I have opted for this rather sleek board after having had nothing but trouble with Z170 boards of the high end regime.
After having read many articles that all came to the conclusion that the 7700k overclocks alike across low-2-highend boards
I decided to try this sleek approach and invest the saved money in some fast RAM and other goodies of actual use.
The board gave me problems when I booted it up 1st time with all 4 modules installed, it blocked. Took 1 and 3 out and it booted with 16GB just fine on XMP. I already had the newest Bios on a stick, flashed it, put all RAM back in and it booted all 4. That's OK with me with such brand new stuff, no points lost, also the RAM is NOT on the QVL !
Win10 installation and application installation went without any errors, the machine felt good right away.
I then tested the RAM and CPU and default settings, XMP applied ( with the ALL CORE option active ).
FAILED ALL TESTS ! nothing worked p95, aida64 etc.. they all either dropped cores or aborted due to errors.
I gradually tested vCore, VCCIO ( was already at 1.2v through XMP I guess ) and VCCSA ( already at 1.35v and somewhat alerting ). None of those stopped the erorrs from occuring.
Turned out I had to up the vDIMM to 1.37228v as lowest where it would stop dropping cores or abort aida64. It then ran stock clocked and XMP for 1h each test. The VCCIO and VCCSA have been upped by the Auto setting to 1.20 and 1.35 Volts.
Since I had a rather golden Sandy that ran 5.0GHz and booted till 5.245GHz my expectations were high, my goal was set.
I have to say, I focused mainly on IPC but that much I can say, this CPU offers a lot more than more IPC compared to Sandy Bridge. The overall boost is present, everywhere...if you reach the 5GHz or higher.
Bringing the CPU to stable 5GHz and RAM at XMP3600 seemed impossible for the first week. It was easy to get it to 4.8 by only upping the vCore to 1.30v and leave the rest alone, on auto. No Phase boosting, no LLC ( default is LLC2, not LLC1 btw ). During stress testing random errors occured, and the only thing that helped was raising VSSIO even further to 1.225v, again, everything else as low as possible.
4.9 at first seemed to not need any higher Voltage but again during the 1h tests it failed sometimes so I upped the vCore by 0.01V and could "almost" stabilize it again for 1h runs aida64 and p95 ( both versions, 26.6b3 w/o AVX and 28.1b1 with AVX ). Random errors and differencies when those occured and how they occured let me to raise Phase management to 140%
and PhaseControl to Extreme in Asus UEFI Bios. LLC was left at default 2 and all other voltages dialed back to 1.31vCore, 1.225 IO, 1.35 SA, 1.3728vDIMM. It then ran stable for hours and hours. As this was my jumpboard to 5 I left it priming all night and aida64 all morning. Temps were OK, with the rather big external radiator and HK-IV waterblock I stayed well below
critical temps. I have not taken notes of the temps form those intermediate steps, iirc they were mid/high 70°C in p95-avx.
The jump to 5.0 was not that easy. It may be the specific die, it may be the RAM, it may be the budget board coming to its limits here but those settings took 1 week to find out and stabilize and temps the system and I can live with
I knew the vCore needs an UP, just how much would tell me how much headroom I will have left. As it did most smaller lightload benchmarks till 5.2 on my quick&dirty test I want more in the long run ( delid etc.. ). More GHZ but cooler same time.
I was not able to stabilize the IMC at 5G/3600 no matter what I dialed in...and I tested every day on every free hour.
It would maybe hold it 45min but still drop a core or get so hot I had to stop ( 90+ region ). I totally forgot about LLC in those days.
After almost settling with 4.9GHZ as last stable to achieve I went back to the table, took a book and a pen and took NOTES and started again ,based on my 4.8 and 4.9 settings which I saved.
I upped the vCore till 1.330v and reached a point where it would do each and every test but not p95-avx. That would still drop cores...eventually. Usually, those drops I could catch with VCCIO, up it by 1 tiny bit and you have it. Well, it made it better but cores still dropped. Less but it occured. I gradually upped vDIMM and at 1.3992 dropping ALMOST stopped with VCCIO at now 1.30v. I then remembered LLC, upped it step by step and finally, at LLC6, 1 below max. it finally ran it all stable. Temps in p95-avx get hairy but it does not drop, aida64 runs 4h+, gaming etc..no BSODs, no erros, no none-posting upon reboot or coldboot.
Lowering any of those values by any bit results in drops. Making sure no volts are too high I had aida64 running and lowered each one indvidually. Each lowering of any dial by 1-2 steppings caused a crash in aida stability test. That told me I am as low as possible for now and hopefully 1 step above just enough.
I run that setting for about 5 days now, after elaborating it for almost 2 weeks until it runs any test stable and temps are such that I can live with them. Delidding is still needed as when the summer comes I still want to be below 75°C in everyday life computing and gaming. I dont do p95 for a living and HandBrake is the most severe I day, maybe once every 3 month for 3 movies. It's a game rig and sometimes I load as many VMware machines as it can hold just for fun and exploiting new OS's, mainly Linux.
While I type this I have the RAM at 4133MHz @ 17-17-17-36-360-2T/1.40v and it runs through the simple stuff, GeekBench, Passmark etc.. What can I say, lets see what aida64 and p95 think about this
At 5.0/3600 this is already fast and has a very low latency. Let's see if we can top this. In a manner that you can feel it in the cockpit, either TiR5, ViVE or Rift and a roaring MiG-21Bis on take-off.
What I would like to do is try this CPU in the Asus Apex or Formula. I had such bad luck with the M9E and M9ACK that I really would like to know if that much UP in price is that much better and easier in overclocking.
This board is almost less than 1/2 of the price of the M9Formula which I almost opted for ( love that VRM cooling ).
