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

BSEL Mod (Xeon Overclocking) Failed - Any Observations?

spetznatz

spetznatz

macrumors regular
Original poster
Jan 5, 2006
235
36
I followed the instructions here:

http://www.o0o.it/pro

To do the BSEL mod. My (previously 2.0 GHz) MP booted up after brain transplant - hooray. Eight-core rendering in Cheetah, no problem.

However the chips seem to be running at 1.86/1066 (According to Marcel Bresink's Hardware Monitor).

So questions -- if I somehow got the procedure wrong (wrong pin covered, say), should the chips be running at all?

If I got the procedure right (correct pin covered) then what are the possible explanations for it not working?

I'd really appreciate any hints, advice, input!!!
 
Can you post a screen-grab of your "System Information" window from Hardware Monitor, like this:

GarageSale_1206022422_555.jpg


It's where your "true" clock and FSB speeds will be shown.
 
Cal -- here you go...as you see, both are reporting 1.86 GHz

The core step of my processors was later than the ones used in the article, would that make a difference?

Also, shouldn't the level 2 cache be 8MB?
 

Attachments

  • Harware Monitor.jpg
    Harware Monitor.jpg
    53.5 KB · Views: 591
spetznatz said:
I followed the instructions here:

http://www.o0o.it/pro

To do the BSEL mod. My (previously 2.0 GHz) MP booted up after brain transplant - hooray. Eight-core rendering in Cheetah, no problem.

However the chips seem to be running at 1.86/1066 (According to Marcel Bresink's Hardware Monitor).

So questions -- if I somehow got the procedure wrong (wrong pin covered, say), should the chips be running at all?

If I got the procedure right (correct pin covered) then what are the possible explanations for it not working?

I'd really appreciate any hints, advice, input!!!
Click to expand...
Can that mod be applied to a 3.0 8-core 2007 mac pro?
 
spetznatz said:
Cal -- here you go...as you see, both are reporting 1.86 GHz

The core step of my processors was later than the ones used in the article, would that make a difference?

Also, shouldn't the level 2 cache be 8MB?
Click to expand...

Yeah. Are you absolutely certain that you covered the correct pin? Was your insulation tape high quality? I mean, could the connector pin have "punched" through and made contact.

I don't know enough about the core step to comment, sorry.

The L2 cache in Hardware Monitor is per core-pair. ASP shows 8 MB per processor.

Frozonecold said:
Can that mod be applied to a 3.0 8-core 2007 mac pro?
Click to expand...

I don't believe so, although I could be wrong. I believe its purpose is to bump the the FSB from 1066 to 1333 MHz, matching the motherboard and up-clocking the processor by x 1.25 into the bargain. Your FSB is already running at 1333 MHz.
 
Not to much to add except recheck your pins.

This is only good for 266MHZ FSB Chips.

Tracer
 
Thanks, Cal, Tracer --

Yes, the only thing that I can think of is that the tape is somehow at fault. We used this brown insulating tape, because it was the only one with the required 'stickiness'. The other two black tapes just wouldn't stay stuck.

If the wrong pin had been covered (and was insulated) then it shouldn't work at all, should it?

Maybe it's semi-conductive insulating tape...:confused::confused::confused:

Is there any other method of blocking the pins? insulating paint, perhaps?
 
spetznatz said:
Thanks, Cal, Tracer --

Yes, the only thing that I can think of is that the tape is somehow at fault. We used this brown insulating tape, because it was the only one with the required 'stickiness'. The other two black tapes just wouldn't stay stuck.

If the wrong pin had been covered (and was insulated) then it shouldn't work at all, should it?

Maybe it's semi-conductive insulating tape...:confused::confused::confused:

Is there any other method of blocking the pins? insulating paint, perhaps?
Click to expand...

Rear defogger kit works great. Just be sure to isolate the pins correctly.

The processor still works even though the BSEL trick did not work properly.

That should explain the reason why it reverts to 1866Mhz.
 
Well, it seems like they're working at stock speeds. Either the pins are still contacting or the over-clock ability has been disabled. Do you have any Intel part numbers for them, maybe starting with Q or SL?
 
Pressure -- "Rear Defogger" -- can you be a bit more specific?

Cal -- will look into it...they came in plain boxes...
 
Pressure said:
Not really, it is just called Rear Window Defogger kit.

It's conductive ink.
Click to expand...

Yes, and I don't thin kit will work in this case: we need to insulate the pin, not to make a connection like it was the case on old Athlons, where conductive ink was used to "unblow" some fuses on the CPU surface.
 
pipomolo42 said:
Yes, and I don't thin kit will work in this case: we need to insulate the pin, not to make a connection like it was the case on old Athlons, where conductive ink was used to "unblow" some fuses on the CPU surface.
Click to expand...

no, you use conductive ink for the BSEL mod...

You can read more about it here.
 
That link was for socket 775 processors, both Mac Pros use socket771 processors.

Tracer
 
Here is the datasheet for the 5300 processor in a 771 socket.

Table 2-1 and 2-2 on page 20 show what BSEL mods are needed. (connect BSEL2 to VTT)

Table 4-2 on pages 74 and 75 identify pin G30 as BSEL2 and E30 as a VTT.

Figure 3-7 on page 55 shows the location of G30 and E30 when viewed from underneath.

Join these with a conductive pen and then insulate G30 (as on oOo.it) you should be golden.

Well, in theory, anyway...

(no guarantees, though) :)
 
Thanks for all the info, folks.

Cal -- The core step for my processors is SLAEL

...now how do you say 'Conductive Ink Pen' in German?
 
Cheers, Duke! (My wife didn't have a clue, and she's a native speaker...)

Cal -- I have a horrible feeling you may be right...

Actually, I think due to the 8MB cache v. 4MB on my previous models, my Pro actually feels extremely responsive...I don't think it's a step backward in speed, anyway. It's just what it will do to the re-sale...
 
Any program that's capable of really using multiple cores intelligently will work better in spite of the apparent drop in clock speed. More and more applications, especially on the Mac side, are going to be doing this. By the time you come to sell, potential buyers should have come to realise that a 1.87 GHz octo beats a 2.0 GHz quad, hands down.

Have you tried the conductive pen? It might just work and it shouldn't do any irreversible harm. My second processor swap only took 10 minutes or so. It's only the first time that seems difficult!
 
Not tried it yet, but got a friend coming over from London on the weekend. He can help me get the procs out again (I can't manage it on my own due to being somewhat 'manually challenged' ;))

Thanks for all you help on this. Much appreciated.
 
I did this same mod 3-4 months ago and mine reacted the same way. But I can assure you that my cores are running @ 2.33GHz even though the monitoring software in OSX only states 1.86GHz. When I boot up into Windows and run CPU-Z it shows 1333MHz FSB and 2.33GHz. For whatever reason OSX just doesn't know how to report it correctly. Unfortunately, the only way that I know of to find out what they're actually running at is to boot up Windows.
 
Something I noticed when upgrading from 4 x 2.66 to 8 x 2.66 was that most of my xbench and my cinebench 9.5 (single CPU) scores were virtually unchanged. Have you any tests that you ran before the upgrade. You could find a relevant section to compare with your performance now?
 
Stangboy -- Interesting. Will check it out later.

Cal -- the fact that I'm getting 52,000 MFlops in PowerFractal vs. the original article's 64,000 is what alerted me first. I'm seeing a 60% increase in rendering speeds in Cheetah3D (the Mac's best-kept secret ;)). I would have expected a bit more...
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back