• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

e6850 purchase

S

sdadept

Limp Gawd
2FA
Joined
Jul 3, 2004
Messages
433
getting ready to pull the trigger on this one but I need to know what most of you are using for FSB on this processor. Should this processor be run @ 1333 with ddr3 and a motherboard that supports it or something else? I didn't post this in the memory sub board because I'm really looking for peoples very specific experience with this intel processor.

and does the FSB have to match the memory clock?
 
I have that processor and have my FSB bumped up to 378 (1512 FSB) to make the processor run at 3.4mhz

9hr Orthos stable with stock voltage, temps hold steady at 50c. Im sure I could get more if I played with the voltage but Im happy where it is.


Using 2gig Corsair Dominator DDR2 memory at 4-4-4-12



Gigabyte GA-P35-DQ6 mobo
 
great, you are hitting on exactly what I'm trying to figure out. Does the memory not need to run at exactly the same clock as the FSB? What speed is that ddr2 memory running at?

thanks
 
tsonka said:
I have that processor and have my FSB bumped up to 378 (1512 FSM) to make the processor run at 3.4mhz

9hr Orthos stable with stock voltage, temps hold steady at 50c. Im sure I could get more if I played with the voltage but Im happy where it is.


Using 2gig Corsair Dominator DDR2 memory at 4-4-4-12



Gigabyte GA-P35-DQ6 mobo
Click to expand...


To the OP:

What the memory is running at depends on what ratio Tsonka is running? If 1:1 (aka 2.0 on the Giga), then his memory is running at 2x 378 or 756. He didn't say what his memory is rated at so I don't know if he is underclocking his memory or not. This is where I get confused myself, but I believe the cpu fsb that corresponds to 378 is 1512 (378x4). Regarding your question on whether the memory should be run the same as the fsb, I'll leave that question to the more learned members here. I'm still learning.:confused:
 
The memory in most core 2 duos and quads is either dual or quad pumped meaning if your running your cpu at 1600 fsb than you need only 800mhz memory
 
Since I am very new to overclocking I am using alot of "auto" settings in my bios, the speed of the memory is at 907hz which is where auto setting put it at.

I am using this memory

http://www.newegg.com/product/product.asp?item=N82E16820145168


As for what it needs to be or should be, i dunno, hence why I use alot of auto settings. But being 9hr Orthos stable tells me it ok where it is at


The only settings I have changed are:

CE1= off

FSB to 378

Memory timings to 4-4-4-12

Memory voltage = +.3 over stock

Put PCIe freq to 100mz instead of auto because I read that some auto features will bump this up when the cpu FSB goes up.


Everythign else is either "auto" or at what ever setting the bios sets itself at.
 
sdadept said:
getting ready to pull the trigger on this one but I need to know what most of you are using for FSB on this processor. Should this processor be run @ 1333 with ddr3 and a motherboard that supports it or something else? I didn't post this in the memory sub board because I'm really looking for peoples very specific experience with this intel processor.

and does the FSB have to match the memory clock?
Click to expand...

Due to the low price of the G.Skill and Crucial Micron-D9 based DDR2 modules (80$), it would probably be most cost-effective to buy the E6750 and not the E6850.

With that RAM, the only reason to buy the E6850 is if u intend to overclock above 4.0Ghz (which is actually possible with this line of CPUs.
 
yea, I'm just looking for the fastest setup I can get. Money doesn't really matter as the company is paying for it. In fact, I'm not even really looking to overclock as stability is extremely important to me. I just want the fastest combo I can get without the thing being a space heater in my office (which is why I'm not doing the quads). So really I'm trying to figure out if DDR1333 memory or DDR2(something else) is the best way to go with this.

I'm still not clear on exactly how the memory matters to the FSB and how that all fits together.

from what I can tell, people are saying that the DDR1066 ? memory with FSB running at 1333 is good with the e6850? (my apologies, I'm really trying to understand this all).
 
Your comment about memory freq has me wondering now ............ should my memory be at 800mhz and not the 907 that bios has placed it at? Or is the increased memory freq a byproduct of increasing the timings and voltage?
 
