the benefit is much easier to find with the digital stopwatch. The thermal heat output of greater than 1:1 ratios does not seem to be a real benefit. The reasoning for such higher bandwidth ability is for either tightening up the timings or overclocking a high FSB and keeping a 1:1 ratio with the CPU:DRAM.

 

Very interesting observation, indeed. I wonder if getting the most out of ram on the 800 vs 1066 strap will show the best benefit then? For example E6700 at default settings on the 1066 strap vs. E6700 OC on the 800 strap, both at 1:1 when possible?

I do believe having the ram on the highest strap should yield the best performance. At least in theory on the Intel platform.

 

Oh I did notice that your 1:1 was clocked a bit higher then your other.

Link to comment
Share on other sites

For example E6700 at default settings on the 1066 strap vs. E6700 OC on the 800 strap, both at 1:1 when possible? [/Quote]

 

You can't make the 1066 strap when at default settings with a 1:1 ratio. You would have to drop the multiplier and raise your FSB to 533 to make the 1066 (or other such divider/multiplier manoevers). Keep in mind that the CPU bus is 266 and quad pumped which makes 533. Thus when on default you have the 533:533 (1:1) ratio. You need to overclock the CPU to gain a higher DRAM speed at 1:1.

 

Oh I did notice that your 1:1 was clocked a bit higher then your other.

 

My 800Mhz (1:1) is clocked at 3.2Ghz. My 1000MHz (5:4) is clocked at 3.4Ghz. Thus the 5:4 is clocked a bit higher than the other. If you notice, the 1000Mhz shows some higher cache speeds and can pretty much keep up with the 800. If there was a far greater difference on all benchmarks then yes, the move would make more sense. As it is the difference is minimal. To get 1:1 at 800 I have to keep the FSB to 400 X 2 = 800. I can not make a 500Mhz FSB to get 1:1 at the 1000 level so I have to divide the FSB/DRAM to achieve the 1000MHz Dram speed.

 

In the 800 strap we have the following ram ratios (FSB:DRAM):

 

400,533,667,800 or as some quote them 1:1, 3:4, 3:5, 1:2

 

In the 1066 strap we see the following.

 

400, 533, 667, 800 but the actual ratios have altered, 400 mode is now a down clock so the following applies.

 

400=4:3, 533=1:1, 667=4:5, 800=4:6

 

Notwithstanding. I personally find that the difference is minimal.

 

That's all I was responding to and that is: "The manipulated divider does not show any "great or marked" difference with the non manipulated divider. In my opinion, the "Law of Diminishing Returns" becomes more marked after 1:1.

 

Yes, some benchmarks will show some minimal differentiation. However, on the real life platform with a load of programs/games and a digital stopwatch it seems not to translate. If I hold a 1lb grapefruit in one hand and a 1.05lb grapefruit in the other hand, I will not feel much of a difference. Yes, the scale will show it but the difference in real life (eating of the grapefruit, cutting into pieces, etc) the difference is quite minimal. I find this the case with Core2 and this difference is far more marked with A64.

 

Edit: Very nice hardware by the way :)

Link to comment
Share on other sites

If you keep a 1:1 relationship with your CPU:DRAM and clock your processor and ram up high with your current Micron D9's then I would stick with them. Keep in mind that the Micron D9's that were used in those products were the ones that Corsair tested and found NOT to make their stringent requirements for the 8500's (and above) and were down binned, so if you find that your ram is limiting you then a move to the 8500's may well be beneficial. My personal opinion is that you very likely can make some great 1:1 with that dram.

 

For example, if you were able to run your FSB at 450 and your dram at 450 with 4 - 4 - 4 - 10, then you would have a very good overclock. If you could make your FSB at 500 and your dram at 500 then you would have some excellent overclock.

 

  1. A FSB of 350 with dram at 700 (4 - 4 - 4 - 12) and a multiplier of 9 would give you ~3.15Ghz
     
  2. A FSB of 450 with dram at 900 and a multiplier of 7 (4 - 4 - 4 -12) would give you ~3.15Ghz
     
  3. A FSB of 525 with dram at 1050 (5 - 5 - 5 - 15) and a multiplier of 6 would give you ~3.15Ghz

 

The third event would give you a far faster system than the other two. However, the current Core2 chipsets seem to have a limiting factor. I have made 1000Mhz on two sets of 6400C4's so I am a bit confused why this would be the case. I am in the process of obtaining some 8500C5's and will let you know what I find.

 

So if you can make #2 (as a theoretical event) then that is where I would take the dram. I am not certain how high you can take your dram or fsb but as long as you can maintain 1:1, I personally think you will be better off with the highest FSB and highest DRAM at a 1:1 level that you can make.

Link to comment
Share on other sites

If you keep a 1:1 relationship with your CPU:DRAM and clock your processor and ram up high with your current Micron D9's then I would stick with them. Keep in mind that the Micron D9's that were used in those products were the ones that Corsair tested and found NOT to make their stringent requirements for the 8500's (and above) and were down binned, so if you find that your ram is limiting you then a move to the 8500's may well be beneficial. My personal opinion is that you very likely can make some great 1:1 with that dram.

 

For example, if you were able to run your FSB at 450 and your dram at 450 with 4 - 4 - 4 - 10, then you would have a very good overclock. If you could make your FSB at 500 and your dram at 500 then you would have some excellent overclock.

