Finally took the upgrade plunge again, but having a few issues

[HumblePie]

So I decided it was time to upgrade due to deals I found. Picked up an I7 6700K combo with an MSI z170a-g45 board. Also snapped up some G.Skill DDR4 3200 Ripjaws as a 32GB Kit (16GBx2).

Also was doing the same for a friend so basically ordered doubled of everything. Except he cheaped out a bit on the memory and went with standard 2133 DDR4 memory Ripjaws.

I was hoping to get these chips past 4.4Ghz, but can't seem to do it with either. They both run fine at 4.4Ghz at 1.35 volts and not a lick past that no matter what juice I give them. Not sure if it's the CPUs or the boards.

On top of that, the memory has been a bit of an issue. My friends does not go past 2133 which surprised the hell out of me. Regardless of the board used. Mine won't go past 3100 speeds. Which is annoying the hell out of me completely. This was with playing mix and match on the parts to achieve the best results. Didn't matter. Same results no matter which part I used. I also did update the BIOS on both boards to the current 2.5 version put out by MSI. I also try upping the voltage on the SA and IO all the way up to 1.3v and got not difference in the outcome. I however can get to PC3800 speeds on my chips if I run either chip by itself no problem. If I put them both in though in dual channel configuration, then I'm capped. I don't think the problem here is with the memory.

So basically my friends PC is humming along fine now at 4.4Ghz with memory at 2133 and mine is at 4.4Ghz at 3100 memory speeds. Temps are about where I see every where else when running prime95 v289. Spikes go up to about 71C using HWMonitor.

I was really hoping to get 4.7Ghz or more out of this setup as I was already pushing 4.4Ghz from my previous 2600K chip and was getting 4.6Ghz from his 2500K previous chip. I know the architecture allows for a bit more performance per clock speed over the older chips we have, but many bench marks I've seen for games I care about tends to give the 6700K a 10%-15% increase in performance for the same clock speeds. Which is why I was hoping to knock out a bit more juice from these newer chips.

Caveat, I was hoping to keep hyperthreading and C-State intact as I had that on my previous chip with no problem. I know I could probably eek out some more overclocking on these chips if I disabled those I bet, but my personal preference of buying a 6700K over a 6600K is because of the hyperthreading.

Google searching hasn't yielded much insight into trying to get further in my overclocking endeavor with these chips. Anyone on these boards have any thoughts?

 

[Carfax83]

Well you left out the most important bit of info. What are you using to cool the CPU?

Also, whilst I myself don't have a Skylake CPU, that voltage seems kind of high just to hit 4.4ghz. My 5930K is overclocked to 4.4ghz and it only uses 1.21-22v with adaptive voltage.

Are you overclocking via multiplier or by bclk strap?

As for the memory, it's probably your individual CPU. Some IMCs can tolerate higher RAM frequencies than others. But you shouldn't really exceed 1.2v for SA if you're on air.. Actually, I wouldn't go past 1.1v personally..

Also 16GB DIMMs are quite large, and stress the IMC even more. Although I thought that Skylake supposedly had an improved IMC that could run higher frequencies compared to Haswell-E :\

And the 6700K is a great upgrade from a 2600K. To get a perspective, your current CPU is the equivalent of a 2600K at well over 5ghz..

 

[TidusZ]

Don't disable hyperthreading, not worth it

 

[mikeymikec]

PC3200

This always throws my brain through a loop

 

[HumblePie]

Well for general oc info I was following this site info http://www.overclock.net/t/1570313/skylake-overclocking-guide-with-statistics

As well as a few others. Also, both chips were basically doing the same thing when it came to overclocking. Could only hit 4.4 stable when at 1.35v and couldn't go over that regardless of juice given. Overclocking so far that I did has been by mmultiplier only.

As for cooling friend has a hyper212 air cooler and I am using a corsair h110i gtx cooler. Tried various pastes I have which include original as3, as5, and formula 7. Wasnt seeing much difference at idle or load.

 

[HumblePie]

This always throws my brain through a loop

Why? It is what is listed on the memory.

 

[Yuriman]

As others have said, you're running into memory controller limitations. Skylake was only validated to run 2133, so you're running the memory controller with around a 50% overclock already. It's not uncommon to hear of people who can't get more than 2800 out of their CPUs.

