OK, Intel guys, beat this! :)

**Mafio** · 02-24-2013, 03:55 PM

awesome stuff there.
keep going.

**Don_Dan** · 02-25-2013, 03:39 AM

Crazy project Knut, I like it!

**chinook** · 02-25-2013, 02:01 PM

Your work is an inspiration to us all

**sin0822** · 02-25-2013, 02:14 PM

really nice job there

**knopflerbruce** · 02-28-2013, 07:48 PM

http://hwbot.org/submission/2361730_...he_30sec_108ms - this is on the 2nd board. One dead mem channel and no VRM-sinks... scary

, but pulled only 1000w. Though, overheating alarm started screaming after 5 seconds due to simple air cooling (2x Noctua U9DO's and 2 U12DO's).

**Movieman** · 02-28-2013, 07:58 PM

Originally Posted by knopflerbruce

http://hwbot.org/submission/2361730_...he_30sec_108ms - this is on the 2nd board. One dead mem channel and no VRM-sinks... scary

, but pulled only 1000w. Though, overheating alarm started screaming after 5 seconds due to simple air cooling (2x Noctua U9DO's and 2 U12DO's).

You darned well don't leave to much for someone to beat you short of breaking into Intels research lab at 2am..
wait a minute, hmmm, that might be the answer!

**delacroixp** · 06-17-2014, 11:24 PM

Great project ... beautiful metal crafting .
Few people really appreciate the raw savagery of parallel processing with 4 CPU's.

One is never enough!

It's all good

Pascal

**FlanK3r** · 06-30-2014, 01:05 AM

wow, this is true AMD machine

. I want this one unlocked CPU!

**lutjens** · 06-30-2014, 05:20 PM

I find it funny how some folks seem to think that a high core count CPU can't overclock much due to it's many cores. This is proof that the high core count chips from both AMD and Intel are capable of so much more than they are from the factory and can actually overclock quite well. I hope we get to play with unlocked 18-core Xeons this fall and if we do, we should see 4.5GHz overclocks out of it (albeit with considerable power consumption).

**knopflerbruce** · 07-18-2014, 02:11 AM

They do overclock alot worse than regular desktop chips, though. I ran them folding stable at 3.6GHz/1.15v for a long time (hot summer days included), but going higher was difficult - mostly because of the insane power draw (1250W, but peaks at 1300W+, causing my EPS cables to melt...). I have a feeling 4 GHz could be possible, unless something self ignites at 1.4+v :p Need a PAIR of ax1500i's for that project... Some PROPER EPS cables, and I wish I could add one more EPS connector to the board The free "slot" is not for power, someone at SM said when I asked if this was possible.

**lutjens** · 07-22-2014, 08:14 PM

Originally Posted by knopflerbruce

They do overclock alot worse than regular desktop chips, though. I ran them folding stable at 3.6GHz/1.15v for a long time (hot summer days included), but going higher was difficult - mostly because of the insane power draw (1250W, but peaks at 1300W+, causing my EPS cables to melt...). I have a feeling 4 GHz could be possible, unless something self ignites at 1.4+v :p Need a PAIR of ax1500i's for that project... Some PROPER EPS cables, and I wish I could add one more EPS connector to the board The free "slot" is not for power, someone at SM said when I asked if this was possible.

As you say, there is a power supply problem. The CPUs themselves seem capable of the higher speeds, but limitations with the power supply itself and with the VRMs on a server board that were never intended to push that much current are the limiting factor. You are running four CPUs...running a dual CPU system shouldn't draw nearly as much power through the EPS cables and this issue should be far less of a concern than with a quad configuration. For those who want to run one high core count CPU on a Rampage V Extreme-caliber board, a highly overclocked 18-core wouldn't be a real problem for the board or it's power delivery system (as long as a decent power supply is used). While the CPU won't break any clock speed records, it should still be able to get with a few hundred MHz of a chip with fewer cores (I'm guessing at around 4GHz for an 18-core CPU while maintaining decent power draw levels and thermals).

My $0.02....

**knopflerbruce** · 07-23-2014, 12:18 PM

Originally Posted by lutjens

As you say, there is a power supply problem. The CPUs themselves seem capable of the higher speeds, but limitations with the power supply itself and with the VRMs on a server board that were never intended to push that much current are the limiting factor. You are running four CPUs...running a dual CPU system shouldn't draw nearly as much power through the EPS cables and this issue should be far less of a concern than with a quad configuration. For those who want to run one high core count CPU on a Rampage V Extreme-caliber board, a highly overclocked 18-core wouldn't be a real problem for the board or it's power delivery system (as long as a decent power supply is used). While the CPU won't break any clock speed records, it should still be able to get with a few hundred MHz of a chip with fewer cores (I'm guessing at around 4GHz for an 18-core CPU while maintaining decent power draw levels and thermals).

My $0.02....

You're sort of right with the EPS argument, however the two SM 1P G34 boards I know of only have a single 8 pin EPS connector. It helps a little bit, but not enough :p When/if there are 2P unlocked xeons on the marked I hope the mobos come with at least 3x EPS connectors, or maybe 2x 8pin EPS + 2x 6/8pin GPU power connectors dedicated to the CPUs. That'll be needed when pushing the limits... 18 cores on RVE will be easy :p 2x 8pin EPS for just one CPU is plenty.