Overclock.net › Forums › Intel › Intel - General › Xeon Hacking and Overclocking X79/X99 [Beyond X58]
New Posts  All Forums:Forum Nav:

Xeon Hacking and Overclocking X79/X99 [Beyond X58] - Page 112

post #1111 of 1201
Quote:
Originally Posted by 99belle99 View Post

I gave up on that idea. I just seen the ad yesterday and thought maybe I should get it. I was able to pick up all three: X79 Sabertooth, i7 3960X and 32 GB's of RAM for €600. Seemed ok yesterday but thinking about it more now I could pick up a more modern CPU for similar money. Like a i7 Kaby Lake. But after thinking about it more I am still happy with my X58 system. Plays all my games brilliantly.

If anything, I'd get a 1650 or 1650v2 as they are pretty cheap now, Ivy has a bit higher IPC but Sandy usually clocks higher. Look at it this way, a 1650 at 4.6ghz is equivalent to a 6850k @ 4ghz in most cases, not bad for an $80-90 chip.
Main
(18 items)
 
Server
(8 items)
 
 
CPUMotherboardGraphicsRAM
E5-1660 @ 4.6ghz 1.275v Asus x79 Deluxe Powercolor PCS+ R9 390 32gb Patriot Viper III @ 2400 11-12-11-23 1t 
Hard DriveHard DriveHard DriveHard Drive
4x Toshiba 4tb 7200rpm 2x Toshiba 5tb 7200rpm Samsung 850 EVO 500gb Samsung 840 EVO 500gb 
Hard DriveCoolingOSMonitor
Crucial BX100 500gb Noctua NH-D15S Windows 10 Pro 2x Lenovo L220x 
Power
Corsair RM850i 
CPUMotherboardGraphicsRAM
2x Xeon x5670 ASUS Z8NA-D6C Windforce 7950 48gb NANYA DDR3 1333mhz 
Hard DriveCoolingPowerCase
2x Toshiba 2tb 7200rpm Raid 1 2x Raijintek Aidos 750w ToughPower Gold DIYPC D480-BK 
  hide details  
Reply
Main
(18 items)
 
Server
(8 items)
 
 
CPUMotherboardGraphicsRAM
E5-1660 @ 4.6ghz 1.275v Asus x79 Deluxe Powercolor PCS+ R9 390 32gb Patriot Viper III @ 2400 11-12-11-23 1t 
Hard DriveHard DriveHard DriveHard Drive
4x Toshiba 4tb 7200rpm 2x Toshiba 5tb 7200rpm Samsung 850 EVO 500gb Samsung 840 EVO 500gb 
Hard DriveCoolingOSMonitor
Crucial BX100 500gb Noctua NH-D15S Windows 10 Pro 2x Lenovo L220x 
Power
Corsair RM850i 
CPUMotherboardGraphicsRAM
2x Xeon x5670 ASUS Z8NA-D6C Windforce 7950 48gb NANYA DDR3 1333mhz 
Hard DriveCoolingPowerCase
2x Toshiba 2tb 7200rpm Raid 1 2x Raijintek Aidos 750w ToughPower Gold DIYPC D480-BK 
  hide details  
Reply
post #1112 of 1201
I wonder if it would be possible to run 2 of those ES 2695 v3 with the hack in one board? It's fun just thinking about it, hehe..
post #1113 of 1201
Quote:
Originally Posted by Hexbyte View Post

I wonder if it would be possible to run 2 of those ES 2695 v3 with the hack in one board? It's fun just thinking about it, hehe..
Yes. It required more EFI work to get the second processor stripped and programmed, but was eventually accomplished in that thread with a Z10PE-d16 and the Asrock dual board (don't recall the name). Possible some others. See the anandtech thread.

I was going to go that route, but after some testing, I found 2 Taichi w/2696 produced higher throughput than would be possible with a dual board at pretty reasonable power consumption (~250W measured at the wall under 18C load).

