Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

HaypurTiryading

macrumors member
Original poster
May 20, 2018
72
29
Turkey
s anyone know clockgenerator (pll) chip/chips of cMP? It should be on cpu tray. I can't access my 4,1 right now but hi-res photos is ok for investigating. I'm looking for further software overclocking possibilities.

Also, looking for ram overclocking possibilities and tightening timing options in a software/uefi or OpenCore like boot manager.

I'll edit the first post after photos come.

Thanks
Muhammet
 
LMAO. I admire your turnip-wringing tenacity! The only thing I can think of is a ZDNet overclocking utility that only worked in Windows on early Mac Pros. But if you boot into MacOS it retained the CPU overclock. I don't know if there's one for RAM.

Let us know how you get on!
 
  • Like
Reactions: HaypurTiryading
LMAO. I admire your turnip-wringing tenacity! The only thing I can think of is a ZDNet overclocking utility that only worked in Windows on early Mac Pros. But if you boot into MacOS it retained the CPU overclock. I don't know if there's one for RAM.

Let us know how you get on!

My aim is detecting PLL (Clock generator) chip and after that editing the bus frequency hex via software like an RW-Everything, SetFSB, SetPLL, Clockgen etc.

I think mine X5680 is 133 x 25 = 3.33GHz. I saw PLL's have hardware or software test modes. We can enter test mode via physical jumpers (resistors, check the first and last link) or virtual jumpers (00 01 like hex PLL register, check the first link). In test modes, PLL's have a predefined BCLK or bus speed like a 133/166/200/233/266/299 etc. If I can push the PLL 133 to 166, mine X5680 run at 166 x 25 = 4.15GHz. :) Also, if PLL supports precise clock adjustment bingo!

#edit: I forgot the ram overclock. After 133 -> 166 adjustment, rams probably will work at 1666Mhz if they handle. There are some programs about ram but they are very risky to play.

ZDNet Utility changes the early Mac Pros CPUs front-side bus speed. The method is the same. Programs are different. Also, ZDNet Utility is PLL specific program. So, it won't work with the newer Macs.

If someone interesting in the topic, I will suggest RW-Everything utility in Windows. It's a very handy and advanced program that can read lots of IO/PLL etc. chip and edit values in realtime. I heard there are dos/terminal variants in Mac OS but I didn't try yet. Also, changes in efi/bootrom manager possible after soft overclock is success. I think.

If someone posts cpu trays hi-res images, I will check the PLL's datasheet.

These websites are my starting points. Some of them are written on a Deutch or Russian. Google Translate is your friend. :)

Deutch



Russian

General oc guide for first gen i7-Xeon
 
Last edited:
so were you able to actually do this? I'm interested in doing this in my Mac Pro 5,1

I'm sorry. I could not be successful.

I took PLL chips some pictures few months ago. If I can find, I will share. There was a option for different fsb's when I am trying but everytime time Mac freezes. Maybe we can not reach PLL through Windows.
 
Does anyone interested in topic? A six core processor overclocked to 4.5 GHz can have almost six core, 3.0 GHz, two processor power in cinebench. I looked datasheet. We can't do software overclock or change IC's registry in Windows. I could'nt read the ic data. Somehow, it was corrupted. I didn't try in macos. There is no software for overclocking or reading ic registry. I don't know.

Maybe in overclock in OC/efi mode successable. Further investigation is needed. Hard programming for ic is also possible. Check the datasheet base clock input table.

Mac Pro 2009 - 4.1's IC chip is Silego SLG84401T

Datasheet: https://www.datasheets360.com/pdf/8402918097453877061

20200126_123142_HDR-02.jpeg
 
I find the topic in general very interesting so don’t give up due to apparent lack of interest.

I’m running 2009 xserve (mac pro 4,1 near clone) not sure if it has the same pll chip on it though. I still have the E5520 installed but have a pair of spare L5640s to use to play with. (I’ve got a single proc xserve3,1 and cMP 3,1 and a super micro X8dt6 dual proc X5680 hack).
 
