Throttlestop overclocking Desktop PCs

[Retrorockit]

I found another victim for a TS overclock. I rescued an XPS410 from a scrap pile at a computer shop that was downsizing. It's the same as the E520 in my sig. but a Mid Tower version. The capcitors look perfect. I have much better cooling otions than I had when I went 4GHz with a QX6800. Including a BTX Peltier setup. Maybe I can push one back up the charts at CPUZ again. 65nm CPU 266fsb and BTX, nothing but the best. It was free, I gave him $10 anyway.
My comment on VRM cooling was related to the Dells most of us are palying with here. They have no VRM cooling at all (but good airflow there) and if you add some hetasinks they gain quite a bit. But a high end aftermarket boards will already have cooling there.
RAM settings don't come into play much with multiplier overclocking, they just run the same as before. Many feel you get better benchmarks at lower clock speeds with BCLK overclocking, but the speed is harder to come by. I started a separate thread for it because it gets too political,and the info. for this method gets lost in the other discussions.
On an unlocked 2 CPU board i'd like to see what a pair of X5687 32nm quads could do.

 

[BouziGolouM]

@Retrorockit I know aftermarket board have better vrm cooling and like you said better vrm cooling=less voltage needed, was just wondering how far this can be applied to a dell board or hp board for an example. Taking my t3500, wondering if I could squeeze +1 or +2 on the multi, just for fun...
Also, as not familiar with OCing with FSB, would I have any gains in OCing through the fsb over just using the multiplier?
Have not tried through fsb yet, will, but wanted to heard maybe someone's experience on this.
Also for ram, the only thing i've heard is that at high oc, it was better just populating 3 out of the 6 dimms can't say.
I might try this with an old tower I got, but does anyone had chance to test it on AMD cpu yet?

@Retrorockit you mentionned you'd like to see what a pair of quad xeon would do, In a week or two I should have 2 6 cores running on an unlocked 2 cpu board.

 

[Aaron_Henderson]

6 cores, between 4.5 and 4,7ghz, i get almost 1000 in cinebench, IPC wise, it's about the same as a newer i3 or first gen ryzen, and multicore perf is about the same as a stock 6700. I'm using an Asus Sabertooth x58 and just got a wonderful evga sr-2 that I'm gonna play with this weekend.
I've been able to boot at 4.8 but it wasnt stable enough to bench it, raised the VCORE to almost max (without using the Overvoltage jumper) and didn't had luck, I know I could get a lil bit higher by playing with the other voltages, but not there yet. I also wonder if I can reach an higher OC if I boot with only 3 sticks instead of 6 and I wonder if max OC is hinder by ECC ram.
Also wonder if I better cooling on the VRMs would help in reaching an higher clock, just don't have any idea how to as it already have a ceramic heatsink on it.
voltages are now sitting between 1.264 and 1.448 (oc'ed with an offset), didn't reached higher clocks with a set vcore. Still happy with that though.

I wonder how well I can oc my sr-2 though, only have 1 dual qpi link cpu at the moment, so wouldn't be able to OC dual cpu yet but ill comeback with the result and pretty sure I can hit something near the 2000 mark on cine with that board.

@Retrorockit found the mod you're talking about, yeah, that's a nice one, found a way back to put an evo 212 myself a couple years ago, but the way i did it wasn't as neat as you nor the page 8 mod.

Dang...lucky dog you getting a hold of the SR-2...I'm not jealous lol

 

[Retrorockit]

