Dell Workstation Owners Club

[lexluthermiester]

Internal Speaker for giving Diagnostics Beep Codes

This.

 

[ManGupta]

Be careful putting heatsinks on the VRM in these. There are components in between the MOSFETs that are sticking up in the way of doing that.

Did not get you. Exactly what components are creating problem. The original Dell small Aluminium Heatsinks seems to be sitting very well over the VRM area without hurting any VRM Component.




But you need to put a Silicon Thermal Pad below the Aluminium Heat Sink for VRM.
screen capture windows 7


I took the Scythe NinJa off of the T5500 MB and tried the Dell 0P041K cooler. The mounting bolts are all wrong, but the size and bolt pattern are correct. 50% more heat pipes and comes with a fan. Male to male bolts is the issue. I don't have an original heatsink to try the old hardware. But this is the cooler I would go to on a t3500. The fins are folded down on the sides to form a duct so all the air from the fan passes over the heatpipes.

Actually though ...... some amount of airfow escaping thru the side of fins ........ may be blessing in disguise ... as these escaping airflow .... will pass over the adjacent VRM Mosfets ... cooling them down a bit.


A little more on fans for ManGupta. There are variable speed fans that are not PWM. They have a blue thermistor sticking up from the motor in the back. They vary speed based on the temperature of the air passing through them. This may be useful in your environment. The thermistor is a resistor, and if you add a resistance between the legs it will be a resistor in parallel which willl make the motor run faster. Normally used as an exhaust fan. Some Nidec Beta V TA350DC comes like this. You can see the sensor here.
https://www.aliexpress.com/item/Fre...Thermal-Sensor-Fan-12V-1-8A-3/1956098195.html
This is a 38mm thick fan. This lets fans move more air without a lot of RPM. I always prefer these over 25mm thick fans.

Thanks for the Aliexpress fan link ....... Fan seems impressive with 5700 max rpm and 130 cfm. GREAT.

The cooling mods I posted above would apply to the T3400 directly because it's a classic BTX. The T3500 is a modified BTX, but the cooling works on the same principles. 1 powerful fan, and good control of airflow. But instead of a duct they use the HDD tray, the RAM shroud, and the case and cover to form one around the heatsink. 100*F. ambient is an interesting problem. I think my BTX overclocking cooling mods could handle that easily. Too bad I wasn't asked this a week a go when I had my 65W Q9505S quad core in there. I might have gotten Prime 95 running in the high 40s.
But I did the BIOS mod so it's probably not supported anymore. The temperatures I got above on the X5470 would represent a 10% overclock on a Q9650 CPU.

Its Winter season now. Any Cooling Mods done by me can be Authentically Tested here only at the Peak of Summer season (May-June). Will give cooling Mods one more shot then. At this time Cooling Mods may seem to be Okey Dokey ..... But during the Peak Summer ..... it may come out that what was okey in Winter may not necessarily pass the Reality Check in Peak Summer.

 

[Retrorockit]

I have tried to add heatsinks to these as part of my modding. My T3500 required drilling holes in the back of the heatsink so it could sit flat on the MOSFETs. The T5500 had too many of them to do that. If you want to add a heatsink check for this. I would suggest separate sinks for each MOSFET because without a spring loaded mount, and thermal paste, expansion and contratcion of the heatsink will cause it to come loose if it's glued on. Like Lex said they made changes in that area that eliminated the need for it. The ones with the heatsink may have different parts there. Photographs of the heatsink version are in all the manuals. But they seem to be uncommon in the wild. I suppose they could be using a thick thermal pad to absorb the difference in height. The MOSFETs aren't all the same height either.

 

[ManGupta]

I have tried to add heatsinks to these as part of my modding. My T3500 required drilling holes in the back of the heatsink so it could sit flat on the MOSFETs. The T5500 had too many of them to do that. If you want to add a heatsink check for this. I would suggest separate sinks for each MOSFET because without a spring loaded mount, and thermal paste, expansion and contratcion of the heatsink will cause it to come loose if it's glued on. Like Lex said they made changes in that area that eliminated the need for it. The ones with the heatsink may have different parts there. Photographs of the heatsink version are in all the manuals. But they seem to be uncommon in the wild. I suppose they could be using a thick thermal pad to absorb the difference in height. The MOSFETs aren't all the same height either.

Yes the purpose of using a thick thermal pad must be to absorb the difference in height. How else one would mount the heat-sink in such an uneven area.

But your point of expansion and contraction of aluminium leading to loosening of the firmness hence the effectiveness of heat-sink is valid.
The Silicon Thermal Pad seems to be very soft rubber-like substance with some thermal conductivity that fills the spaces between the uneven heights of VRM Components.

