The Intel 6xx Series 875P-T thread

[Guest LightSpeed]

All of the higher end prescott are 'J' stepping. The lowest you can get for J is 3.0 (530E-J) up to 3.8 (570E-J).

Part #

BX80547PG3000EJ-530J

BX80547PG3200EJ-540J

BX80547PG3400EJ-550J

BX80547PG3600EJ-560J

BX80547PG3800EJ-570J

I think the 3.4J is the way to go...time for another CPU

What is taking so long for DFI to release an official BIOS for the 6xx CPU?

[BigStan]

If the 550J had been in stock anywhere when I bought my 530J, that's the proc I would have bought. I really liked my 550 D0 a lot.

[sam11811]

hey everybody a official bios has been released. It's about time long live dfi

[Guest LightSpeed]

I am not jumping for joy with this BIOS as it is the same crap as it was when it was BETA. D0 support is still broken, and it seems impossible to hit the same FSB as with previous BIOS using a 640.

To me it looks like DFI just made this official in order to alleviate the complaints of using a beta bios and nothing more. I think it is pathetic how DFI can sweep blatant bugs under the woodwork and call things official.

After using this board for over 4 months and waiting for it to work properly I am starting to give up hope.

How many boards have to stop working from customers using the current BIOS with older CPUs? I already see quite a few people having problems, and they ALL have D0 CPUs

Luckily I stopped wasting my time with the D0 and it is sitting on my desk instead of corrupting windows and BIOS like it does for everyone else.

WAKE UP DFI!

[Angry_Games]

Lightspeed:

 

you complain bitterly about overclocking problems when no guarantees have ever been made about overclocking, and never will be, not by DFI nor any company.

 

we make boards for enthusiasts, and they are designed for overclocking, but it would be the dumbest thing we could do to try and guarantee that you would get any type of overclock.

 

we guarantee the boards to run only at stock speed. We provide overclocking functions in the bios for you, but make no guarantees that they will succeed.

 

before you complain about that statement, find me ONE single motherboard company that will guarantee any overclock past default speed.

 

you won't.

 

I've already stated before that there is no bug with D0 cpu's. You should use the search function and find where I posted this (in fact, i think you were on the CC for email when i asked TW about this for you, yet you weren't satisfied with their answer). Since you weren't satisfied with their answer, here are a couple of suggestions:

 

1. modify the bios yourself to fit your needs (or find someone to modify it for you)

 

2. don't overclock

 

3. accept that overclocking is not guaranteed at all, only default speeds are guaranteed.

 

4. you always have the choice of moving to a different board if the the DFI board is not up to your liking. No one is forcing you to continue to be unhappy with your DFI board. THis forum certainly is not forcing you to continue being unhappy. I would rather see you with a different brand of motherboard but happy than to see you complain day after day about your overclocking woes with DFI. If you are really that unhappy, sell the board you have and move on to something different.

 

 

i only have D0 2.8 cpu's and I don't have any problems on any of our LGA775 boards other than my 2.8 ES is a piece of junk that won't overclock to even 3.2Ghz, yet my retail 2.8 does 3.7Ghz without a hitch on the 875P-T (265x14).

 

im not really sure what else you want us to do Lightspeed. Show me a board that does not run at default speeds and I will have it replaced for you. Overclocking...we provide this forum to try and help, and to have a community for help, but we cannot guarantee you any kind of overclock (as I repeat this again).

[OCZTony]

Guys

 

I have been bashing on this board now for a week or so and i have a few points I think i can now share.

 

1 The 6XX issue (14 multi) does need sorting and can be done thru bios, it is possible with windows now with software.

 

2 The boards seem to max out "around" 270 to 280fsb, i don't think we will get them higher than this without hardware mods.

 

3 hynix dimms either are good or bad, nothing in between, for some reason one of my pairs here is good but the others are real poor...they clock better on the Asus boards i ahve.

 

4 This poor clocking does not mean the board is slow, it just means you have to rethink how you overclock, the asus hates VX, the 875P-T will run it at 250 2-2-2 if you ram enough juice i have a feeling we need to look past the 14 multi and go for 1:1 around 250 to 260fsb max.

 

5 There are a host of other tweaks that we can apply to bios, dram latency, upclock ratio's, strap settings etc etc that should get the board even faster

 

6 Angry is doing a fine job here, he does NOT write the bios files and i know he DOES pass your comments on, I will help him here as i have contact with the engineers back in TW and i may be able to get some progress if DFI feels its worth it. You have to remember DFI are moving on to other chipsets and other boards, time is money and they are real busy at the mo with NF4 intel etc.

[Guest LightSpeed]

I am not talking about overclocking...I am talking about DEFAULT clock speeds, and I also hate to inform you that you are WRONG when you say there is not a bug, and it happens at ANY FSB..including STOCK. Notice how all the results you see are at DEFAULT clock speeds

Have a look here and tell me otherwise...this is BS already

I will re-paste the important data for you to see in case you missed it

 

D0 prescott @2.8GHz 'proper' L1/L2 cache

 

L1 cache: 10571

L2 cache: 7641

 

E0 prescott @3GHz: (new beta BIOS and apparently correct speeds)

L1 cache: 21069

L2 cache: 18484

 

N0 prescott @3GHz (note L2 is slower due to longer pipeline, not BIOS bug)

L1 cache: 21066

L2 cache: 16372

 

D0 prescott @3.4 with 'wrong' values:

L1 cache : 23878

L2 cache : 20948

 

Funny...the numbers seem to scale quite well according to CPU speed here...BUT then we have this crock of BS:

 

D0 prescott 3.4E with 'correct' values???