I've had excellent results adding cooling to Dell boards. I would advise adding small heatsinks to each VRM MOSFET. A big metal heat sink will expand and contract and unless you can make a spring mount and flexible interface like the factories do, or they may not stay put. On the T3500/5500 they willl hit components betweem the MOSFETs anyway. The ones I like best are the copper pin fin sinks from Enzotech. Alphacool may offer the same thing. But 10 in a packege isn't enough to do a high powered MB. They can be bent like wires to clear things. I don't trust the ones with tiny fins close together. Air is sticky and forms a boundary layer that doesn't move in small spaces. If you have a fan blowing right onto them they'r probably OK and better than nothing. I have been known to buy coppper chipset coolers and cut them down to get what I need. But copper is much harder to work with than aluminum. The copper sinks are heavy and will fall off over time if you use the 2 sided tape they come with. There is a permanent 2 part thermal epoxy from Arctic. Some people mix it with TIM to make it removable. Some techs rely on Superglue due to it's very thin interface. I haven't tried that myself. It's probably next on my list. The actual result I got was more Voltage available allowing a higher overclock. The lower Voltage required was from moving the Delta fan off of the MB header.
BCLK vs multiplier overclocking is not an either/or question as many BCLK people seem to believe. I would do a multiplier overclock then drop a multipler and start bumping the fsb up. The worst case scenario for this was the 65nm C2X chips which would only move 1 whole 266, or 333 fsb multipler at a time. I would go to Overclock.net X58 Xeon Club for fsb overclocking. Call it BCLK there or you'll get flamed, (fsb is sooo LGA775 you know). I have nothing against the technique, but I don't want this thread to end up with 1,000 pages of RAM timings and QPI Voltages etc. so that TS info. can't be found. Also I know absolutely nothing about it so I can't help you anyway. I've only overclocked locked BIOS Dells which is what led to this forum. Several of us here post in both forums.
If you post that you got 4.7GHz on a multiplier overclock all hell will break loose over there. I hope to see it! But those guys really do know their stuff. My guess is you will lose some clock speed but gain performance in some of the benchmarks. That's the theory anyway.

 

[DarthBubba]

There is a permanent 2 part thermal epoxy from Arctic. Some people mix it with TIM to make it removable. Some techs rely on Superglue due to it's very thin interface.

Wakefield makes a thermal epoxy with a 7.93 thermal conductivity that may be better than Arctic's - I don't know for sure because Arctic is not big on telling their thermal conductivity rating. Wakefield's IS electrically conductive (where Arctic's is not supposed to be) so care is required.

Anyway, Retro, can you please tell us where you attached the fan wires on the Dynatron G17?

It just disappears behind the back chassis wall.

Thanks,

-DB

 

[Retrorockit]

I didn't connect it because I may use another fan. But with the blue thermistor that's visible sticking out of the motor it doesn't matter on that fan. Just any 12V. source you like. It's not a PWM fan. It responds to the temperature of the air coming off of the cooler. The Molex 4 pin sticking up in the air would be as good a place as any. If I use a PWM fan I just power it the same way and use a splitter off fo the original CPU/ PWM wire for a signal. I've run 2 motor GFB series 3.4A. fan s that way before. But I definitely wouldn't run the 1.6A. fan off of the MB header that had a .9A. fan
It's called Thermal Speed Control.

 

[DarthBubba]

I didn't connect it because I may use another fan. But with the blue thermistor that's visible sticking out of the motor it doesn't matter on that fan. Just any 12V. source you like. It's not a PWM fan. It responds to the temperature of the air coming off of the cooler. The Molex 4 pin sticking up in the air would be as good a place as any. If I use a PWM fan I just power it the same way and use a splitter off fo the original CPU/ PWM wire for a signal. I've run 2 motor GFB series 3.4A. fan s that way before. But I definitely wouldn't run the 1.6A. fan off of the MB header that had a .9A. fan
It's called Thermal Speed Control.

What was the temperature drop with the G17 cooler in your T3500, and with which fan (if any) pulling air through the fins?

Also, from reading your posts you seem to favor high amperage fans. Am I misreading or is there an advantage to high amperage fans?

On both of my T3500s simply placing a junk-box-special 80mm fan up against the "exhaust" side of the standard CPU heatsink dropped the CPU temp from 80+ degC to upper 60s degC while running Prime95. Dead stock the T3500 CPU (W3530 - no OC) ramped up to 83 or 84degC during Prime95, but no further. The CPU didn't seem to be throttling per CPU-Z.

 

[Retrorockit]

I haven't run my T3500 yet. I'm busy with 1500W. E Bike projects right now. The high amperage fan thing is because I got into Throttlestop overclocking with Dell BTX computers. They came with 1.3A. to 1.6A Delta fans. The 3.4A 2 motor GFB 120x50mm was what fit in the BTX fan housing. The 1.8A. 150x50mm replaced the whole fan and housing and ran off of the MB header so it was a no brainer. 275cfm and quiet too. I have this stuff sitting around so when I neeed a fan I grab one. Mangupta is trying to cool his sytem with 100*F. air. So he needs lot's of it.1.3A. should help. These are all industrial fans that Dell has used for years. The aftermarket stuff looks like toys to me. The thing is they spend almost all of their time at 30-40% speed. Did I mention they're cheap?
Dell seems to wait until the CPU gets pretty hot before the fan speeds up much. Did you try HWinfo64 or some other fan control utility to get the RPMs up sooner? The Dell fan is powerful enough. If you're not complaining about the niose you haven't heard it at full speed yet.

 

[DarthBubba]

I haven't run my T3500 yet. I'm busy with 1500W. E Bike projects right now.

You're wasting time on BICYCLES when there's so much more important stuff to waste your time on?!?!

<much deleted>

Dell seems to wait until the CPU gets pretty hot before the fan speeds up much. Did you try HWinfo64 or some other fan control utility to get the RPMs up sooner? The Dell fan is powerful enough. If you're not complaining about the niose you haven't heard it at full speed yet.

I've had the fans ramp up, but not to the noise levels you and others have reported. Maybe when the summer arrives and the ambient temp goes up.

BTW, can you tell me the model number of the fan in the earlier picture - the small numbers under the "TA350DC"? Or at least tell me the size of the fin stack on the exhaust side? I've ordered a Dynatron G17 and no one in the order chain seems to be able to tell me the dimensions.

Also, any chance that you could grab your multimeter and find out the cold temperature resistance of the thermistor on that fan? I may have to jerry-rig my own temp-control fan.

Thanks,

-DB

 

[Retrorockit]

The Dynatron provides a 92x25mm fan. The heatsink is 93mmx70mm. M35105-57 is the fan in the photos.
Here's the data sheet for the various Nidec Widebody fans.
https://datasheet.octopart.com/M34789-33-Nidec-datasheet-643872.pdf
Thermistor=4.04 Ohms about 72*F.
This was an OEM Dell fan in the older pre BTX Dimensions with the higher powered Pentium 4/D CPUs. It was shrouded to the CPU heatpipe cooler and pulled air out of the sytem.

 

[SoNic67]

I am getting those errors:

MSVCR120.dll was not found
mfc12u.dll was not found

I have the C++ redistributable 2008 (x64), 2010 (x64), 2013 (x64), 2017 (x64).
What am I missing?

 

[ManGupta]

I am getting those errors:

MSVCR120.dll was not found
mfc12u.dll was not found

I have the C++ redistributable 2008 (x64), 2010 (x64), 2013 (x64), 2017 (x64).
What am I missing?

Not very sure, but I think Windows Updates may fix that.

 

[SoNic67]

Just checked... nothing there, just Antivirus definitions.

LE: Go the 32 bit versions of C++ redistrib too, from here:
https://www.itechtics.com/microsoft...ownload-links/#Visual_C_Direct_Download_Links
That fixed it.

 

[DarthBubba]

The Dynatron provides a 92x25mm fan. The heatsink is 93mmx70mm. M35105-57 is the fan in the photos.

It's a lot easier to find the fan with that number - looking for Nidec TA350DC on Ebay returns too many hits, most of which are not thermally controlled. I've ordered two, one for each T3500 I have here.

Thermistor=4.04 Ohms about 72*F.

As it turns out the thermistor may be used differently than I thought. Peeling back the label on one of the fans and peeking at the additional circuitry should be enlightening.

Thanks for the reply and info.

 

[Retrorockit]

The thermistor could be remotely cabled to be used on the air oulet to power a fan on the inlet side, or elsewhere.

 

[DarthBubba]

The thermistor could be remotely cabled to be used on the air oulet to power a fan on the inlet side, or elsewhere.

We sort of have an equivalent to that with the case front intake fan being controlled by the CPU temp sensor, don't we? Presuming that the CPU cooler's "air duct" formed by the RAM duct and HDD tray are still present.

Anyway, I'm seriously tempted to try and cool the CPU with an extra AIO liquid cooler I have on hand. There's a 1366 backplate for it, but getting the 120mm radiator to exhale to the outside of the case looks problematic without resorting to some irreversible metal cutting.

 

[DarthBubba]

Well, lotsa goodies arrived today via USPS and the Big Brown Truck. The following information is a bit overly detailed to make searching Ebay for the parts easier.