My standard memory info blurb:

Basic OC Intel Math.
CPU runs at the CPU mulitplier x FSB.
Memory runs at 2x FSB (assuming a 1:1 memory multiplier, aka 2.0)

The chipset is designed for and in general runs best (more stable) when the memory is set to run at 2x FSB. However multipliers are available to allow other memory speeds.

-------------------------------------
PC6400 is probally fine for any C2D quad or otherwise based on the following.

Using a CPU mulitplier of 8 (all c2d have this multi except the E6300 and you are not even looking at that one)

Using an FSB setting of 400MHz (all good enthusiasts OCing boards should be able to do that)

Setting the memory mulitplier to a 1:1 ratio (the memory runs at twice the speed of the FSB, which is the way Intel designed the chipset, other ratios are available but in general you want to run at 1:1)

if you do the above your CPU is running at a very respectable OC of 8x400 = 3.2GHz and your memory is running 2 x 400 = 800 MHZ right on spec and not overclocked at all. This leaves you room to increase the FSB some more. Most decent OCing memory will get to 430MHz FSB (860MHz memory speed) before you need to loosen the timings to go higher, see my sig. So typically you should be able to get somewhere near 430x8 = 3.44GHz. Of course top notch cooling and proper setup etc etc.


Some C2Ds have multi of 9 which allows a lower FSB to reach the same CPU frequencies but you end up under clocking your ram. for example, 9 x 357FSB = 3.2GHz CPU speed but your memory is only running at 2x357 = 714MHz under its rated speed and you loose memory bandwidth. If next available memory mulitplier/divider (check each proposed board for the exact memory dividers it supports,) was 2.5 and you enabled it your CPU would still be at 3.2 but your memory would run at 2.5 x 357 = 892MHz which might be too high on fast timings and require you to loosen timings up just to boot. In this case ram rated higher than PC6400 would be preferred.

So it just depends on how hard you are going to push it, good very respectable OC without stressing stuff pc6400 will do very well. You want to push as hard as you can get something rated a bit higher. 4GB (4x1GB typically reduces the ability to overclock just a tad)

Thats the facts. Download the manual for the board you are considering and do the easy math based on available memory dividers and see what you can do (in theory).

Different boards label/call the memory divider different things but the manual should be able to tell you how to calculate the memory speed with the different dividers. If not just post back or try Google.

my 2 cents: go pc6400, better quality 4 4 4 12 that does not take a lot of voltage 1.9 v. The stuff that requires greater than 2.0V to work at advertised timings does not impress me. Any C2D OCed to 3.2 or so runs like a scalded cat and not a lot is gained for increases in speed over that vs the stress on components and the need for top of the line stuff. A 3.2 target can easily be met with a $130 board, and even pc6400 value ram and a $35 heatsink. But it just depends on what you want to do and the budget.


D9 are a much misunderstood thing, a couple of years ago Micron Memory made a batch of memory chips with a big long part number that started with D9, these chips turned out to be legendary in their ability to do very high overclocks. Well they have not been made in a long time. ALL Micron production starts with D9 now and they make good chips and typically Microns overclock well. BUT like all companies they speed bin the parts. There are great D9 chips, good D9, and not so hot (which probally get sold under some other name/re-branded). Crucial is the retail arm of Micron.
(This is the simple version of a very long complicated story. ) In summary, D9 these days means nothing more than it is Micron memory, there are suffixes to the D9 standard part number that do indicate the quality/speed of the part and you would need to post in the memory forum for advice as to what sticks might contain the more desirable Micron chips.
 
hi Bill, thanks a ton for explaining that in the cliff notes version!

so that brings up another interesting point. If memories are nominally running around 800 these days, what's with all the 1066 and 1333 memory running around? Are those quad pumped busses at lower speeds?

also note, I'm not looking to overclock at all. This is for a hardcore developer workstation that needs to be supremely fast but rock stable as well.

also, since I will be using a 2560x1600 30" monitor I will need serious bandwidth to the graphics card which unless I'm missing something relies on the FSB as well (thus making a high FSB speed important to this computer).
 
