Overclocking Escapades: socket LGA1366 Intel Core i7 920 D0 stepping

[Tech Tweaker]

CPU: Intel Core i7 920 D0 stepping (SLBEJ)
Motherboard: Asus Rampage III Formula
Memory: G.Skill 6GB (3x2GB sticks) F3-12800CL8T-6GBRM (DDR3-1600/PC3-12800) 8-8-8-24-2T @1.6V
Graphics: Nvidia GTX 780 Ti or GTX 580
CPU cooling: Corsair H100i with 2x Scythe Gentle Typhoon AP-15's
SSD: Samsung 840 Evo 120GB (Overkill as it turns out, as the SATA bus on this motherboard isn't able to take advantage of its full speed. I probably need to downgrade this system to a SATA II drive at some point, maybe an OCZ Vertex 2 or something.)

I actually tested i7 920 #1 a while ago, but didn't get around to testing the other 2 920's I have to see which one was best. Mainly just because I had one CPU and knew it was stable, and needed the system to be stable for some testing I was doing with other parts.

#1 Batch No. 3943A524
2.8GHz (21x133MHz): 1.10V (1.1 V)
3.2GHz (21x152MHz): 1.10V (1.1 V)
3.3GHz (21x158MHz): 1.208V +1.31-1.402V QPI/DRAM (1.2V +1.375V QPI/DRAM) (62°C-60°C-58°C-60°C)
3.4GHz (21x162MHz): 1.256-1.264V +1.303-1.402V QPI/DRAM (1.25V +1.375V QPI/DRAM) (64°C-62°-61°C-61°C)
3.5GHz (21x167MHz): 1.304-1.312V +1.323-1.402V QPI/DRAM (1.3V +1.375V QPI/DRAM) (70°C-65°C-67°C-66°C) (vCore needs to be lowered and retested, as it was apparently higher than necessary)
3.6GHz (21x172MHz): 1.312V +1.33-1.396V QPI/DRAM (1.3V +1.375V QPI/DRAM) (71°C-68°C-69°C-67°C) (vCore needs to be lowered and retested, as it was apparently higher than necessary)
3.7GHz (21x177MHz): 1.312V +1.33-1.396V QPI/DRAM (1.3V +1.375V QPI/DRAM) (72°C-69°C-69°C-68°C) (vCore needs to be lowered and retested, as it was apparently higher than necessary)
3.8GHz (21x181MHz): 1.312V +1.33-1.396V QPI/DRAM (1.3V +1.375V QPI/DRAM) (72°C-69°C-70°C-68°C) (vCore needs to be lowered and retested, as it was apparently higher than necessary)
3.9GHz (21x186MHz): 1.232V +1.33-1.396V QPI/DRAM (1.225V +1.375V QPI/DRAM) (67°C-63°C-64°C-63°C)
4GHz (21x191MHz): 1.256-1.264V +1.323-1.402V QPI/DRAM (1.25V +1.375V QPI/DRAM) (69°C-65°C-66°C-64°C)
4.1GHz (21x196MHz): 1.296-1.304V +1.316-1.396V QPI/DRAM (1.2875V +1.375V QPI/DRAM) (74°C-70°C-71°C-70°C)
4.2GHz (21x200MHz): 1.336-1.344V +1.323-1.396V QPI/DRAM (1.325V +1.375V QPI/DRAM) (79°C-76°C-77°C-75°C) *Isn't actually fully stable at these settings apparently, as it randomly crashes when Speed Step, C1E, and C-States are enabled for power saving. Odd thing was it crashed while at the desktop doing nothing, but was perfectly stable for a 2-hour gaming session of Battlefield 1 when EIST, C1E and C-States were disabled.