Pretty sure the RAM is advertised as "DDR4-3200". "PC3200" was used back in the early days to show that RAM had 3200MB/s of bandwidth. DDR4 3200mhz would be something like PC12800 (just made the number up, I forgot how to calculate it).

 

[HumblePie]

As others have said, you're running into memory controller limitations. Skylake was only validated to run 2133, so you're running the memory controller with around a 50% overclock already. It's not uncommon to hear of people who can't get more than 2800 out of their CPUs.

Pretty sure the RAM is advertised as "DDR4-3200". "PC3200" was used back in the early days to show that RAM had 3200MB/s of bandwidth. DDR4 3200mhz would be something like PC12800 (just made the number up, I forgot how to calculate it).

ahh you are right on the number. PC4 256000 is what is the PC rating on it. Listed as 3200 still.

This is the memory I am specifically using.

http://www.newegg.com/Product/Produ...m_re=g_skill_ddr4_3200-_-20-232-091-_-Product

It says rated for 3200 bandwith on the z170 chipset platform. There are some rated for even higher speeds along with full 64GB kits. Such as this kit. http://www.newegg.com/Product/Produ...m_re=g_skill_ddr4_3200-_-20-232-208-_-Product

I however did read of the motherboards tested for these chips and the MSI board I have isn't one of them.

http://www.gskill.com/en/product/f4-3200c16d-32gvk

Guessing I'm going to have to sell this motherboard and get another then. Sucks.

 

[ShintaiDK]

As others have said, you're running into memory controller limitations. Skylake was only validated to run 2133, so you're running the memory controller with around a 50% overclock already. It's not uncommon to hear of people who can't get more than 2800 out of their CPUs.

Pretty sure the RAM is advertised as "DDR4-3200". "PC3200" was used back in the early days to show that RAM had 3200MB/s of bandwidth. DDR4 3200mhz would be something like PC12800 (just made the number up, I forgot how to calculate it).

The memory controller is actually validated above 4000Mhz.

 

[ShintaiDK]

ahh you are right on the number. PC4 256000 is what is the PC rating on it. Listed as 3200 still.

This is the memory I am specifically using.

http://www.newegg.com/Product/Produ...m_re=g_skill_ddr4_3200-_-20-232-091-_-Product

It says rated for 3200 bandwith on the z170 chipset platform. There are some rated for even higher speeds along with full 64GB kits. Such as this kit. http://www.newegg.com/Product/Produ...m_re=g_skill_ddr4_3200-_-20-232-208-_-Product

I however did read of the motherboards tested for these chips and the MSI board I have isn't one of them.

http://www.gskill.com/en/product/f4-3200c16d-32gvk

Guessing I'm going to have to sell this motherboard and get another then. Sucks.

You can just down clock the memory to 2133. If it doesn't change OC, its not due to memory/mobo combination.

Did you overclock the uncore too?

 

[Yuriman]

Thanks Shintai.

And yeah, lower your memory clocks and see if it makes any difference. And if not, I wouldn't be concerned that you're only getting 3100/3200mhz; the performance difference is negligible, and certainly not worth the time and effort of replacing the board.

Heck, 4.7ghz is less than 7% faster than 4.4. The difference will be completely imperceptible even in entirely CPU-bound scenarios.

 

[Majcric]

Op, What is your BCLK? 102.10 And are you using the XMP profile? I think your memory timings are why you can't get past 4.4.

If you want to enable the XMP profile out of the box, exchange your memory for some with timings of say 15-15-15-35 at around 3000/3200. The memory with higher timings seems to sffect the BCLK as result raises system agent and VCCIO voltage. So the system is already pushed to its max.

My chip will only do around 4.6 at around 1.35-1.36 or I can put it at 4.5 and not touch the voltage. Guess which choice I take?

 

[HumblePie]

Bclk has not been touched. At 100 default still.

XMP for the chips are shown in the link above. Basically xmp is 3200 16 latency for most. The base 3200 settings if not using xmp or manual adjustments are 18s. So looser timings.

 

[Deders]

Don't want to get you down on your new hardware, but from what I've seen around launch, MSI boards don't seem to be able to overclock as high as Gigabyte or Asus. I would have hoped they would have sorted this out by now.

I remember a few gens back when the MSI z'x'7-g45 would outperform it's bigger brothers.

 

[Carfax83]

Did you install the DIMMs in the correct slots? I know with quad channel boards, if you don't install the DIMMs in the correct slots, it can cause a lot of stability and performance problems.

