Haswell PCI-E coldbug

[K404]

I found that if I took the CPU colder than -120, 3D would stutter like crazy. Scores were in the toilet.

 

I shall name this....cold-stutter

[pro]

What board and bios?

[K404]

Gigabyte OC board, BIOS F5B, PCI-E Gen2 set in BIOS

Edited June 6, 2013 by K404

[Massman]

I've heard similar complaints of "coldslow'-type issues, not never seen an isolated case.

 

Interesting topic!

[crustytheclown]

I've seen something similar with an early i7 940... but it was only one cpu at about the same temps Kenny describes...

I remember raisining a voltage insanely high to get past it but i have to get a bios screen in front of me to remember which one...

[K404]

Oh, it's definitely not isolated I've had PCI-E related problems on Ivy at -140....but this is Haswell

[Xtreme Addict]

Oh, it's definitely not isolated I've had PCI-E related problems on Ivy at -140....but this is Haswell

 

PCIE for sure doesn't like you To have problems on IB you need damm luck

[K404]

You want to know about bad luck on Ivy? I saw 6.4GHz, ONCE, screenshot only. Ivy and me did NOT get on well.

[xxbassplayerxx]

I've seen something similar with an early i7 940... but it was only one cpu at about the same temps Kenny describes...

I remember raisining a voltage insanely high to get past it but i have to get a bios screen in front of me to remember which one...

 

Many X58 boards had a sort of "QPI Degradation" to them, but it was unrelated to temperature. As QPI approached 4GHz (BCLK at 220), the QPI (link between X58 Chipset (which handles PCI-E lanes) and CPU) would stutter really badly and cut 3D frames massively. This is why QPI Slow Mode was useful for 2D but not for 3D.

 

I uploaded this a few years ago while messing with my RIIIE:

 

Edited June 7, 2013 by xxbassplayerxx

[Bobnova]

Capacitor capacitance goes down with temperature, it's possible that the little 0.1µF coupling caps drop enough that the signals don't last long enough to transfer.

If they're on the board for Haswell you could try replacing them and see what happens

[Moose83]

I had similar problems on Ivy with Asroc Z77 Extreme 9. Onces i go under -120 3d was an Diashow:D

[Gamer]

Stick with 10o bclock, team au had similar issues at computex

[K404]

If I stay at 100 strap, My uncore is limited to 3.9GHz.

[jjjc]

If I stay at 100 strap, My uncore is limited to 3.9GHz.

 

B10 bios fixes uncore multiplier on 100 strap. Seems to be working for me, 40x applied no problems at least - just prelim testing.

 

http://www.mediafire.com/download/7v1c3176s876v56/B10.ZIP

[sin0822]

or just use 1.25x if you don't want to use that BIOS that is still 100bclk and 100 pcie

[pro]

Stick with 10o bclock, team au had similar issues at computex

 

it was related to using the LN2 switch actually

 

all is ok now just use new GTL and OC buttons

 

http://gigabytedaily.blogspot.tw/2013/06/z87-windows-xp-downloads-oc-button.html?spref=fb

[K404]

Awesome! Thanks for that link!

 

But....Windows 7 & 8 Z87 OC Buttons driver is in the owners trash (error 404)

[sin0822]

i am pretty sure you can just use the XP link for both? I thought that it how it was linked on facebook?

[Massman]

Fyi, we had a funky cold issue with Z87 last weekend.

 

At -135°C, 01 Nature was ~70FPS; at -107°C, 01 nature was ~ 220FPS. Reboot solved the issue though.

[CL3P20]

Fyi, we had a funky cold issue with Z87 last weekend.

 

At -135°C, 01 Nature was ~70FPS; at -107°C, 01 nature was ~ 220FPS. Reboot solved the issue though.

 

PCI-E link 'cold slow' .. strange.. sometimes I get it.. other times not. Could be tied to input voltage ratio's. I have had both: VGA signal 'drop' to black screen, and VGA/PCI-E cold slow.. but adjusted Input and IO's and have not had it return.. yet