But the contact with heat will loosen over a period of time. Perhaps two screws containing spring (like those in CPU Heat Sink) if tightened in the two holes (in addition to the thermal pad) may secure firmly the heat sink.

That is all this if Heat Sink is really required. Else as Lex said it may not be needed at all. It may be only required in case CPU draws very high power (much more than rated 130 Watts most X58 processors consume). This scenario can only occur in case of Extreme Overclocking which Dell MB do not allows at all.

 

[lexluthermiester]

That is all this if Heat Sink is really required. Else as Lex said it may not be needed at all. It may be only required in case CPU draws very high power (much more than rated 130 Watts most X58 processors consume). This scenario can only occur in case of Extreme Overclocking which Dell MB do not allows at all.

That about sums it up.

 

[Retrorockit]

We actually do overclock the T3500, and dual CPU OC work is being done on the T5500/7500 also. But your ambient temperature situation may make it desirable to do this.
I never could find that heatsink available anywhere, so I can't really say what solution Dell used for the issues I encountered. Alphacool makes some small copper finsinks for MOSFETs.
https://www.amazon.com/Alphacool-Copper-Heatsinks-6-5mm-10-pack/dp/B077VQTB6Q
Enzotech also.
https://www.aquatuning.us/air-cooli...9I5ZAIUIwnaIZ0VXLQV5DqXtsbiX4mpRoCzIcQAvD_BwE

 

[sepheronx]

I'm still waiting on the Dell 5 pin to 4 pin fan adapter. Then I got to get a new GPU as I used the gtx1060 for my home theater/retro gaming build.

But can someone guide me through getting that w3680 to 4ghz?

 

[ManGupta]

I'm still waiting on the Dell 5 pin to 4 pin fan adapter. Then I got to get a new GPU as I used the gtx1060 for my home theater/retro gaming build.

But can someone guide me through getting that w3680 to 4ghz?
For OC You Need to go thru that thread started by Retrorockit
https://www.techpowerup.com/forums/threads/throttlestop-overclocking-desktop-pcs.235975/

yeah Shipping from China is a loooooong wait, I too am waiting for it,

although I went ahead and added 5 fans drawing power directly from PSU and running at full speed.

Moreover I do not trust the PWM reliability of Dell MB (Dell has designed it for running silent (in an office environment), not for high CFM for cooling).

Whenever I did stress testing (before adding 5 fans) Core temps went over 100 C leading to thermal throttling ....... still Dell's Two Front Fans speed did not increased ....... this makes me wonder whether or not to opt for PWM feature of Dell's MB using Dell 5 pin to 4 pin fan adapter. Because ........... Now, after adding 5 fans running on full speed all stress tests completes successfully without any Thermal Throttling.

 

[SoNic67]

SLG8XP519T But support is hard to find for that one.
The ICS932S421BGLF is supported by SetFSB and is the PLL for the T5500, and T7500. So if this method appeals to you a single or dual CPU setup on those should work. An SATA controller card is reported to be necessary to prevent OS crashes. The T5500 MB is a direct replacement for the T3500. Cost about $40. But the chipset might prefer RDIMMs.

confirms 16GB RDIMMs on some X58 MB for 96GB RAM. Says not all support it. Of course no one over there is going to try it on a T3500

I have a T3500 with X5675 CPU. Right now I am "full" 24GB of RAM and although I am tempted to try the RDIMM, I can't see the gain. I am already using 4GB od those in PrimoCache as a cache, and it's enough IMO. Maybe I'll play more with this: https://www.ebay.com/itm/49Y1415-8GB-DDR3-1333MHz-RDIMM-Memory-IBM-System/181697687564
6x$23=$138... is it worth it thou?

Also I have a 3Ware 9650SE SATA RAID controller on my board, so I would rather try to OC (after moving my SSD on that controller).

 

[Susquehannock]

But can someone guide me through getting that w3680 to 4ghz?

Simplest way I found is install Intel XTU (here) and change all multi to 30x. ThrottleStop is another option of course.

I have a T3500 with X5675 CPU. Right now I am "full" 24GB of RAM and although I am tempted to try the RDIMM, I can't see the gain. I am already using 4GB od those in PrimoCache as a cache, and it's enough IMO. Maybe I'll play more with this: https://www.ebay.com/itm/49Y1415-8GB-DDR3-1333MHz-RDIMM-Memory-IBM-System/181697687564
6x$23=$138... is it worth it thou?