#2 Batch No. 3911A234 (seems to have been lapped at some point...? Not by me. Surface of the IHS is oddly shiny and flat, even though the etching of the model number, batch number and so forth are still readable.) Got it in a parts combo with a motherboard back in Summer 2017 I believe it was. Doesn't seem to clock very well, so may just resell it with the same motherboard it came with on that basis.
3.8GHz (19x200MHz): 1.232-1.240V +1.33-1.396V QPI/DRAM +1.283-1.29V IOH (1.225V +1.375V QPI/DRAM +1.311V IOH) (68°C-67°C-68°C-67°C) *still testing for lower CPU vCore
3.9GHz (20x195MHz): 1.336-1.344V +1.33-1.396V QPI/DRAM +1.283-1.29V IOH (1.325V +1.375V QPI/DRAM +1.311V IOH) (85°C-85°C-83°C-83°C)

Tried for 4GHz with this chip, couldn't get it stable even with 1.45v and wasn't willing to go higher as it was hitting 88°C within the first few minutes of firing up prime95 Blend test. This chip apparently is a terrible overclocker.


#3 Batch No. 3945A854 *Haven't tested this one yet, but upon searching online I found several posts from people asking about this batch number and one post about someone claiming to have ran one at 3.8GHz with a very low 1.17v. So, hopefully its from one of those "Golden" batches and clocks well at low voltage. Certainly can't be much worse than the batch 3911A234 chip, as that one needs a lot of voltage.
3.8GHz (19x200MHz): 1.184-1.192V +1.263-1.356V QPI/DRAM +1.283-1.29V IOH (1.175V +1.325V QPI/DRAM +1.311V IOH) (67°C-63°C-65°C-61°C)
3.9GHz (20x195MHz): 1.24V +1.257-1.356V QPI/DRAM +1.283-1.29V IOH (1.225V +1.325V QPI/DRAM +1.311V IOH) (75°C-71°C-71°C-69°C)
4GHz (20x200MHz): 1.280-1.296V +1.263-1.349V QPI/DRAM (1.275V +1.325V QPI/DRAM) (77°C-74°C-74°C-71°C)

 

[EarthDog]

Throwback thread, lol! Awesome!

 

[Johan45]

Nice TT hope that is a sweet heart for ya. Hard to find a true golden chip

 

[dejo]

if you are gonna bench for boints- you will prolly want to work with slow mode on that board. the one that I had was limited to around 220bclk when not in slow mode

 

[Tech Tweaker]

if you are gonna bench for boints- you will prolly want to work with slow mode on that board. the one that I had was limited to around 220bclk when not in slow mode

I've already benched and submitted with i7 920 #1 and this board back in August 2017. Board topped out at 219.36MHz with my Xeon L5506 in it, Slow Mode made no obvious difference for me. Really I thought Slow Mode was just for if you were running with Sub-Zero cooling, so I've only ever engaged it a few times when I was at my max OC to see if I could get more.

My Rampage II Extreme also topped out at 219.82MHz (nearly 220MHz) (before the board died of an electrical short), there seems to be a theme for topping out at or around 220MHz with these X58 boards.

 

[dejo]

most of my asus 1366 boards topped out at 215-220 without slow mode, many would do around 235-240 with it enabled. But I did have one deluxe board that would do in the mid 250s in high speed mode

 

[UltraTaco]

Thank you for sharing! I assume all those tests HT on?

I'm gonna check in with my numbers(signature). I disabled ht after 3.6 because temps get warmer. Probably could go higher with it on now that I'm no longer in FL and radiator is cleaned up, but don't feel like revisiting prime at this time..

Cheers! 920 still going strong!

 

[Tech Tweaker]

Thank you for sharing! I assume all those tests HT on?

I'm gonna check in with my numbers(signature). I disabled ht after 3.6 because temps get warmer. Probably could go higher with it on now that I'm no longer in FL and radiator is cleaned up, but don't feel like revisiting prime at this time..

Cheers! 920 still going strong!

Yes, HT on. I don't turn HT off unless I'm running a benchmark that doesn't benefit from having the additional threads generally.

 

[UltraTaco]

Aha, I see. Thank you. Even more impressive then!!

