247 bclk on Bloodrage PCI-E hardmodded

[G.Foyle]

Hi I'd like to share another high bclk with recently discovered Bloodrage mod:
Foxconn Bloodrage with PCI-E frequency mod
Xeon W3520 retail batch #3844A704 coldbug at -105°C
Corsair Dominator GT 1866
Kingpin's Dragon F1 EE + LN2
Bios G25
Vcore = 1.45-1.52 V

Max bclk = 5070 MHz - best non-Classified W3520 frequency


Max clock = 247 MHz


Max SuperPi 16M = 3 : 17 . 687 at 241.4 bclk

It was 32M stable as well, and we had screenshots to prove it, but they got lost in a hard drive failure

Thanks to The_Fox for his excellent tips here and here, r1ch and Lord Trident for discovering the mod. Oh Foxconn, why didn't you do it right from the start?

Thanks
-Team PCLab.pl

[Solarfall]

why is your qpi link only at 2897? i wonder is that some sort of a bug on the board/cpu-z since it should be wayy over 4000 at those kinds of speeds

[G.Foyle]

We disabled QPI Fast Mode - it makes QPI multi 12x but hurts 3d performance.

[Solarfall]

We disabled QPI Fast Mode - it makes QPI multi 12x but hurts 3d performance.

ohhh i see, thanks for the explanation

[sofos1990]

Awesome bclk frequency
After this mod for high bclk, the mobo is very competitive

[The-Fox]

Great clocks xoqolatl, glad to see my settings helped you
BTW, have you tried to fiddle with the settings and see if you can get some higher clock ?

We disabled QPI Fast Mode - it makes QPI multi 12x but hurts 3d performance.

It is most probably using different bClk as the i7 Doesn't support x12 multy for QPI.
According to CPU-Z its 2/3 the "Real" bClk but Massman and Saaya allege that its 1/4 of the Real bClk and that CPU-Z is inaccurate, thus the main reason to hurt 3D performance.

Can you test 3D Performance with QPI Slow mode and see if it really hurts 3D benches and by how much ?

Thanks

[G.Foyle]

I did run 3dm03 on slow QPI once, but it was so slow I didn't even bother to save a screenshot.
Your settings were very good, but I had to change:
PCI-E to 114 or 115 MHz
QPI CPU Latency to 80-90
Vpll to 1.86 V
CPU clock amplitude to 0.9 V
I also fiddled with CPU clock skew but I don't know if that helped.

[roller11]

Fox,

with slow mode, my framerate goes down my about 66% so yes, it's a huge factor in 3D.

[The-Fox]

I did run 3dm03 on slow QPI once, but it was so slow I didn't even bother to save a screenshot.
Your settings were very good, but I had to change:
PCI-E to 114 or 115 MHz
QPI CPU Latency to 80-90
Vpll to 1.86 V
CPU clock amplitude to 0.9 V
I also fiddled with CPU clock skew but I don't know if that helped.

Interesting, I actually find that Clock Skew made things worse for me, I played with it for a while, on H20 and cold, but behavior was the same .

The PCI-E didn't seem to matter, 110-120 is all the same I think

Maybe on my next session I would try some of your settings like Vpll, CPU Clock Amplitude and QPI CPU Latency.

Fox,

with slow mode, my framerate goes down my about 66% so yes, it's a huge factor in 3D.

Is that on the EVGA Classified ?
I kind of started to think that the BR is doing the QPI SLow mode differently but it is probably all the same I guess.

[roller11]

Fox,

slow mode tried on both the Classified and the giga x58 UD5, same result.

xoqolatl,
q about your max CPU clock...you show your max is 5070 and that's with a 16M superpi. Normally a chip can do a higher Max CPU clock running, say, CPUZ than superpi. This implies that you have some kind of wall at 5070 Mhz, is this the case?

[G.Foyle]

I had a validation at 5103 (243x21) but the file got corrupted.

[roller11]

So it appears that there is some kind of wall at 5070 even with the hard mod.