Does anyone interested in topic? A six core processor overclocked to 4.5 GHz can have almost six core, 3.0 GHz, two processor power in cinebench. I looked datasheet. We can't do software overclock or change IC's registry in Windows. I could'nt read the ic data. Somehow, it was corrupted. I didn't try in macos. There is no software for overclocking or reading ic registry. I don't know.

Maybe in overclock in OC/efi mode successable. Further investigation is needed. Hard programming for ic is also possible. Check the datasheet base clock input table.

Mac Pro 2009 - 4.1's IC chip is Silego SLG84401T

Datasheet: https://www.datasheets360.com/pdf/8402918097453877061

View attachment 919017
I'm really interested in this but I don't really know anything about overclocking at this level in general 😅
 
I find the topic in general very interesting so don’t give up due to apparent lack of interest.

I’m running 2009 xserve (mac pro 4,1 near clone) not sure if it has the same pll chip on it though. I still have the E5520 installed but have a pair of spare L5640s to use to play with. (I’ve got a single proc xserve3,1 and cMP 3,1 and a super micro X8dt6 dual proc X5680 hack).
I'm really interested in this but I don't really know anything about overclocking at this level in general 😅

I couldn't reach ic/pll chip from windows. I use rw everything for chip's registry reading/writing and all data was corrupted. Everytime cMP freeze. We need efi based readings or pin mod for ic/pll chip, i think. I'm not expert at efi or pin mod. If community help us, everbody will have faster computer. :)
 
  • Like
Reactions: gesù21092002
For info, ThrottleStop (Windows) can OC the Xeon on 5,1.

And the OC seems can stay after warm reboot (similar to the Thunderbolt card).

However, it seems something else also limiting the CPU's performance (the primary suspect is the VRM, which can't deliver enough power to the CPU). Therefore, even we can clearly see the performance improvement (e.g. in CineBench), but is isn't that much (if compare to the clock increase ratio).

But what I haven't see yet is to disable some cores, then overclock the CPU (e.g. to simulate X5698).

Even though the multi thread performance will be degraded. But for something that only need single thread performance, we may still able to benefit from OC.
 
For info, ThrottleStop (Windows) can OC the Xeon on 5,1.

And the OC seems can stay after warm reboot (similar to the Thunderbolt card).

However, it seems something else also limiting the CPU's performance (the primary suspect is the VRM, which can't deliver enough power to the CPU). Therefore, even we can clearly see the performance improvement (e.g. in CineBench), but is isn't that much (if compare to the clock increase ratio).

But what I haven't see yet is to disable some cores, then overclock the CPU (e.g. to simulate X5698).

Even though the multi thread performance will be degraded. But for something that only need single thread performance, we may still able to benefit from OC.
so let's say that a Mac with dual xeon x5690 3,47 GHz scores about 6700 points in Geekbench, and I hypothetically overclock both at 3,9 GHz what kind of score increase can I expect? (sorry if the question is dumb)
 
For info, ThrottleStop (Windows) can OC the Xeon on 5,1.

And the OC seems can stay after warm reboot (similar to the Thunderbolt card).

However, it seems something else also limiting the CPU's performance (the primary suspect is the VRM, which can't deliver enough power to the CPU). Therefore, even we can clearly see the performance improvement (e.g. in CineBench), but is isn't that much (if compare to the clock increase ratio).

But what I haven't see yet is to disable some cores, then overclock the CPU (e.g. to simulate X5698).

Even though the multi thread performance will be degraded. But for something that only need single thread performance, we may still able to benefit from OC.

I tried different ThrottleStop versions. Overclocking options are blocked. Also, I tried different softwares like SetFSB, Clockgen, CPU-Tweaker, rw-everything, Intel XTU, Tpu ThrottleStop. Programs see the system but overclocking options are blocked/missing or ic/pll registry data corrupted.

Capture.PNG

Capture2.PNG

[automerge]1590581514[/automerge]
so let's say that a Mac with dual xeon x5690 3,47 GHz scores about 6700 points in Geekbench, and I hypothetically overclock both at 3,9 GHz what kind of score increase can I expect? (sorry if the question is dumb)

Windows scores, you can check youtube videos or forums.

Single 4.5 GHz X5690 = ~1000-1100 Cinebench r15 point.
(Stock was, ~800. Dual stock ~1600)

2 x X5650-60 = 1200-1300 Cinebench r15 point.

By the way, the stock voltage allows overclocking up to 4-4.2 GHz on a pc side.
 
Last edited:
  • Like
Reactions: gesù21092002
I tried different ThrottleStop versions. Overclocking options are blocked. Also, I tried different softwares like SetFSB, Clockgen, CPU-Tweaker, rw-everything, Intel XTU, Tpu ThrottleStop. Programs see the system but overclocking options are blocked/missing or ic/pll registry data corrupted.

View attachment 919110
View attachment 919112
[automerge]1590581514[/automerge]


Windows scores, you can check youtube videos or forums.

Single 4.5 GHz X5690 = ~1000-1100 Cinebench r15 point.
(Stock was, ~800. Dual stock ~1600)

2 x X5650-60 = 1200-1300 Cinebench r15 point.

By the way, the stock voltage allows overclocking up to 4-4.2 GHz on a pc side.
I can't remember the exact details now. The last time I did that may be 2 years ago. But I am sure both ThrottleStop and Intel XTU can mod the multiplier (not just from me, but also tested by other users).
[automerge]1590592117[/automerge]
Just found one of the record. This was captured when I OC my W3690 to 4GHz in XTU.
4GHz.PNG


Anyway, I think the X5690 is locked by default. May be that's why you can't OC it on cMP.
 
Last edited:
This is from another cMP user. He OCed his W3690 to 4.12GHz and 4.25GHz.
39443054_1827611780608840_2973204077162266624_o.jpg

The benchmark result also suggested that OC was working. The cMP clearly perform better than the stock settings.
 
  • Like
Reactions: theMotoMan
This is from another cMP user. He OCed his W3690 to 4.12GHz and 4.25GHz.
View attachment 919162
The benchmark result also suggested that OC was working. The cMP clearly perform better than the stock settings.
This is from another cMP user. He OCed his W3690 to 4.12GHz and 4.25GHz.
View attachment 919162
The benchmark result also suggested that OC was working. The cMP clearly perform better than the stock settings.

Our cpu's work +1 standart multiplier, means X/W5690 works 3.58GHz under load, all core. Not 3.46GHz.
3.58 -> 4.25 = %15 overclock gain. Not impressive for top tier cpus but below X5680/W3680 every gain is very acceptable. We should not underestimate %15. Nearly all of us using very good video cards for our older cpus. Gain = higher fps = lower bottleneck.

Yes, W series xeons have unlocked multiplier and increasing will work out of the box but X series can only overclock by bclk/fsb. That's why I am looking ic/pll overclocking options. Bclk overclock can be achievable by os accessed re-programming, efi accessed re-programming or hard wire mod. (133 bclk to 166bclk..., see the datasheet for slg84401t)

As I said before, bclk overclock is not possible thru os/efi because ic chip is write protected somehow or it's data flow is one way. We can't read the ic data but datasheet shows manual bclk via ic pin mod is probable.
 
Yeah, I have the OC table for the 5600 series Xeon, they always work at 3.6 or 3.73GHz as long as there is no thermal restriction.
X5600 turbo table.jpg


And if we OC the CPU to 4.25GHz, there should be 18% improvement (4.25-3.6)/3.6 x 100% = ~18%

But as you can see, there is something bottlenecking the CPU (e.g. power), and if we OC the CPU to 4.25GHz, the improvement is just (866-781)/781 x 100% = ~10%

Since 10% is way outside normal error margin already. And in fact, we can further push the CPU clock speed to achieve even higher result.
39454306_1829177700452248_4194734287003582464_o.jpg

Therefore, I will say that's a good prove that we can actually overclock the CPU on cMP. However, as you said, for X5690, that will be much harder to achieve.

Anyway, the performance improvement may not be that much if compare to the clock speed increment. But I bet this very depends on the actual workflow. Some compute task can benefit more than the others. Gaming may do better.
 
If I had a spare dual cpu tray for my 4.1, I would have already tried hard pin mode for ic chip. We also can lower the ram speed or loosen the ram timings via Taiphoon Burner. bclk overclock also affects the ram. 133 bclk to 166 bclk looks easy according to datasheet. Just one pin triggers to increasing the bclk but I don't know how to trigger that input. Maybe it triggers via some 3.3-5v volt, resistor, gnd, or shunting somewhere?

Look at that, Linus overclocks TI-84 calculator. :)

We need a expert technicians that show us how to mod or guide our machines.
 
  • Like
Reactions: h9826790
I found this thread not to be a mac user very interesting, I am trying to upload my ram memories to 1600mhz in a hp 2072la laptop, but selecting clok generator in rw evrything gives me sm bus error, how can I solve it?


Sin-t-tulo.png


in h20 efi gave the same error but to solve it by altering the following data in an INI in rw I don't see something similar

Sin-t-tulo.png


[MULTI_FD]
Flag=1 << Change this to 0, save and try again

Also, when you do above edit, also make this change >>
; Supports on WIN flash.
[Option]
Flag=0 <<< Change to 1
;Flag (wW)
; default : 0.
; 0 : Auto flash mode.
; 1 : User option mode, including option, start, exit buttons.
; (Option button will disable on secure flash mode.)
; 2 : User flash mode, including start, exit buttons.
 
I found this thread not to be a mac user very interesting, I am trying to upload my ram memories to 1600mhz in a hp 2072la laptop, but selecting clok generator in rw evrything gives me sm bus error, how can I solve it?


Sin-t-tulo.png


in h20 efi gave the same error but to solve it by altering the following data in an INI in rw I don't see something similar

Sin-t-tulo.png


[MULTI_FD]
Flag=1 << Change this to 0, save and try again

Also, when you do above edit, also make this change >>
; Supports on WIN flash.
[Option]
Flag=0 <<< Change to 1
;Flag (wW)
; default : 0.
; 0 : Auto flash mode.
; 1 : User option mode, including option, start, exit buttons.
; (Option button will disable on secure flash mode.)
; 2 : User flash mode, including start, exit buttons.

Hello. Here is not hp or software overclocking forum. This is the section dedicated to Mac Pro only. Although I am not a competent person so I can't answer your question. I believe that you can find the answer to your question in different forums like bios-mods etc.
 
Yup, single/dual X series are locked and overclock couldn't be possible if someone can't find a IME Exploit or something. Our cMP's PLL chip locked in BCLK overclock mod. I saw some exploit for X99 Xeons but I don't know it is suitable for X58.
 
Yup, single/dual X series are locked and overclock couldn't be possible if someone can't find a IME Exploit or something. Our cMP's PLL chip locked in BCLK overclock mod. I saw some exploit for X99 Xeons but I don't know it is suitable for X58.
I’m thinking about getting w3690 cpu, how could I overclock it?
 
Ah, I've finally found my people. I have no experience with overclocking but similarly am trying to get every last ounce of power out of my 2009 4,1–>5,1 with 5690s, whether that's through CPU or RAM overclocking. I've chipped away at all the other optimizations, from bifurcated x16 pcie nvme ssd and gpu upgrades to ram slot optimizations. Been even looking at liquid cooling for fun. Anyone have any more luck on this topic?
 
1635742617075.png


so simply by looking at the datasheet.. in order to change FSB from 133 to 166, we need to change FSB(pin 49) from 0 to 1. To make this happen, it seems like pin 49 need to be isolated from existing signal line and then tie to pin 48 which is FSA (which already tie to 1) on the SLG84401? There are some resistors leading to these pins and perhaps removing the one connecting to pin 49 and solder it to 48 would work.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.