Intel sent me a QS 7350K processor to test out and give feed back on. QS stands for Qualification Sample and this is a sample that would be exact to the parts shipping in Retail.
Looking at the batch information we can see the chip was manufactured the 39th week of 2016. A great week if you were seeking a 7700k batch.
http://www.xtremesystems.org/fugger/a2.jpg
What makes this chip special is that is an i3 dual core entry level processor with the K skew. K means that the CPU is fully unlocked for overclocking and that is just what the people wanted.
These processors should be available everywhere by the end of the month for around $180 US.
http://www.xtremesystems.org/fugger/a4.jpg
A whole bunch of supporting hardware that I tested on:
Asus Z270 Maximus IX Formula
Asus Z170 Maximus VIII Impact
Gigabyte Z270X-Gaming SOC
Gigabyte Z170 Designare
MSI Z170 Platinum
Intel 750 SSD 1.2TB PCIe
Galax 4000C19 HOF B-Die
Galax GT 710 VGA
Enermax MAXREVO 1500W
Lots of LN2:
http://www.xtremesystems.org/fugger/a1.jpg
I first tested before I delided the CPU, I found the TIM under the 7350K is sufficient for almost reaching the top speed of the chip. Using an F1 LN2 pot from K|ngp|n to get my initial measurement. The VID was 1.312v before delidding and the TD between the CPU and the pot was 15c @ stock. Without any voltages set for cold bug or max frequency memory I set off for the top speed of the CPU pre delid.
I set vcore to 1.65v and and multiplier 60x I was able to boot into windows at -100c and run XTU, I scaled up to 6.3Ghz in XTU before I reached the end of the TIM's capability at -125c.
http://hwbot.org/submission/3420848_...0k_6496.81_mhz
There was no issue with the original TIM cracking and I had a lot of hope that the chip would keep scaling after delid.
http://www.xtremesystems.org/fugger/a3.jpg
After delidding a bunch of 6700K's and 7700K's, the 7350K lid pops much easier.
http://www.xtremesystems.org/fugger/a7.jpg
The 7700K on the top, the 7350K on the bottom, as you can see the 7350K is a much smaller die.
I cleaned up the old TIM and installed some fresh Thermal Grizzly Kryonaut thermal paste. It is ok to use a little too much here, make sure you cover the entire edge of the CPU.
My TD dropped from 15c to 5c and my VID dropped to 1.25v, I was able to run Cinebench R15 at 5.1Ghz @1.3v and my temps were under 50c max. A huge improvement at stock. I recommend delidding if you have a delid tool and proper thermal paste. DO NOT unnecessarily endanger your CPU with a razor blade or vice to delid to get those extra degrees if you are just running an AOI or HSF, the original TIM works great. But to possibly get that extra 100Mhz you will need to delid and use the best TIM.
All the motherboards worked great and had no problem running the XMP settings with the most recent BIOS. My XMP profile is a 4000C19 profile and the 7350K IMC had no problem delivering, my cache and IMC were just as strong as the 7700K in terms of reaching high frequencies with minimal cooling on CPU and RAM.
Sticking with XTU and a delidded CPU I set off again for max frequency of the CPU. I set my rails for no cold bug to find that I really needed to push these rails hard, much harder than the 7700K to reach full pot on the Asus board, the Gigabyte and MSI needed a smaller voltage bump to run at full pot. I really love how all the manufacturers use different labels for the rails that need to be adjusted so check the correct guide for each board to find these settings as all of my boards needed different settings. Gigabyte was the easiest to hit full pot settings but the BIOS is not mature and a few bugs made it hard to tune memory settings and reboot beyond -125c. Both MSI and Asus had no problems rebooting at full pot and both MSI and Asus have B-Die profiles to reduce tuning time drastically.
All three boards have OC controllers, Gigabyte is not fully functional yet, it does not have an LCD display, it has voltage read points for all rails, it does not accept a thermal probe input, no voltage adjust, and it is more of a fan controller than an OC controller. The MSI is smaller, the buttons work only when you use the supplied (short) cable (does not work plugged directly into the mobo), slow mode switch, no voltage adjust, no LCD, no input for temp probe. The Asus Panel is much larger than the other two combined, it has a long cable, an LCD display, two temp probe inputs, slow and and pause buttons, fan headers, voltage control of some rails (menu on LCD), stylish plastic case. Asus wins.
http://valid.x86.fr/75l61g
Even with full pot I could not get a stable 6.7Ghz for screen shot, I tried on all boards and failed. My top benching speed was 6.5Ghz 2c4t for XTU, only 200Mhz more than I ran before delidding, I really hoped it would have scaled more but I was already pushing voltages beyond safe levels.
http://hwbot.org/submission/3423481_...0k_1178_marks/
Just to recap, with an AOI hitting 5Ghz will be very easy and I would expect to see a lot of entries over 5.1Ghz for benching on AIO and water cooling. Hitting 6Ghz was easy on phase change, you will need to find the lowest vcore to reach the max frequency with stability. With LN2 these are the most fun and least expensive to really enjoy Kaby Lake overclocking.
Also during my tests I had a chance to use the smallest socket 1151 heat sink made by Intel, I think it came with a Celeron. I was able to boot into windows at 5Ghz @ 1.35v but I was not able to complete XTU.
A huge thanks to Intel, Asus, Gigabyte, MSI, Galax, and Enermax.
Original Images:
http://www.xtremesystems.org/fugger/IMG_1479.jpg
http://www.xtremesystems.org/fugger/IMG_1484.jpg
http://www.xtremesystems.org/fugger/IMG_1487.jpg
http://www.xtremesystems.org/fugger/IMG_1488.jpg
http://www.xtremesystems.org/fugger/IMG_1490.jpg
http://www.xtremesystems.org/fugger/IMG_1491.jpg
http://www.xtremesystems.org/fugger/IMG_1512.jpg
Bookmarks