The Dynatron G17 CPU cooler (see message #605) arrived with its own fan and a fan mount already attached to the exhaust side of the fin stack. It appeared to be NIB from the seller. The Nidec Beta V TA350DC Model M35105-57 thermally controlled fan (mine is actually a -58) also in message #605 above has the same frame size (90mm) as the G17's fan (except for the 38mm thickness of the blade tunnel), but with slightly smaller mounting holes in the corners. A bit of careful drill work should open up the mounting holes such that the M35105-58 can snap on to the G17 directly. The M35105-58 draws 1.8 AMPS @ 12VDC so powering it from a Molex connector is warranted.

The other thermally controlled fan that arrived is a Nidec Beta V TA300DC Model M33406-16G6, which is 80mm square and 25mm thick. It fits the exhaust side of the stock T3500 CPU cooler pretty exactly. I'll need to jerry-rig a way to attach it to the cooler with a bit of vibration isolation between the cooler and the fan as it "sings" when held against the cooler's fins directly. In informal testing the W3530 CPU in my T3500 stayed in the mid-to-upper 60*C range during a Prime95 run. The fan audibly slowed as the CPU temp dropped from 82*C. I tried running it from 5VDC and 7VDC but it wouldn't spin up.

 

[SoNic67]

It fits the exhaust side of the stock T3500 CPU cooler pretty exactly.

Do you OC the T3500? Why the fan?

 

[lexluthermiester]

Do you OC the T3500? Why the fan?

ThrottleStop can be used to set CPU's to their maximum performance in non-overclocking systems. When doing this, it's important to keep the CPU cool and mounting a fan to the stock heatsink is an easy way to do so.

 

[SoNic67]

That makes sense now. Thanks.

 

[Retrorockit]

The T3500 with one of the unlocked Xeons is a perfect candidate for Throttlestop overclocking. But you get the automatic overclock settings that Intel allows. So far most people are satisfied with a smalll fan on the stock cooler. It's unknown at this point if added cooling will allow a higher overclock. 4.1 to 4.3Ghz is typical so far. Unfortunately none of the 2 CPU Xeons are unlocked (except some LGA771).

 

[DarthBubba]

Do you OC the T3500?

Not yet I'm waiting for a reasonably priced W3680 to present itself.

Why the fan?

Silicon devices last longer when kept cool; however, my vacuum tube stereo equipment runs better when hot (within limits).

But seriously, I'm not gonna let the CPU stay near its TDP for extended periods of time. There are derating tables for solid state electronics based on decades of unexpected failures that provide guidance here. Not that anyone runs Prime95 for 24 hours straight, but there are some games and video processing tasks (in my case) that can stress the system for prolonged periods. I don't want my video rendering times to lengthen because the CPU is protecting itself via throttling.

ThrottleStop can be used to set CPU's to their maximum performance in non-overclocking systems. When doing this, it's important to keep the CPU cool and mounting a fan to the stock heatsink is an easy way to do so.

What he said .

 

[lexluthermiester]

The T3500 with one of the unlocked Xeons is a perfect candidate for Throttlestop overclocking.

This too, but there is only one 6-core(edit) unlocked Xeon(W3690) and one unlocked i7(990X) compatible with the T3500/T5500/T7500 series of workstations. If you can get one, they're bad ass!

 

[Retrorockit]

There are 4 unlocked Xeons and each one has a somewhat equivalent i7 series. W3570,W3580 are dirt cheap ($20) 4C/8T 45nm Nehalem Xeons or i7 965 Extreme and i7 975 Extreme, and W3680,W3690 are the 32nm 6C/!2T Xeons and are about $80. The i7 980X,i7 990X are their equivalents. The Xeons usually support faster memory speeds. 1333 vs. 1066. But with the memory controller on the CPU support can vary somewhat.

 

[lexluthermiester]

There are 4 unlocked Xeons and each one has a somewhat equivalent i7 series. W3570,W3580 are dirt cheap ($20) 4C/8T 45nm Nehalem Xeons or i7 965 Extreme and i7 975 Extreme, and W3680,W3690 are the 32nm 6C/!2T Xeons and are about $80. The i7 980X,i7 990X are their equivalents. The Xeons usually support faster memory speeds. 1333 vs. 1066. But with the memory controller on the CPU support can vary somewhat.

Sorry, I was referring exclusively to the 6core versions(corrected the post). Most people generally prefer them over the quads. May have missed it, were we able to confirm the W3680/i7-980X were unlocked?
EDIT; Found the specs on the 980X to confirm unlocked status.