Thanks...and you're the lucky guy I thought I briefly answer as there were several questions like this and also others far less sensible - good thing I did not post the 5.3 GHz results
The whole point of the posting exercise was JUST to do a direct comparison of 'before and after' re delidding, and with CL Ultra on the die. I am not doing any top-speed validation runs yet with this semi-incomplete setup.
The screenshots with the windows resource manager (and other tools) indicating 100% load on all cores came in two batches:
4.8, 4.9 and 5.0 and then 5.0, 5.1, and 5.2. For all these tests ('before and after'), I used Aida64 Extreme Edition 'System Stability Test' to allow for a direct comparison of 100% load - which I stated in the original post, btw
The only real differences were ambient temps (up to 3 C variance) ...and the first 3 tests were done on a Sabertooth Z77 (which is now in a VM) and the last 3 on a Maximus V Extreme. Because of the ambient and motherboard differences, I quickly ran Cinebench (two cycles each) right before and after the delid on the same mobo /ambient temp - Cinebench is NOT, and never has been, my be-all-end-all stress test.
I also keep on mentioning (hoping that one day someone will actually 'get it') that this is a partial work-related 'test' system and as such has MULTIPLE instances of SQL Server installed, including SQL ENTERPRISE which is a big memory and thread hog - in fact just the enterprise version has multiple times the memory usage than the next largest user (other than Win 7 systems stuff) via Resource / Processes.
Then there is also a RAM Disk...the SQL has a pres-set minimum of 8 gb of RAM and can use up to 16, the RAM disk has another 8 GB. SQL Enterprise is also paired with Microsoft IIS web server, addressed in localhost by the VM systems at all times trying to mimic real-world conditions, even when stress testing.
Some folks here seem to suggest that I should take 90 % of the RAM and assign it to Prime95 ...please folks - do the math, as much as I always do Prime95 testing with whatever is available, the system decides that automatically.
The are two issues I'm trying to deal with now. The first is that the delidding / cooling upgrade is not yet finished...the IHS is not seated quite right (about .75 to 1 mm 'too far back') but I don't think that that is a big issue - certainly does not seem to affect performance, and I did get the CL-U application on the die just right, so I am in no hurry to take the IHS off the die again. Of bigger concern is the fact that the close-water-loop block is a bit loose on one of the four corners as the corresponding rear retaining nut has worked itself deeper into the plastic retainer slot - I can twist the whole screw / nut assembly even when on full lock
To compensate, I added a much thicker layer of MX4 on the die / water block area ON PURPOSE than I otherwise would, knowing that within a week or so, a brand new custom loop / Koolance block / big 360 rad will go on anyways
The second issue I am still researching...the Watt numbers are getting quite scary once past 5.1 Ghz, never mind 5.3. A 3770K is rated at max 77 watts, but at the speeds I am running and I shared with you, the chip pulls almost three times its rated max (see also Sin0822's table on that on this site). I'm getting a kill-a-watt meter on the weekend to do some more benching when benching, so to speak.
Edited by Joa3d43 - 2/14/13 at 3:33am