Thank you BillParrish, that was a very good explanation. We appreciate you taking the time to explain it so well.

I have one further question. In my case my Gigabyte P35-DQ6 board with a E6850 is running the cpu at 1333 and my memory at 800 with a 5:6 divider. This is the settings I get with everything in auto. Is this OK, does the divider lessen my performance? Or should I do as you stated above and run at 3.2 GHz?
 
michaLcoughliN said:
mine should be arriving tomorrow :D then overclocking.. then thread based on my experiences :p
Click to expand...

Congradulations, they finally shipped your stuff:D NCIX sent me a confirmation on the E6850 I ordered, so Mr. Purolator should be visiting me as well on Monday morning:p
I don't expect to get very far with my old DS3 but you never know. Looking forward to seeing how you do with the P5K Deluxe, I will probably have to get one, or a P35 Gigabyte, not sure which yet.

sdadept: I'd stay away from DDR3 as it isn't ready for prime time yet:p Just wanted to suggest Mushkin memory as another option, I have had really good results with it:cool:
 
Jacez44 said:
Due to the low price of the G.Skill and Crucial Micron-D9 based DDR2 modules (80$), it would probably be most cost-effective to buy the E6750 and not the E6850.
Click to expand...
Its early days yet. But, from what ive read, the e6850 is clocking significantly higher than the e6750.
 
BillParrish said:
my 2 cents: go pc6400, better quality 4 4 4 12 that does not take a lot of voltage 1.9 v. The stuff that requires greater than 2.0V to work at advertised timings does not impress me. Any C2D OCed to 3.2 or so runs like a scalded cat and not a lot is gained for increases in speed over that vs the stress on components and the need for top of the line stuff. A 3.2 target can easily be met with a $130 board, and even pc6400 value ram and a $35 heatsink. But it just depends on what you want to do and the budget.


D9 are a much misunderstood thing, a couple of years ago Micron Memory made a batch of memory chips with a big long part number that started with D9, these chips turned out to be legendary in their ability to do very high overclocks. Well they have not been made in a long time. ALL Micron production starts with D9 now and they make good chips and typically Microns overclock well. BUT like all companies they speed bin the parts. There are great D9 chips, good D9, and not so hot (which probally get sold under some other name/re-branded). Crucial is the retail arm of Micron.
(This is the simple version of a very long complicated story. ) In summary, D9 these days means nothing more than it is Micron memory, there are suffixes to the D9 standard part number that do indicate the quality/speed of the part and you would need to post in the memory forum for advice as to what sticks might contain the more desirable Micron chips.
Click to expand...

Sounds like someone hasn't done their RAM research. D9GMH and D9GKX are the IC's you want to get. Those two chips replaced D9DQWs, which were actually labeled fat bodies. I don't know why people call GMH fat bodies too, since it's actually a smaller IC than some on the market. DQW's were called that since they were actually pretty big.

D9GMH and D9GKX, depending on bin and the PCB layers, can achieve DDR2 1000mhz @ cas 4 with around 2.3-2.4V ish, atleast from what I've seen. Bill is right in that if someone says some sticks are guaranteed D9's, you should probably search around a little more (though usually when enthausists say D9, we mean GMH or GKX, GCT and those others will still clock decently, but not as well as the others).

As for sticks which have been consistantly been found as D9GMH//GKX, those would be...:

Buffalo Firestix DDR2 800 cas 4
Crucial Ballistix DDR2 667 cas 3
Crucial Ballistix DDR2 800 cas 4
Crucial Ballistix tracer line (667 and 800)
Crucial Tenth Anniversary
TeamGroup Team Xtreem DDR2 667 cas 3
TeamGroup Team Xtreem DDR2 800 cas 4
G.Skill DDR2 800 HZ line

Almost all PC8500 are D9GMH or D9GKX, and some PC8000, though I'd be a little more skeptical since I've heard of elipda IC's hit 1000 mhz. I didn't list any OCZ, but OCZ does have D9GMH//GKX sticks, but they're higher end. Corsair does as well, but with all the revisions corsair has of their memory, I wouldn't bother trying to look for those specific sticks, especially since Corsair seems to like to bin all their chips to hell.
 
so my question is still then, is it ok for the FSB and the memory clock to run differently. Because most of the setups I see people talking about with e6850 have the FSB running at 1333 but the memory running at 800 or 1066. What gives?
 