Two reasons:
1. I can push ram timing - running with 2133CAS10-1T - produces repeatable 10% improvement in my target apps over 2133CAS15 (which is all I could get with ECC on a C612 board that blocked me from changing timing)

2. My target apps are not scaling to dual socket systems and it will be some time before I can address that.

So, the V4's get to live in the Z10PE and the V3's now outperform them core-for-core in 2 taichi boards.
post #1114 of 1201
Oh, that's super cool. I checked the thread only briefly so I did not see it. I'll look for it now!

There probably aren't a whole lot of applications that would support that many cores (especially on dual socket systems?) I was just thinking of benchmarking with it for fun. Not that I really have any reason to nor can I actually afford it, haha.
post #1115 of 1201
Quote:
Originally Posted by Hexbyte View Post

Oh, that's super cool. I checked the thread only briefly so I did not see it. I'll look for it now!

There probably aren't a whole lot of applications that would support that many cores (especially on dual socket systems?) I was just thinking of benchmarking with it for fun. Not that I really have any reason to nor can I actually afford it, haha.
Yeah. its fun. You might start at the end and work backward from the end to find the V3x2.efi conversations and links.
post #1116 of 1201
I actually read through it all from start to finish, haha. Seems like only 2 people really cared to use it in a dual socket system and that the performance overall was lacking compared to the improvements of single socket systems.
post #1117 of 1201
Quote:
Originally Posted by Hexbyte View Post

I actually read through it all from start to finish, haha. Seems like only 2 people really cared to use it in a dual socket system and that the performance overall was lacking compared to the improvements of single socket systems.
I've found it to creates a system that is pretty sensitive to temperature.

Because I planned on moving it to a dual system, I just cobbled together the single system to tests and ended up with a noctua sink on one and an H110 on the other.

There was a noticeable difference in performance between the two running a ~3 hour compute job. I dug in and figured out it was entirely temperature/power throttling subtly reducing performance throughout the job. So, I set then both up with the same clocks, voltages, and H110 cooler and now they both perform the same.

From memory, I think it was 2:40 for the H110 and 3:15 for the noctua which is 160m vs 195m or the noctua was 22% slower.

So, a small sample of people might not give you the whole answer. The difference in temps was not huge (10-12C), but its effect on current and thus the hard TDP cap evidently was enough to make a big difference on a job making use of 16-18 cores...

As I mentioned, the other wild-card is memory. A dual system will clamp you to stock ECC 2133,CAS15,CR2 memory speeds where a single system would generally allow you to use unbuffered dimms at whatever timing the BIOS will allow and the memory will tolerate... That can be another 10% difference right there...
Edited by cekim - 4/14/17 at 11:47am
post #1118 of 1201
Hmm, I was under the impression from the other thread that the issue generally wasn't heat but the amount of power your system could use, so what you are saying is new to me. I guess the RAM does limit your options a lot but it should at least be a static increase/decrease so to speak, no throttling or clocking differently!

Considering your experiences tell a different story than the people in the other thread I think this is still an interesting (and fun) option to consider for a system! :]
post #1119 of 1201
Quote:
Originally Posted by Hexbyte View Post

Hmm, I was under the impression from the other thread that the issue generally wasn't heat but the amount of power your system could use, so what you are saying is new to me. I guess the RAM does limit your options a lot but it should at least be a static increase/decrease so to speak, no throttling or clocking differently!

Considering your experiences tell a different story than the people in the other thread I think this is still an interesting (and fun) option to consider for a system! :]
power + heat are the limiting factor. The presumption is that heat is only a binary issue (throttling), but it doesn't seem (empirically) to be that simple. It looks like there's a sliding scale and/or the additional resistance caused by hotter gates is showing up in current consumed (resistivity).

Don't have a solid answer for you, just these two systems...
post #1120 of 1201
That's okay, you've given me a lot of information about things I had not considered before and I'm thankful for that!
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel - General
Overclock.net › Forums › Intel › Intel - General › Xeon Hacking and Overclocking X79/X99 [Beyond X58]