Swapped 975xe for 980x

vvhocare5

Distinguished
Mar 5, 2008
768
0
19,060
Well, my 980X showed up a couple of days ago and I was curious how it would perform in the same motherboard. Its an Intel DX58SO, bios had been updated in preparation for this.

This is not about how best to OC a 980X - it is about the differences using the same motherboard and changing from a 975XE to a 980X.

The existing 975XE was a great performer and easy to OC. I OC'd it to 4.2GHz just by increasing the multiplier. Left the voltage on Auto so it would clock down when not in use. Increased memory speed to 1600 (at 7-7-7-20) from the default of either 1066 or 1333 (RAM is DDR3-2000). In Prime95 it would hit 85C - I use a Thermalright Ultra 120 Extreme for CPU cooling. The CPU voltage would auto increase to around 1.35 under full load. Other voltages were left at Auto or defaults. Gaming or other high load situations the 975 would run at 75C and idle at 33-35C. That was more than enough for me for the performance I was getting. I can do more but it begins to take a lot of work.

So the 980X shows up and I swap it in - maybe 20 minutes.

Lets see if it will boot? Its supposed to drop in and run right? Same interface, right? Power on, no explosions (always a good sign). Nope, bios error screen, ok go in, reset to defaults since there is a different CPU in there now. Most Processor selections are on AUTO, memory is still at 1600.

No boot - bios error screen. Drop memory to 1333. Again, no boot - bios error screen.

Drop memory to 1066 (default) and bite the bullet and select Auto - rats, rats, rats...
Finally we get a boot running basically stock 980X settings and speeds. Windows 7-64 starts up and says settings have changed and it needs a restart. OK, reboot, no issues back in Win 7. I play around with a few games and apps.

So first observation is the memory controller is different than in the 975 - same RAM, same motherboard, same settings and the 980X wont start. I play around a little with RAM settings without a lot of success and decide thats enough for now. Ill spend some time on that later. And its better to work on CPU OC with RAM at stock or less anyhow.

Lets try OC'ing the CPU - thats what I really wanted to do first anyhow but got sidetracked with the memory.

So I take the easy route that I used on the the 975. Most things on Auto, set multiplier to 30 and.....no boot. Huh? Again?

OK, CPU settings on Auto and after some repeated attempts it only boots at about 3.46, anything higher than that and its no good - Windows crashes just as the desktop begins to load - and does a reboot (this is the only failure mode it exhibits).

Hmmm - ok lets go to a 30x multiplier - which is about 4.0GHz and increase the the CPU voltage. 1.15 no good, 1.20 - no good, 1.25 - we boot into Windows.
Seems stable, play various games - launch Prime95 - most cores run but one generates an error after 15 minutes. CPU voltage to 1.30 and Prime95 runs flawless.

Now the amazing thing is the heat - or lack thereof. Task manager shows all 12 threads at 100%, I leave, grab a beer come back its still at 62C. Seriously? No errors in Prime95 - all cores running. I tried the other options in Prime95 and the ones that exercise the CPU made it run even cooler, it would drop into the 50's.

So its stable enough at this point - I know some prefer to run it for 24hours, but thats not me.

Lets try increasing the clock again. 31 multiplier same voltage - at 1.30v same pattern begins to emerge. Windows will start but Prime95 will generate a couple of errors after a while. CPU Voltage to 1.35v and it runs error free again. Temps in low 60's still. multiplier to 32 which is 4.25GHz windows starts, but Prime errors after a while, temps still in the mid 60s. I really dont want to run at 1.4V so I back down the settings to 4G and 1.30v

That was my first couple of days. These were the highlights and I was running memtest in there, various games and so on. Its interesting but tedious.

The low heat stuns me, Im going to look at other ways to measure it, not entirely sure what Im seeing personally. Im running a variety of tools and they are reporting the same measurements.
At this same voltage and clock the 975 was at 85C+

Power at idle is reported at 68W, very similar to what the various sites report, no more 6W at idle Im afraid, temps are 30-33C across the 6 cores.

The 980X does not like AUTO for OC'ing, Im not sure if its a bios issue in ramping up power fast enough for it or the DX58SO isnt up to it, but I had to select manual voltages to get it to OC at all.

I never saw 70C in anything I tried. Not sure what the uber OC types can do with this but I suspect there is some headroom out there. But the limiting factor may be the voltage one has to pump into it...

My personal style is to have parts drop their clocks and voltages when Im not using them, mainly for better life expectancy.

Im going to go back and play around with the other multipliers and voltages to see what their effects may be on CPU multipliers and try to get the RAM up a little - but without increasing the latency past 7-7-7. Im curious about the effects of QPI voltage since we have more cores connected.

And of course different parts, mb's, PSU's, RAM and so on will generate different results...
 

ubernoobie

Distinguished
May 29, 2009
886
0
19,010
DO NOT PUSH 32NM CHIPS PAST 1.3V, YOU WILL REGRET IT LATER ON. Sorry about the caps but, please do not wreck your i7 980x by increasing voltage to 1.3+. The 32nm chips are much more delicate to voltage than 45nm. 45nm max voltage is around 1.45 using intel's process. 32nm chips in the core i3/i5 have max voltage of 1.4v. Any higher and it may fry your processor regardless of temperature. For 32nm chips, you will have to start changing from looking at temperature to all out voltage since voltage is what 32nm chips hate. Good luck on your 1k+ chip
Edit: You can push it past 1.3v but exceeding 1.4v WILL fry ur cpu so stay around 1.3
 

vvhocare5

Distinguished
Mar 5, 2008
768
0
19,060
No problem with the caps - appreciated. For now 1.3v is as far as I am personally comfortable with. Any insight into multiplier improvements by changing other voltages or restrictions on the other voltages?