Besides buffering, only advantage is price. If decide to try, you may want to hold off until someone can confirm whether RDIMM modules are even possible in a T3500/X58 board. As you may have seen in post #360, i have three of those same modules (#49Y1415) unbranded IBM. Work great in T5500 board, but will not boot with them in any way in the T3500.

 

[sepheronx]

thanks!

 

[ManGupta]

i have three of those same modules (#49Y1415) unbranded IBM. Work great in T5500 board, but will not boot with them in any way in the T3500.

Susquehannock,

Did you tried with a X56XX Processor in T3500 ... ??

 

[Susquehannock]

Susquehannock,

Did you tried with a X56XX Processor in T3500 ... ??

post #364

[edit] CPU = W3680.

Yes. Meant to mention that. Only tried those IBM modules in T3500 in my signature so far. Since it is a single QPI chip was considering whether that was a factor. System runs so well with overclocked W3680 that I have been reluctant to go back to quad core and drop a X5687 to try it. Found X5690 at good deal so may try that once it arrives just to confirm if QPI is a factor.

 

[ManGupta]

post #364

I strongly suspect that the Integrated Memory Controller on W36XX series processor is the culprit. Actually I ran Sis Sandra Software and it says that Registered ECC DIMM upto 48 GB is supported.

 

[Susquehannock]

I strongly suspect that the Integrated Memory Controller on W36XX series processor is the culprit. Actually I ran Sis Sandra Software and it says that Registered ECC DIMM upto 48 GB is supported.

Sure hope you're right. Having 3x8gb=24 already, attaining 48gb would mean another $66 vs over $200 switching to UDIMMs. At which point my T5500 board would be of no use unless going dual CPU.

 

[lexluthermiester]

I strongly suspect that the Integrated Memory Controller on W36XX series processor is the culprit. Actually I ran Sis Sandra Software and it says that Registered ECC DIMM upto 48 GB is supported.

It's could very well be possible. Someone needs to take the plunge and try it though..

 

[ManGupta]

Sure hope you're right. Having 3x8gb=24 already, attaining 48gb would mean another $66 vs over $200 switching to UDIMMs. At which point my T5500 board would be of no use unless going dual CPU.

Just for testing RDIMM full 48 GB capacity may not be necessarily required (unless we want to test the capacity as well). For just testing if T3500 MB with X56XX series Processor supports Registered DIMM or not ....... can be tested without full capacity ..... even single channel RDIMM.

Sure hope you're right. Having 3x8gb=24 already, attaining 48gb would mean another $66 vs over $200 switching to UDIMMs. At which point my T5500 board would be of no use unless going dual CPU.

T5500 is good system, not just 2 CPU but also it has active chipset cooling (unlike T3500) and also larger capacity PSU.

 

[Retrorockit]

I have a T3500 with X5675 CPU. Right now I am "full" 24GB of RAM and although I am tempted to try the RDIMM, I can't see the gain. I am already using 4GB od those in PrimoCache as a cache, and it's enough IMO. Maybe I'll play more with this: https://www.ebay.com/itm/49Y1415-8GB-DDR3-1333MHz-RDIMM-Memory-IBM-System/181697687564
6x$23=$138... is it worth it thou?

Also I have a 3Ware 9650SE SATA RAID controller on my board, so I would rather try to OC (after moving my SSD on that controller).

X58 at OCN is confirmed to support 48GB w/o resorting to RDIMMs.
The RAID controller would be interesting on a T5500. The T3500 PLL chip is not supported by SetFSB, but maybe Clockgen, or SetPPL, or even RWEvertyhing may allow access.
The T5500/7700 PLL is supported. And it's reported that SATA crashes unless you use a PCIe controller card.

@ManGupta I'm digging into my parts pile and getting the 0P041K heatsink mounted onto a T5500MB. The key to swapping heatsinks on these is the threaded spacers in the Thermalright mounting kits. They're the right thread and the right height ( or can be adjusted if needed). I'm trying to duplicate the spring loaded OEM mounting they came with.

 

[Susquehannock]

Just for testing RDIMM full 48 GB capacity may not be necessarily required (unless we want to test the capacity as well). For just testing if T3500 MB with X56XX series Processor supports Registered DIMM or not ....... can be tested without full capacity ..... even single channel RDIMM.

T5500 is good system, not just 2 CPU but also it has active chipset cooling (unlike T3500) and also larger capacity PSU.

This is true. Was referring to my particular situation using T5500 board in T3500 case. No need for the extra power as configured and active cooling on northbridge does not become a major factor unless utilizing the CPU riser.

What I gather from my experiments thus far. If it does turn out to be the memory controller on X56xx that means no overclocking with RDIMMs since, unless I'm mistaken, all X56xx are dual QPI with no known unlocked CPU. And since T5500/5520 chipset boards don't seem to like single QPI, using a known unlocked such as W3680 is not an option either.