Check Six
After having read many articles that all came to the conclusion that the 7700k overclocks alike across low-2-highend boards
I decided to try this sleek approach and invest the saved money in some fast RAM and other goodies of actual use.
The board gave me problems when I booted it up 1st time with all 4 modules installed, it blocked. Took 1 and 3 out and it booted with 16GB just fine on XMP. I already had the newest Bios on a stick, flashed it, put all RAM back in and it booted all 4. That's OK with me with such brand new stuff, no points lost, also the RAM is NOT on the QVL !
Win10 installation and application installation went without any errors, the machine felt good right away.
I then tested the RAM and CPU and default settings, XMP applied ( with the ALL CORE option active ).
FAILED ALL TESTS ! nothing worked p95, aida64 etc.. they all either dropped cores or aborted due to errors.
I gradually tested vCore, VCCIO ( was already at 1.2v through XMP I guess ) and VCCSA ( already at 1.35v and somewhat alerting ). None of those stopped the erorrs from occuring.
Turned out I had to up the vDIMM to 1.37228v as lowest where it would stop dropping cores or abort aida64. It then ran stock clocked and XMP for 1h each test. The VCCIO and VCCSA have been upped by the Auto setting to 1.20 and 1.35 Volts.
Since I had a rather golden Sandy that ran 5.0GHz and booted till 5.245GHz my expectations were high, my goal was set.
I have to say, I focused mainly on IPC but that much I can say, this CPU offers a lot more than more IPC compared to Sandy Bridge. The overall boost is present, everywhere...if you reach the 5GHz or higher.
Bringing the CPU to stable 5GHz and RAM at XMP3600 seemed impossible for the first week. It was easy to get it to 4.8 by only upping the vCore to 1.30v and leave the rest alone, on auto. No Phase boosting, no LLC ( default is LLC2, not LLC1 btw ). During stress testing random errors occured, and the only thing that helped was raising VSSIO even further to 1.225v, again, everything else as low as possible.
4.9 at first seemed to not need any higher Voltage but again during the 1h tests it failed sometimes so I upped the vCore by 0.01V and could "almost" stabilize it again for 1h runs aida64 and p95 ( both versions, 26.6b3 w/o AVX and 28.1b1 with AVX ). Random errors and differencies when those occured and how they occured let me to raise Phase management to 140%
and PhaseControl to Extreme in Asus UEFI Bios. LLC was left at default 2 and all other voltages dialed back to 1.31vCore, 1.225 IO, 1.35 SA, 1.3728vDIMM. It then ran stable for hours and hours. As this was my jumpboard to 5 I left it priming all night and aida64 all morning. Temps were OK, with the rather big external radiator and HK-IV waterblock I stayed well below
critical temps. I have not taken notes of the temps form those intermediate steps, iirc they were mid/high 70°C in p95-avx.
The jump to 5.0 was not that easy. It may be the specific die, it may be the RAM, it may be the budget board coming to its limits here but those settings took 1 week to find out and stabilize and temps the system and I can live with
I knew the vCore needs an UP, just how much would tell me how much headroom I will have left. As it did most smaller lightload benchmarks till 5.2 on my quick&dirty test I want more in the long run ( delid etc.. ). More GHZ but cooler same time.
I was not able to stabilize the IMC at 5G/3600 no matter what I dialed in...and I tested every day on every free hour.
It would maybe hold it 45min but still drop a core or get so hot I had to stop ( 90+ region ). I totally forgot about LLC in those days.
After almost settling with 4.9GHZ as last stable to achieve I went back to the table, took a book and a pen and took NOTES and started again ,based on my 4.8 and 4.9 settings which I saved.
I upped the vCore till 1.330v and reached a point where it would do each and every test but not p95-avx. That would still drop cores...eventually. Usually, those drops I could catch with VCCIO, up it by 1 tiny bit and you have it. Well, it made it better but cores still dropped. Less but it occured. I gradually upped vDIMM and at 1.3992 dropping ALMOST stopped with VCCIO at now 1.30v. I then remembered LLC, upped it step by step and finally, at LLC6, 1 below max. it finally ran it all stable. Temps in p95-avx get hairy but it does not drop, aida64 runs 4h+, gaming etc..no BSODs, no erros, no none-posting upon reboot or coldboot.
Lowering any of those values by any bit results in drops. Making sure no volts are too high I had aida64 running and lowered each one indvidually. Each lowering of any dial by 1-2 steppings caused a crash in aida stability test. That told me I am as low as possible for now and hopefully 1 step above just enough.
I run that setting for about 5 days now, after elaborating it for almost 2 weeks until it runs any test stable and temps are such that I can live with them. Delidding is still needed as when the summer comes I still want to be below 75°C in everyday life computing and gaming. I dont do p95 for a living and HandBrake is the most severe I day, maybe once every 3 month for 3 movies. It's a game rig and sometimes I load as many VMware machines as it can hold just for fun and exploiting new OS's, mainly Linux.
While I type this I have the RAM at 4133MHz @ 17-17-17-36-360-2T/1.40v and it runs through the simple stuff, GeekBench, Passmark etc.. What can I say, lets see what aida64 and p95 think about this
At 5.0/3600 this is already fast and has a very low latency. Let's see if we can top this. In a manner that you can feel it in the cockpit, either TiR5, ViVE or Rift and a roaring MiG-21Bis on take-off.
What I would like to do is try this CPU in the Asus Apex or Formula. I had such bad luck with the M9E and M9ACK that I really would like to know if that much UP in price is that much better and easier in overclocking.
This board is almost less than 1/2 of the price of the M9Formula which I almost opted for ( love that VRM cooling ).
Check Six
Last edited: