Xeon Owners Club

[lexluthermiester]

B67 and Q67 are not supposed to work with Quad Channel, ECC and Lga2011 socket These are lga1155 chipsets. But chinese made them work.

We're not talking about Chinese boards, we're talking about ASUS boards.

I'm sure that ecc feature doesn't work and the board sees those sticks as a regular ram without this middle chip.

Registered DIMMs will not run in such a mode as it does not exist. The ECC & buffering chips on those types of DIMMs can not be disabled as they are the direct connection to the IMC on the CPU. There's something else going on there. What does HWInfo say about them?

 

[Edwired]

We're not talking about Chinese boards, we're talking about ASUS boards.

Registered DIMMs will not run in such a mode as it does not exist. The ECC & buffering chips on those types of DIMMs can not be disabled as they are the direct connection to the IMC on the CPU. There's something else going on there. What does HWInfo say about them?

I'm curious about that too

 

[Caring1]

Not all RDimm has ECC.
ECC only works on server boards designed for it.

 

[Edwired]

I only spotted this in Aida64

Error Correction
ECC Supported, Disabled
ChipKill ECC Supported, Disabled
RAID Not Supported
ECC Scrubbing Supported, Disabled

Does this mean if I was to buy ecc type ram the above would be showing enabled. Head scratching moment here

 

[lexluthermiester]

Not all RDimm has ECC.
ECC only works on server boards designed for it.

Incorrect. All Registered DIMMs are ECC enabled. Additionally, X58 supports ECC. I'm running ECC in my personal T3500's.

I only spotted this in Aida64

Error Correction
ECC Supported, Disabled
ChipKill ECC Supported, Disabled
RAID Not Supported
ECC Scrubbing Supported, Disabled

Does this mean if I was to buy ecc type ram the above would be showing enabled. Head scratching moment here

If you install ECC RAM, those fields would then show enabled.

 

[Caring1]

Incorrect. All Registered DIMMs are ECC enabled. Additionally, X58 supports ECC. I'm running ECC in my personal T3500's.

"Although most registered memory modules also feature error-correcting code memory (ECC), it is also possible for registered memory modules to not be error-correcting or vice versa. Unregistered ECC memory is supported and used in workstation or entry-level server motherboards that do not support very large amounts of memory.[1]"

A direct quote from Wikipedia on Registered Dimms.

 

[lexluthermiester]

"Although most registered memory modules also feature error-correcting code memory (ECC), it is also possible for registered memory modules to not be error-correcting or vice versa. Unregistered ECC memory is supported and used in workstation or entry-level server motherboards that do not support very large amounts of memory.[1]"

A direct quote from Wikipedia on Registered Dimms.

Ok, there is that. However, I've never seen it nor ever heard of it as there would be little point to the Register Buffer without the added features of ECC. So in practical reality, you're not going to find it the wild.

 

[Rumcajs997]

Unfortunately I'm not able to answer why it works. All I know is that it works and seems like I am not the only one that got it working - https://www.overclockers.com/forums...Vaw3tJPuB79g4a78z7AQi9aVY&cshid=1622350033922

 

[Zyll Goliat]

Well here is the thing RAM controllers actually are built-in to the CPU instead of the north/southbridge on X58 BUT also seems like something else is also preventing Reg ECC to work on those branded motherboards...maybe it's possible to change something via software/bios and that could fix this or is some additional "thing" on the workstation boards that making REG ECC operational......but yeah seems like @Rumcajs997 and some other guys are correct and that only 2rx8 memory Reg ECC working....why and how...who knows....

 

[Edwired]

Well here is the thing RAM controllers actually are built-in to the CPU instead of the north/southbridge on X58 BUT also seems like something else is also preventing Reg ECC to work on those branded motherboards...maybe it's possible to change something via software/bios and that could fix this or is some additional "thing" on the workstation boards that making REG ECC operational......but yeah seems like @Rumcajs997 and some other guys are correct and that only 2rx8 memory Reg ECC working....why and how...who knows....

Sure that is a valid point about the memory controller been inside of the cpu instead of the north/south bridge unlike the p45 chipset that had the memory controller in it. In this case results and people testing out the theory has proven that it works with the 2rx8 ram is pretty amazing results. In my mind something in the bios/dram area of things causing the no post issue usually tells me the dram spd timing is abit funky.

To change the subject abit
Question for anyone if they can answer it about the Write CAS Latency (tWCL) regarding to the xeon x5675 and Corsair XMS3 CMX4GX3M2A1600C9. Now the fun part is that the ram I have that I looked at the timings, it stated it needs tWCL to be at 9T according to the spd specs and the bios selected 8T which works ok but when I set it to 9T manually. The whole system point blank refused to post regardless. Kinda strange

 

[Rumcajs997]

To be honest those transfers for Triple Channel are rather weak IMO. It`s all standard settings, no OC yet but they are on the level of Dual Channel Sandy Bridge - https://www.techpowerup.com/forums/...da-64-cache-and-memory-benchmark-here.186338/

It`s strange as I have got Chinese Atermiter X79 Turbo with Quad Channel ECC Registered 1600mhz and they are fast as hell, as far as I remember I get Aida64- 45gb/s where these triple channel are getting less than 20GB/s.

BTW. Do you guys remember the temperature of your X5670 Xeon? I know it`s 32nm so it will be hot but IMO it is too hot. With Noctua nh-c12p se14 at max fan speed, I am getting 65C under Prime95 load with all settings standard. That seems way too high, I was hoping to get at least 4.2Ghz with this cooler but if it`s already 65C without touching anything... I think it`s installed correctly as when I touch the cooler closer to the bottom it it`s pretty warm so it transfers the heat.

 

[Zyll Goliat]

Well that memory bandwidth is about right for the X58 platform If I recall correctly temps on my E5645 OC on 4,2Ghz under the prime 95 was around 85C so yeah those Xeons knows to get hot tho' depends some people can reach higher clocks with less power but all in all it is kinda normal if you reach 80c under the Prime95 that's mean that in normal tasks like gaming you probably never going to go above 65c more or less.....

 

[80-watt Hamster]

BTW. Do you guys remember the temperature of your X5670 Xeon? I know it`s 32nm so it will be hot but IMO it is too hot. With Noctua nh-c12p se14 at max fan speed, I am getting 65C under Prime95 load with all settings standard. That seems way too high, I was hoping to get at least 4.2Ghz with this cooler but if it`s already 65C without touching anything... I think it`s installed correctly as when I touch the cooler closer to the bottom it it`s pretty warm so it transfers the heat.

One of my 5670s can crunch BOINC at 4.2-4.3 on a 212 EVO in the mid-70s at modest fan speeds. Reducing (or would that be increasing?) the memory divider can get you a few C if you don't need the bandwidth. My Sabertooth also seemed to give it more volts than it needs, so you may be able to set a slight negative offset.

As far as P95 temps, my somewhat faulty memory is telling me that 60+ is pretty common for small FFTs on basically any processor on which I've run it.

 

[Edwired]

To be honest those transfers for Triple Channel are rather weak IMO. It`s all standard settings, no OC yet but they are on the level of Dual Channel Sandy Bridge - https://www.techpowerup.com/forums/...da-64-cache-and-memory-benchmark-here.186338/

It`s strange as I have got Chinese Atermiter X79 Turbo with Quad Channel ECC Registered 1600mhz and they are fast as hell, as far as I remember I get Aida64- 45gb/s where these triple channel are getting less than 20GB/s.

BTW. Do you guys remember the temperature of your X5670 Xeon? I know it`s 32nm so it will be hot but IMO it is too hot. With Noctua nh-c12p se14 at max fan speed, I am getting 65C under Prime95 load with all settings standard. That seems way too high, I was hoping to get at least 4.2Ghz with this cooler but if it`s already 65C without touching anything... I think it`s installed correctly as when I touch the cooler closer to the bottom it it`s pretty warm so it transfers the heat.

All depends on the room temp that plays the factor of the cpu and heatsink temp as i have the same heatsink with noctua redux 1500rpm fan but it controlled by the cpu temp + speed most of the temp on the cores hover around 45c to 53c but the overrall cpu is less than 45c other times it hover less than 35c depending what I'm doing on the pc It pretty good heatsink no issues since.

As for the bench mark this what I'm getting

As the number goes up with the ram speed and the northbridge clock all depends on the ram you have. It's best to find the balance between the two as it running actual 9-9-9-24 as it a bug in the software showing that as it the same in cpuid it shows 8-9-9-24 when the pc goes to sleep or wake up it shows funny numbers

 

[Rumcajs997]

Just so you know, these 2gb Registered 2rx8 sticks are working perfectly fine too. So I got 6x4gb = 24gb for the price of regular 2x4gb = 8gb non ECC ram and 6x 2gb for 13$.
BTW. I turned off this Linux boot that motherboard has and it is booting up quicker now but still painfully slow comparing to other boards I've seen. I think it must be something with MemOK system as when I turn it on with just 1x 2gb stick it booted up in just like 3s. The more sticks I'm adding, the longer it boots. With 6x 4gb it takes like good 15s.

 

[Edwired]

As I have booted my set up many times it takes about 40 to 45 seconds to post due to that it reboots twice. As this happens if the ram or uncore is higher than the stock state if I lower the uncore and dram speed it post every single time therefore faster boot times. This is what I'm thinking

 

[Edwired]

Looks like I'm in the 200 plus blck club
Same voltages I had before but lower cpu multi to 20x

 

[lexluthermiester]

Looks like I'm in the 200 plus blck club
Same voltages I had before but lower cpu multi to 20x
View attachment 202836

Nice!

 

[Edwired]

Look like I got a good chip. Now I got to do some testing to see if it can hold a pint of volka without spilling it

 

[lexluthermiester]

Look like I got a good chip. Now I got to do some testing to see if it can hold a pint of volka without spilling it

That BCLK at that multi and that voltage? Heck yeah you got a good one!

 

[Edwired]

Once I put it to sleep it gone into a deep coma and never recovered other than that I think it the qpi link is having a black hole somewhere in that area of the blck 213. I don't think the uncore having that issue as I was able to go higher but it limited by the dram ratio due to 1600mhz or 2000mhz is available to me once I dropped the lowest dram and uncore and booted to the win 10 and put it to sleep again same thing happened never recovered as I can tell that the mem ok led light would come on solid and not blink once when everything is working. God the OC itch is only starting to get to me now

 

[Rumcajs997]

I wonder, those Xeons don't support more than 1333mhz RAM but is it possible to OC them above that limit? Is there some kind of separate OC of Ram and CPU like some old AMD systems had eventually?

 

[Zyll Goliat]

I wonder, those Xeons don't support more than 1333mhz RAM but is it possible to OC them above that limit? Is there some kind of separate OC of Ram and CPU like some old AMD systems had eventually?

They can work up to the 2133Mhz as far as I know and yes you can OC memory frequency and timings on that motherboard

 

[Edwired]

I'm thinking it the blck that causing the sleep issue above 200. But if I have it on exactly 200 the sleep issue don't happen. Would be nice to get higher ram but seems hard to get a hold of 1866 or higher

Only concerning thing is why is the voltage reading on the board is abit less compared to PSU readout. Or is it something on the board that wearing out overtime?

I did remember that the atx 24 pin that needs adjusting then the voltages came back up abit as I say it needs cleaning or tightening down have to do that another day.

Might need to get contact cleaner or white vinegar to drop it into the atx slot and connector as had abit of spare time to see if it changes the voltages 12v - 5v - 3.3v in realtime by wiggling it side to side (rocking effect) with the system is on. To my surprise it actually works as the old 3.084v jumped up to 3.248v same with the old 4.875v jumped to 5.047v and 11.850v jumped to 12.194v as old values when the system was running last night I throw up a screen shot later.

Worth checking if you notices the voltages differences from the power supply and motherboard

Here the improved voltages reading it pretty close to the power supply reading according to the dmm as I was using pcsx2 to use the gpu to render the graphics for metal gear solid 3 snake eater. If you were to compare it as found on post #7,292 it pretty small margin for the loss of power getting to the conpoments on the motherboard. For the overclock side of things not sure how much of the effect it had before and after. As it was giving me alarm bells in my mind seeing it going lower and lower over time

 

[Rumcajs997]

I'm honestly surprised how shitty the "Noctua nh-c12p se14" is. 14cm fan set at max speed and my x5670 @ 4.1ghz 1.275v throttles under Prime95 after 3min. And it throttles at 3.85ghz 1.225V as well after 10min. The cooler is installed correctly as when I touch its heat pipes under load they are hot as hell. I need to get rid off it and buy some proper cooler.