i7 920 D0

[Pedros]

Originally posted by: Rubycon


You may have pulled the trigger too fast on selling that C0. You could easily get a Dud0! I'd never part with something unless my mind was made up and I did not want it at all. IMHO

Hmmm, let's hope not then

I didn't lost any money changing CPU ... i got a discount on the D0 so... let's see And i still have my C0 so, if i don't like it... i change it back and sell the D0

 

[Pedros]

Originally posted by: Beno
i dont know if my p6t deluxe have a bios optimized for D0s. would like to know if anandtech test that motherboard after reading teh D0 article.
im getting one of these when tank guys have em in stock

Beno,

your p6t will do just fine. Just keep the bios version up to date

 

[Idontcare]

Originally posted by: JAG87

Originally posted by: Rubycon
I've never even seen a core crash in Prime when pushed too far - the machine just halts with a BSOD or kernel panic.

That's very true. I scratched my head at this, only once I've seen a calculation error in Linx during my overclocking session, all other trials have been indisputable BSODs.

Neither of you two have seen Prime95 do this?

Rounding Error Detected

I can generate these thread stop messages all day long by merely lowering my Vcc by 0.05V (from 1.35V to 1.30V) on my Q6600. Never BSODs from it, I have to go another 0.05V lower (to 1.25V) before I can get my rig to BSOD.

 

[ashishmishra]

Originally posted by: Pedros
Hello ashishmishra,

I'm running the chip @ 19x200 BCLK, 1.275vcore, 1.3 qpiv, 1.60v mems. All the remaining is on auto. Well, i tested it with OCCT.
I see lots of people running the same speed at lower voltages with lower temps. So that's why i'm asking this

But, 1.375v is way too much voltage for 3.8Ghz :| What's your batch number?

Actually I think it can do slightly lower but I haven't completely dialed in the vcore yet (I know) I think my batch is 3839A. Also I think if I had load line calibration option for my MSI mother board I am sure it could run at 1.35V if not lower, cause that is where it drops to under load.

 

[Rubycon]

Originally posted by: Idontcare

Neither of you two have seen Prime95 do this?

Rounding Error Detected

I can generate these thread stop messages all day long by merely lowering my Vcc by 0.05V (from 1.35V to 1.30V) on my Q6600. Never BSODs from it, I have to go another 0.05V lower (to 1.25V) before I can get my rig to BSOD.

Yes that is what happens with 65nm and 45nm Core2Quads. i7 OTOH does NOT exhibit this behavior at all. It will just halt the system completely.

 

[Idontcare]

Originally posted by: Rubycon

Originally posted by: Idontcare

Neither of you two have seen Prime95 do this?

Rounding Error Detected

I can generate these thread stop messages all day long by merely lowering my Vcc by 0.05V (from 1.35V to 1.30V) on my Q6600. Never BSODs from it, I have to go another 0.05V lower (to 1.25V) before I can get my rig to BSOD.

Yes that is what happens with 65nm and 45nm Core2Quads. i7 OTOH does NOT exhibit this behavior at all. It will just halt the system completely.

Interesting. It's hard to imagine the silicon in an i7 is suddenly less susceptible to undervolt induced errors given that Yorkfield quads are produced with the same xtors...so something in the error checking must have been improved for more robustness in Nehalem. (i.e. layout and architecture robustness has been improved, not process tech)

Is this something we would expect from the static CMOS implementation or from the 8 xtor sram cells? A steeper cliff between acceptable signal/noise to unacceptable signal/noise?

Where's Dmens, pm, Tuxdave and CTho9305?

 

[chizow]

Originally posted by: Rubycon

Originally posted by: Idontcare

Neither of you two have seen Prime95 do this?

Rounding Error Detected

I can generate these thread stop messages all day long by merely lowering my Vcc by 0.05V (from 1.35V to 1.30V) on my Q6600. Never BSODs from it, I have to go another 0.05V lower (to 1.25V) before I can get my rig to BSOD.

Yes that is what happens with 65nm and 45nm Core2Quads. i7 OTOH does NOT exhibit this behavior at all. It will just halt the system completely.

That's a very interesting observation, is there some white paper detailing and documenting this, or is it something you picked up on from testing? Not saying its incorrect, as the analysis seems spot-on and something I've also noticed come to think of it. i7 lets you know very quickly whether you're stable or not and ANY instability results in a very neat and tidy BSOD citing hardware (rather than kernel, driver or app), often within the first hour of Prime/OCCT/Linpack etc. The nice part about this is it allows for very accurate voltage tweaking, so that even a single incremental increase can lead to (in)stability.

 

[aigomorla]

Originally posted by: Idontcare

Originally posted by: JAG87

Originally posted by: Rubycon
I've never even seen a core crash in Prime when pushed too far - the machine just halts with a BSOD or kernel panic.

That's very true. I scratched my head at this, only once I've seen a calculation error in Linx during my overclocking session, all other trials have been indisputable BSODs.

Neither of you two have seen Prime95 do this?

Rounding Error Detected

I can generate these thread stop messages all day long by merely lowering my Vcc by 0.05V (from 1.35V to 1.30V) on my Q6600. Never BSODs from it, I have to go another 0.05V lower (to 1.25V) before I can get my rig to BSOD.

on an i7 tho idc, it crashes entirely or passes.

You dont get the same errors like a dual die quadcore does

im with Ruby and JAG on this one.

 

[iCyborg]

My experience with i7 also agrees with this: I had about 5-6 prime95 failures, all of them BSODs.

 

[Rubycon]

Note the reason!

http://pics.bbzzdd.com/users/C6FT7/crasht.jpg

Never noticed this until i7. I agree we need some input from one of our resident Intel microarchitecture experts.

 

[Idontcare]

Originally posted by: Rubycon
Note the reason!

http://pics.bbzzdd.com/users/C6FT7/crasht.jpg

Never noticed this until i7. I agree we need some input from one of our resident Intel microarchitecture experts.

Yikes. So clock distribution and skew is coming into play. Definitely a consequence of the static CMOS then.

So its not necessarily a "they decreased sensitivity to rounding errors by making something more robust versus Yorkfield" but more of a "the clock distribution is now the weak(er) link when overclocking and fails earlier than the traditional signal/noise compute failures do".

Rest assured its a feature though, that odor is the smell of progress :thumbsup:.

 

[chizow]

Yep! That's the one. If I had to guess its due to the memory controller being integrated this time around, so if there's any addressing faults the entire CPU shuts it down. But ya a definitive answer from the Intel folks would be great.

 

[Rubycon]

Another observance - throttling seems much more efficient and far less intrusive. Once any core reaches 100°C throttling will drop the multiplier down briefly to keep that temp from going over 100°C. I just watched this for an hour with linpack running and the fan on the TRUE spinning at a snails pace while the cpu was at 191x21 at 1.5 volts. No crashes or errors were observed at all during this test.

 

[aigomorla]

Originally posted by: Rubycon
Another observance - throttling seems much more efficient and far less intrusive. Once any core reaches 100°C throttling will drop the multiplier down briefly to keep that temp from going over 100°C. I just watched this for an hour with linpack running and the fan on the TRUE spinning at a snails pace while the cpu was at 191x21 at 1.5 volts. No crashes or errors were observed at all during this test.

Isnt 100C themal shutdown?

God ruby WTF are you doing to your poor chip.

My chips rarely even see's 70C.

 

[Rubycon]

Stress testing is to stress the system to the limit. 100C is the limit and yes that's where it will throttle. If the temp goes beyond that the motherboard SHOULD power off but not all do and I've yet to reach that point. Next step is to boot up with no heat sink on the cpu.

If I cannot crash it at 100C running a stress program I know it will never crash at 65C with encoding. Well let's hope so.

I have plenty of cpu's in case one blows like a fuse and you know I've never had just that happen before. Of course if this was the only CPU I had around it would've blown at the first attempt to raise the vcore. Mr. Murphy plays that way you know.

 

[aigomorla]

Originally posted by: aigomorla

My chips rarely even see's 70C.

i wasnt kidding about this statement.

HT ON:
http://i125.photobucket.com/al...gomorla/Capture-10.jpg