Sorry Real Life kicked my ass for a couple of days.

To aswer your excellent question about the memory speeds. Its all about performace increase and the fact that memory has a much quicker design cycle than CPUs so memory tends to evolve to faster speeds to meet market demand from OCers.

Speaking of Intel chipsets, Intel designed the CPU and chipset for the memory to run at 2x the FSB. This has to do with optimizing data transfers and timing on the buss between the CPU, its internal cache, the MCH/northbridge which is the memory controller chip and the Memory sticks themselves. They could have picked 1.8x FSB or 2.5xFSB I guess for what was probally a ton of good reasons they went with 2xFSB. And when it is all running at 2x FSB everyone in the boat is rowing at the same speed.

Well back in the stone ages where I came from, mid 1980s, memory was fantastical expensive and slow and at first the cpus were faster, so you had to use a divider to slow down the FSB to accomodate slow memory. You could also upgrade your CPU and use your old memory. Fast forward an epoch (about 10 years in computer time,early/ mid 90's) and memory manufacturing had be perfected, speed vastly increased, standards body formed to create spec etc. and to the point, crazy people where rewiring memory clock chips to speed up the memory etc. and Overclocking became a mainstream market for the board and memory manuf who saw there was money to be made. Also there were "families of CPUs" running at a lot of different buss speed (66MHZ 100MHz 133MHZ) . AMD got into the act. So lots of different CPUs running at different buss speeds and now with memory being cheaper and the ability to go faster, someone got the idea of using mulitpliers on the memory clock frequency to allow faster memory to work with slower cpu buss speeds (with associated buffering and timings waits to keep stuff talking properly). This would increase memory bandwidth (the amount of information being transfered in a unit of time could be increased) and thus system performance.

[I typed that fast, based on my recollections, its not all exacty true but thats the jist of it. ]

Now to your specifice stated goals. Everyone in this forum is geared to OCing and while stabiliy is a concern, fast and faster is the game. In the server world stability is all it is about (well and power usage). The trouble with memory mulitpliers is the buffering and wait states imposed to keep things from going insane. People in the boat are no longer rowing in unison. The memory guys are rowing their ass off. The boat goes faster but the cpu guy has to "stutter" row to keep from wacking the memory guys oars. The boat is faster but its tricky. Everyone has to be rowing just exactly "right" to keep from messing other rowers up.

So in general, running memory in synch (2 x FSB) on Intel chipsets is inherently more stable and if you raise the FSB in general but keep the 1:1 ratio (2x FSB) everyone is rowing at the same speed. From a design standpoint for stability it would in general be preferred by an engineer who designed the system as we tend to be tidy people and dont like complications.

A well done memory mulitplier will work fine and make the boat go faster but the chance of something getting out of wack when stressed or heated is there. OCers will test and use whatever makes the boat go faster so manuf of boards and memory are only too happy to provide the needed settings if it will sell product. It works fine and who cares if every once in a while your game freezes, reboot and give it another go.

Post too damn long now, you get the idea, nothing wroing with running a memory mulitplier to proivde more system performance, a systems engineering purist would not care for it but its proven technology. Some minor chance of compromising system stability if things are pushed too far, and due to the buffering and wait states you dont get quite as much of a boost as one would think, its still can improve performance greatly if your application is memory intensive.

So if you get a 1066 FSB cpu (266 "actual" FSB) 2x266 would indicate memory speed of 533MHz but for more performance you put in 800Mz memory and run a mulitplier of 3.0 so memory runs 266 x 3.0 = 800MHz right on spec. And with a good quality board there would probally never be an issue. (and it would be faster in most applications)

A server administratior would probally not do the above and be happy with 533MHz memory for unquestionalbe rock solid stability.


Dear God I am unstoppable once I get going.
 
You must log in or register to reply here.
Back
Top