L1 cache: 12836

L2 cache : 9278

 

Here is another set of results using the EXACT same timings, and according to DFI the D0 score is correct, yet less than half of what it should be in the REAL WORLD

 

D0 3.4

L1 16 12835 -should be ~23k

L2 1024 9277 -should be ~20k

mem 3032

 

N0 640

L1 16 22470

L2 2048 17463

mem 2325

 

 

Data on D0 CPU bug

You are assuming I am 'complaining' about overclocking, yet the bug I mentioned and can easily reproduce (as can anyone with D0 CPU) is IN FACT A BUG, which not only reports the cache speeds wrong, but gives a higher memory bandwidth score

Before you claim that things are fixed and correct, please analyse the data

Now please tell me why a D0 should be any different with the values you claim are 'proper'

Ignorance is bliss I guess, and the fact that you/DFI claim this is not a bug is simply ridiculous.

[OCZTony]

I take it you are still talking about memtest86. I noticed a cache bug also with the 6XX over the 5XX but upon checking actual performance in XP the 6XXCPU is actually a lot faster.

 

I often wonder if i should get Samuel from X-86 secret.com over here so that he can see he needs to alter support for cache in memtest86 as if this is your issue, thats all it is.

 

Memtest had cache issues with a64, bartons, prescotts, etc etc and they all got ironed out eventually, I used to beta test for Samuel so i got to see the early builds that had issues.

[Guest LightSpeed]

It is not memtest with the issue, as using an older BIOS does not show this anomaly, and also the newer BIOS don't allow as high of an overclock due to the bad relationship between the cpu cache/memory bus; at least that is my theory.

How can memtest be to blame when the #s change with a BIOS flash?

More than that, the OLD #s were correct, and the 'new' ones are screwed, yet it has been 'fixed'? I don't buy it, sorry

I assume the cache bug you are referring to is when the L1 and L2 cache speeds show the same #, which seems to fix itself if you re-start the tests, yet this has no bearing on performance, unlike this BUG

Just sit back and clear your mind of anything else, and tell me HOW 2 different CPUs can have more than 50% difference in cache speed, and the slower one produces scores 20% higher...come on it is not that hard to figure out guys

[OCZTony]

memtest reads registers from bios for timings, it also calculates bandwidth a little like everst does in windows. Now if it can't see the registers correctly it will reports weird timings or bandwidth or cache etc.

 

Now, what you need to do for us, tell us which bios has the best overclocking.

we then could tweak the romsip to apply the 6XX support but have the old bios that clocks well.

 

You tell me what bios you think best and i will ask for the tweaks to be applied to the latest build.

[red930]

From a interested, 3rd party point of view, I would have to say that LightSpeed is correct in his analysis and the opinions expressed hold true to what many have experienced since upgrading from BIOS 11/30. His presentations of such facts, and please forgive me LightSpeed of stating this, leave much to be desired. Stating an opinion is one thing; stating an opinion that belittles the person(s) or company involved is another thing.

 

There are people of this forum who do not possess the technical knowledge, gained through many years of experience, that LightSpeed, me or many other inhabiters of this forum possess. No disrespect to the newer members of our forum, of course.

 

Each of us have a certain portion of experience or knowledge that can contribute to finding a solution to the problems that plague this particular implementation of the 875P chipset DFI has produced. Presenting that knowledge in a negative way does not advance the reconciliation of said problems; it only hinders the solution and can/does cause illwill between our members. That which I have seen since joining this forum.

 

Personally, I am an advocate of positive progression. I honestly believe in teamwork. I lived it when I was in the Marines and use it wholly on a daily basis at home and at work.

 

With that being said, let me be the first to contribute to a solution the to various difficulties that we face....

 

First, I would use 11/30 BIOS revision as the platform to start from. It offers the most "virgin" coding for which to implement later ascendants of the D0/D1 processor.

 

Second, the implementation of CPU revisions, no matter what speed, would be included. I realize the microcode has been implemented, yet the details of the microcode have not been fully embedded to take advantage of each particular CPU.

 

Third. timings are skewed. The bus speed of each individual CPU is correct, but the direct translation of the sub-components hasn't been addressed and is reported incorrectly by said software due to mis-translated bit reporting from the BIOS. This is not due to the monitoring programs fault, it is directly linked to the bit reporting each CPU uses. I could be wrong on this perceived theory though I think it would be one of the good starting points for analysis.

 

On a side note, I am working on compiling a BIOS for the 875P-T which I endeavor to support all LGA775 CPUs currently on the market.

 

Mods: please excuse me for my blatant disrespect for this thread. My post is intended for all of us to combine our knowledge and assist DFI in providing a solution to each particular problem it's customers' face.

 

Steven White

aka AceGoober

USMC Retired

[Guest LightSpeed]

What I have been saying all along is that the board is mis-reading the registers on the D0 stepping CPU, which is apparent when using memtest. There is no reason 2 CPUs should have 700mb/s difference in RAM bandwidth with the exact same timings, especially with lower cache speeds.

I guess my translation to layman terms needs some work, and I am a little tired of going over the same thing, so my patience on this is a little short, sorry

The BIOS that allowed the highest overclocking for the D0 was 11/30, but then it was the last one that read the registers on the CPU properly. I have all the BIOS from 11/30 up to present, including all the beta ones I could find. 11/30, 12/30, 1/25, 2/17, 2/24, 3/25.

With all memtest findings aside, with 3/25 and any CPU I can't go more than 271 in BIOS, but up to 1/25 275 is ok, and 11/30 280FSB is possible.

I will have an E0 CPU to add to the mix so I have D0, E0 and N0 for any further testing. All I would need is a couple extra BIOS chips :nod:

I am willing to help fix things, but it really bothers me when people say things are fixed that I can clearly see are not fixed. I tend to lose patience with it after a couple of months I guess...sorry