ground says:

5 year and 8 months ago – Crazy chip!

Alpi says:

5 year and 8 months ago – Jesus holy christ !! :) Insane clocks ! Grats, epic run !

leeghoofd says:

5 year and 8 months ago –

Hell yeah!!!

SparkysAdventure says:

5 year and 8 months ago – Nice

skulstation says:

5 year and 8 months ago –

Wow, what a cpu man.

SparkysAdventure says:

5 year and 8 months ago – Hold up you destroyed the Westmere 6x record. Somebody get this man an SR-2.

TAGG says:

5 year and 8 months ago –

Thanks guys

@skulstation, the one you have should allso do at least 6GHz (the better one)

@AutisticChris I have one and another one of these crazy chips

@AutisticChris

TaPaKaH says:

5 year and 8 months ago – Dat CPU... What can it do on air and/or for Spi on LN2?

ground says:

5 year and 8 months ago –

20 minutes ago, TaPaKaH said:

Dat CPU... What can it do on air and/or for Spi on LN2?

1.3V for R15 at 4.5 GHz I've heard, not the rarest ambient clock (I actually have at least 3 xeons that can do that, but those will be multiplier limited)

TaPaKaH says:

5 year and 8 months ago –

Are those Xeons 3xxx (single QPI) or 5xxx (dual QPI) series?

I've recently dug through many old Gulftown OC threads but couldn't find any examples of single QPI chips (such as i7-9x0 and W36x0) needing less than 1.36-1.38V for full stability and/or 1.28-1.30V for 32M at 4.5GHz (extrapolating from other clockspeeds at 0.04V per 100MHz). I also wasn't able to find any such chips myself on a few dozen A0 / 3005F chips that I tried over the years. So I guess the dual QPI chips are better on MHz/volt due to having to meet stricter binning requirements induced by the second QPI bus.

TAGG says:

5 year and 8 months ago –

1 hour ago, ground1556 said:

1.3V for R15 at 4.5 GHz I've heard, not the rarest ambient clock (I actually have at least 3 xeons that can do that, but those will be multiplier limited)

0,02V less and 0,05GHz more and you're spot on

@TapPaKah They are dual QPI but only run on one with X58, no idea about single core on LN2 only ran R15 but all the other cascade scores are with this chip

I actually only binned 22 X5690s to find 2 that do what i stated above on air, LN2 test of the 2nd one incomming this week

ground says:

5 year and 8 months ago –

4 minutes ago, TaPaKaH said:

Are those Xeons 3xxx (single QPI) or 5xxx (dual QPI) series?

I've recently dug through many old Gulftown OC threads but couldn't find any examples of single QPI chips (such as i7-9x0 and W36x0) needing less than 1.36-1.38V for full stability and/or 1.28-1.30V for 32M at 4.5GHz (extrapolating from other clockspeeds at 0.04V per 100MHz). I also wasn't able to find any such chips myself on a few dozen A0 / 3005F chips that I tried over the years. So I guess the dual QPI chips are better on MHz/volt due to having to meet stricter binning requirements induced by the second QPI bus.

I'll note that this is my binning data (binned about 50 32nm Westmere 5xxx xeons or so) and not Taggs, and these are the best of those, though my main goal was binning for max BCLK (that chip is actually not included here as it clocks worse). There might have been some better chips I missed but I wasn't aiming for max clocks at ambient

(last 2 are Uncore/RAM freq for superpi 4m as I was only testing these quickly. If you want I can do some more detailed testing I can do that later this week on the E5649 and 270 E5640)

I'm binning at 1.4V Core/1.4V QPI at ambient with a weighted down aircooler. All three of these should about manage R15 at around 4.5 with 1.3V.

 

The E5649 is 100% stable at 4490 Core/3814 Uncore/4040 QPI/890 MHz with 6 DIMMs with 1.344V Vcore, 1.3V VTT, 1.65V DRAM

(used as a daily, actually degraded at some point, was better before). I've heard of a couple x5675s being 100% stable at 4.6+ with under 1.4V in daily systems.

TaPaKaH says:

5 year and 8 months ago –

Did you happen to grasp any difference on core clocks between "classic" batches like 3003A-3031A and "new" batches like 31xxC / 32xxC?

I guess no one cared about the latter on 980X / 990X / W36x0. When they came to market in early 2011, the main binning efforts (and budgets) of the mainstream OC community have shifted towards recently launched Sandy Bridge chips (which also didn't cost $1k per try). But if 31xxC / 32xxC can demonstrate better core clocks compared to 3005F of 3031A then there is a chance for i7-980X / 990X history to be re-written.

ground says:

5 year and 8 months ago –

I mostly binned low end xeons (because thats what I could optain in bulk for cheap), and the C batch seems to be able to reach high bclks (271 at ambient is the highest I‘ve gotten so far, with maybe 30 c batch chips tested), core clocks were, with the couple chips I tested for core clocks, pretty similar to A and B batch (only had a couple of those though). Taggs chip here is a B batch but he has a similarly clocking C batch. If you want my bclk binning data shoot me a PM  If you want to test a couple low multi C batch chips feel free to shoot me a PM about that, I‘m sure I can arrange that...

suzuki says:

5 year and 8 months ago –

My 990x does 5030 @1.53v cb15 with Aio at regular ambient .

Really curious to see what this chip can do on ln2.

Will test iti n november ,when i get home ,on ss.

SparkysAdventure says:

5 year and 8 months ago –

@TAGG When you go for dual Westmere records, could you try memory on LN2 as well? 

If you're going to demolish the competition, you may as well go all out. I also recall hearing Hypers scale on LN2 (but only like one or two people had tested).

superpatodonaldo says:

5 year and 8 months ago – great job TAGG; it shuold deserve more points...