 

  1. A FSB of 350 with dram at 700 (4 - 4 - 4 - 12) and a multiplier of 9 would give you ~3.15Ghz
     
  2. A FSB of 450 with dram at 900 and a multiplier of 7 (4 - 4 - 4 -12) would give you ~3.15Ghz
     
  3. A FSB of 525 with dram at 1050 (5 - 5 - 5 - 15) and a multiplier of 6 would give you ~3.15Ghz

 

The third event would give you a far faster system than the other two. However, the current Core2 chipsets seem to have a limiting factor. I have made 1000Mhz on two sets of 6400C4's so I am a bit confused why this would be the case. I am in the process of obtaining some 8500C5's and will let you know what I find.

 

So if you can make #2 (as a theoretical event) then that is where I would take the dram. I am not certain how high you can take your dram or fsb but as long as you can maintain 1:1, I personally think you will be better off with the highest FSB and highest DRAM at a 1:1 level that you can make.

Interesting, thanks for the heads up. :D:

 

I've heard that the P5W DH is at home (sweet spot) at 333FSB at 10 multi with the right ram. Ever heard of that?

Link to comment
Share on other sites

Interesting, thanks for the heads up. :D:

 

I've heard that the P5W DH is at home (sweet spot) at 333FSB at 10 multi with the right ram. Ever heard of that?

 

 

I have found with the P5W DH that if you set the 2:3 divider over 333fsb then you can encounter "issues".

 

I personally think that if you stay within the 1:1 ratio and do not play with the dividers then you will be (Limited by dram of course) able to make some find numbers with stability. I also have found that you very likely do not have to play much with the MCH and that setting to 1.6v is more than enough unless you are on high end water or LN and looking for extreme overclocks.

Link to comment
Share on other sites

I have found with the P5W DH that if you set the 2:3 divider over 333fsb then you can encounter "issues".

 

I personally think that if you stay within the 1:1 ratio and do not play with the dividers then you will be (Limited by dram of course) able to make some find numbers with stability. I also have found that you very likely do not have to play much with the MCH and that setting to 1.6v is more than enough unless you are on high end water or LN and looking for extreme overclocks.

 

I'll look into that thanks.

 

 

Side note: I found that the (Nayna) PC5400Pro are in fact a bit slower then the PC6400Pro (Micron) at the same 300FSB of PC4800 @ 600 Ram Frequency. I thought that it would be the same but it's not.

Link to comment
Share on other sites

That's understandable since one would expect the same bandwidth and the same timings would effect the exact same results. There are timings that are programmed into the modules that are not "tweakable" in the BIOS and thus there usually is discrepancies in final benchmark results. What one needs to keep in mind is that these differences are not all that great in the real world with a stopwatch. Load F.E.A.R. with the exact same system and only difference being a change in the modules. Use a digital stopwatch. Render a 10MB file in Photoshop and again use a digital stopwatch. Take a 500MB WAV file and compress it to MP3 using that stopwatch. Take a 2GB VOB file and convert it to an DIVX AVI with a stopwatch.

 

Then the realities kick in :)

Link to comment
Share on other sites

Using the WinRar built in benchmark and hardware test found in Tools appears to give you an idea of how well your ram works.

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/E6700/PC5400_RAR.jpg

PC5400 4-4-4-12 1325

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/E6700/OC6400_1351_2_20.jpg

PC6400 4-4-4-12 1351

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/E6700/OC6400_1374_333_10.jpg

PC6400 4-4-4-10 1374

Link to comment
Share on other sites

Using the WinRar built in benchmark and hardware test found in Tools appears to give you an idea of how well your ram works.

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/E6700/PC5400_RAR.jpg

PC5400 4-4-4-12 1325

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/E6700/OC6400_1351_2_20.jpg

PC6400 4-4-4-12 1351

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/E6700/OC6400_1374_333_10.jpg

PC6400 4-4-4-10 1374

 

Pretty minimal difference. Are you running the modules at their rated speeds ie 5400, and 6400? with those timings?

Link to comment
Share on other sites

Pretty minimal difference. Are you running the modules at their rated speeds ie 5400, and 6400? with those timings?

 

No, they were all on PC4800 speed.

 

I now have the PC8500C5D with the fan (why are they sold separate?) and have it at PC8500 @ 300FSB. At 3:5 this ram rocks even though it's still on the 800 strap! One thing I did notice is that if this ram is not running at it's rated speed it is slow. You don't get the full effect of this ram until you get it at it's rated speed!

 

Here is a pic

http://i11.photobucket.com/albums/a154/eastcoasthandle/PC8500/PC8500_CPUZ.jpg

CPU-Z

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/PC8500/8500C5D_1704_RAR.jpg

WinRAR (ram test) 1704

 

http://tinyurl.com/yexu92

Everest benchmark test

 

http://i11.photobucket.com/albums/a154/eastcoasthandle/PC8500/HL2LC_PC8500_1000FSB_125_22.jpg

HL2: Lost Coast

4x/8x everything else maxed!

Link to comment
Share on other sites

No, they were all on PC4800 speed.

 

Ah, that makes sense. If I had your system I would drop the multiplier to 8X and raise the FSB to 400MHz giving you 3.2Ghz with a 1600MHz FSB. Set your dram to 1000Mhz and retest.

 

I think you will be impressed by the results.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...