 

[Tech Tweaker]

Chip #2 definitely isn't a silicon lottery winner. It has what I would consider a really insane voltage jump to go from 3.8GHz to 3.9GHz (1.225v to 1.325v), a .10v increase (a tenth of a volt) is really quite a large increase for only 100MHz more. I could understand said increase if it were at a very high clock speed and going from 4.3GHz to 4.4GHz lets say (Very high clock speed for an i7 920, near to well impossible with air cooling or an AIO cooler from what I've seen so far on LGA1366 and how much heat these chips produce once you go beyond say 1.275v.), but at 3.9GHz that level of increase doesn't make much sense as this isn't what I would consider that high of an overclock. It might be able to achieve say 3950MHz/3.95GHz, but at 3.9GHz with the core temps hitting 85°C it's already at a temperature level that I'd be uncomfortable running it at in a PC that I would want to be 24/7 stable, even if it only hit said temperatures for a few hours at a time. The room I have this PC in became noticeably hot while I was testing this computer for stability, to the point where I had to set up a box-fan in the doorway blowing cool air into the room to cool things down and make it more bearable to be in the room with it.

Not only is it not a silicon lottery winner, but it runs genuinely hotter than chip #1 at the same voltage, by around 6°C on the hottest core and 8°C on the coolest core, even though chip #1 was clocked 300MHz higher.

#1 Batch No. 3943A524
4.2GHz (21x200MHz): 1.336-1.344V +1.323-1.396V QPI/DRAM (1.325V +1.375V QPI/DRAM) (79°C-76°C-77°C-75°C)

#2 Batch No. 3911A234
3.8GHz (19x200MHz): 1.232-1.240V +1.33-1.396V QPI/DRAM +1.283-1.29V IOH (1.225V +1.375V QPI/DRAM +1.311V IOH) (68°C-67°C-68°C-67°C) *still testing for lower CPU vCore
3.9GHz (20x195MHz): 1.336-1.344V +1.33-1.396V QPI/DRAM +1.283-1.29V IOH (1.325V +1.375V QPI/DRAM +1.311V IOH) (85°C-85°C-83°C-83°C)

Up next is the i7 920 I got for $0.99 plus shipping on ebay (or somewhere around there, may not remember accurately as it's been a while ago), because the seller didn't think it worked.

 

[UltraTaco]

Why do you have such high qpi voltage? Fast ram+NB?
Mine was 1.275vtt 1523ram X2 NB. (4ghz cpu)

 

[Tech Tweaker]

Why do you have such high qpi voltage? Fast ram+NB?
Mine was 1.275vtt 1523ram X2 NB. (4ghz cpu)

Well, at the current settings for 3.8GHz (200Mhz x 19) I've got my NB Frequency at 3408MHz and Memory at 1600MHz, my QPI link speed is over 3600MHz/7200MHz effective speed (on CPU's that are rated between 2400(4800)-2933(5860)) (all the Core i7's I've looked at seem to be rated at 2400MHz, but several Xeons are rated at 2933MHz) so it needs the additional voltage to keep everything stable. Since almost no LGA1366 chip that I'm aware of is rated to run 1600MHz RAM speed (most only officially support 800MHz or 1066MHz max), I guess I'm technically overclocking the RAM Bus on every CPU I use and it's technically a fast RAM speed for the time.

And because it was easier to set it at the max I needed for my max OC on my Xeon E5640 (210MHz x 21/4.4GHz) and leave it there, than to have to re-test every chip to find out what each needed for the QPI/DRAM voltage. Basically, I was a little lazy and eliminated that step from my testing because it meant it would take longer to bin each chip if I had more than just the CPU's vCore to worry about. So, I can go from a minimal overclock all the way to max OC without having to worry about raising the QPI/DRAM/VTT voltage as I go. The E5640 got unstable once I went past 4.0GHz (191x21) (1.25v QPI), then needed another bump at 4.1GHz (196x21) to 1.275v QPI, 4.2GHz (200x21) was 1.325v QPI, and 4.3 and 4.4GHz (205-210x21) needed 1.375v set in the BIOS. Though my i7 930 only needed 1.325v max on the QPI/DRAM to remain stable at its maximum speed of 4.1GHz.

 

[UltraTaco]

Understood, good point. Thank you!