Since the seller seems to dragging their feet giving a shipping confirmation on the X5690, I may just drop the X5687 in there later today.

Anyone know if (or how many) Microsoft limits number of hardware changes in Windows 10?

 

[ManGupta]

This is true. Was referring to my particular situation using T5500 board in T3500 case. No need for the extra power as configured and active cooling on northbridge does not become a major factor unless utilizing the CPU riser.

yeah that's correct

What I gather from my experiments thus far. If it does turn out to be the memory controller on X56xx that means no overclocking with RDIMMs since, unless I'm mistaken, all X56xx are dual QPI with no known unlocked CPU. And since T5500/5520 chipset boards don't seem to like single QPI, using a known unlocked such as W3680 is not an option either.

unfortunately ............ I am afraid ....... this seems to be true ......

Since the seller seems to dragging their feet giving a shipping confirmation on the X5690, I may just drop the X5687 in there later today.

If you do not mind ........ what price was X5690 Deal ( I do not understand why ...... but I am getting quotes of Used X5690 at twice that of X5680 ...... in spite of difference of only 0.13 MHz)

Anyone know if (or how many) Microsoft limits number of hardware changes in Windows 10?

No Idea about that ...... just heard that in case of Microsoft OEM licenses there may be issue while changing Motherboards ..... But did not knew about Processor. I do have changed processors few times but so far did not encountered any issued related to windows Activation.

I have now ordered just one 8 GB stick of Registered ECC DIMM ......... just for a reality check ....... will update the forum once shipping arrives.

@ManGupta I'm digging into my parts pile and getting the 0P041K heatsink mounted onto a T5500MB. The key to swapping heatsinks on these is the threaded spacers in the Thermalright mounting kits. They're the right thread and the right height ( or can be adjusted if needed). I'm trying to duplicate the spring loaded OEM mounting they came with.

Seems interesting mods ..... what is the height , breadth and depth of this HS ?

 

[Retrorockit]

Here are the 0P041K photos. The item as it comes with springs and step bolts, and mounted on an early active cooled T5500. The only parts needed were the Thermalright Knurled female spacers, and some Dell HDD,CD/DVD step bolts with nylon washers to make them tight. If you want to you could go get some longer bolts, and reuse the original springs and cups for an OEM spring type mounting. To tighten the mounting you could shorten the spacer by grinding/filing. Or just drill into the back hole so it sits lower on the MB studs. The spacers are really nice because they have a small shoulder on top that indexes into the bracket on the heatsink so there's no slop in the positioning of it. This should work the same for the 2nd CPU riser.
I try to post thumbnails but the whole image always seems to load for me?
This should help with ManGuptas's 100*F. ambient situation, or help prep for a dual CPU overclocking assault.
MB to top fin installed 105mm, OAH from MB installed 115mm.

 

[Susquehannock]

This should help with ManGuptas's 100*F. ambient situation, or help prep for an dual CPU overclocking assault.

Nice mod. Was thinking about ManGuptas's heat situation myself. I realize ambient room temps in India are much higher than mine (30*C vs 22*C/72*F), but why 100*C so quick? Bad thermal paste? Did you remove cooler and re-apply?

A couple of these boxes I got with W3565 would shoot right up to 100*C and thermal throttle exactly as described with only light loads. Turned out the thermal paste was as shipped from Dell - now old and dried out. Not even making good contact anymore. A quick clean and apply of standard white paste dropped full load temps to 70*C levels. Just a thought.

And like you said Retrorockit, lapping the heatsink smooth does help a lot. Even better if you can use a premium paste. Grizzly Kryonaut works well for me. Although expensive. Tried all sorts of fans in both push and pull configurations. Up to 120x38mm Nidecs. Did not seen to make a significant difference over the good old 80mm. Hottest I can make it now with configuration in my signature is 58*C with Prime95 small FFTs. Often stays below 50*C in gaming and the like.

 

[Retrorockit]

100* "F" ambient. He's cooling with 38*C. air.

 

[Susquehannock]

100* "F" ambient.

100*F ambient? Must have missed that somewhere. Was reading 30*C (86*F) ambient in post # 366.

 

[Retrorockit]

Maybe I did the conversion wrong. He said he didn't have AC, and here in South Florida that means 98*F, ambient. But I tend to go overboard on cooling anyway. If I get an exptreme OC going I don't have to go back and start over. I probably didn't pay more than $15 for that cooler or I wouldn't have 2 of them.
He did ask me to continue with my suggestions and that's where I ended up.