Also, did you make sure your motherboard is actually certified by G.Skill to run that memory at that frequency?

*Edit* OK I checked G.Skill's QVL for MSI with your RAM kit, and your motherboard wasn't listed.

You can look at it here.

 

[HumblePie]

Did you install the DIMMs in the correct slots? I know with quad channel boards, if you don't install the DIMMs in the correct slots, it can cause a lot of stability and performance problems.

Also, did you make sure your motherboard is actually certified by G.Skill to run that memory at that frequency?

*Edit* OK I checked G.Skill's QVL for MSI with your RAM kit, and your motherboard wasn't listed.

You can look at it here.

They are in the right slots and said earlier that I checked that the mobo wasnt supported by gskill

 

[Carfax83]

Then I'm fresh out of ideas. Maybe you just got a dud? It's been known to happen.

Although like I said before, a 6700K at 4.4ghz is still an incredibly fast CPU. It would take a 2600K at probably 5.2ghz or more to match it, especially in CPU limited circumstances..

As for the memory, I think that perhaps a future UEFI release may ameliorate the problem.

 

[HumblePie]

Then I'm fresh out of ideas. Maybe you just got a dud? It's been known to happen.

Although like I said before, a 6700K at 4.4ghz is still an incredibly fast CPU. It would take a 2600K at probably 5.2ghz or more to match it, especially in CPU limited circumstances..

As for the memory, I think that perhaps a future UEFI release may ameliorate the problem.

But two different ones experiencing the exact same behavior on the boards regardless of how I mix and match them?

Thinking the board type just sucks for overclocking.

 

[Carfax83]

But two different ones experiencing the exact same behavior on the boards regardless of how I mix and match them?

Thinking the board type just sucks for overclocking.

Looked for some reviews of the mobo, and the only professional one I could find was by Tech Report.

In their overclocking test, they got 4.5ghz. On an Asus board with the same CPU and cooling, they got 4.6ghz. But one review isn't enough to come to any sort of conclusion on a motherboard's overclocking capabilities..

It seems like it's heavily geared towards budget shoppers despite having some enthusiast features, so MSI may have cut corners to lower the cost compared to their more expensive enthusiast parts..

I'm just postulating here. To be honest, the most important factor in how well a CPU overclocks, is the CPU itself. A good motherboard will just make it easier to extract additional performance, but it can't make a poor CPU shine.

 

[ShintaiDK]

Then I'm fresh out of ideas. Maybe you just got a dud? It's been known to happen.

I am kinda in on this if everything else got tested. While 4.4Ghz would be extremely low, it can happen.

 

[Carfax83]

I have just one more suggestion. Ever since I built my newest rig, I was never really able to run my TridentZ DDR4 3200 at full speed. Sure I could boot into windows and run a few benchmarks, but when it came to gaming, it was crash central. Also Windows would eventually BSOD or error out on me.

Lately though after updating my UEFI to the latest version, I tried once again to see if I could run DDR4 3200 speed using XMP. It turns out that I couldn't.....again.

But after doing some research, I had one more trick up my sleeve to play. I disabled XMP, and entered the settings manually. Lo and behold, I can now run at DDR4 3200! :thumbsup:

I don't know what it was about G.Skill's XMP settings, but after entering my memory sub timings and voltages manually, I can now run at DDR4 3200 with stability!

Here's my Aida64 memory and cache benchmark run at DDR4 3200, with the CPU at 4.4ghz and the uncore at 4ghz. So my point is, you might try disabling XMP and entering the settings manually. It might allow you to run your RAM at the frequency it's meant to run at..

 

[HumblePie]

Looked for some reviews of the mobo, and the only professional one I could find was by Tech Report.

In their overclocking test, they got 4.5ghz. On an Asus board with the same CPU and cooling, they got 4.6ghz. But one review isn't enough to come to any sort of conclusion on a motherboard's overclocking capabilities..

It seems like it's heavily geared towards budget shoppers despite having some enthusiast features, so MSI may have cut corners to lower the cost compared to their more expensive enthusiast parts..

I'm just postulating here. To be honest, the most important factor in how well a CPU overclocks, is the CPU itself. A good motherboard will just make it easier to extract additional performance, but it can't make a poor CPU shine.

Well found out part of the reason for the OC issue I was having with the chip I currently have is that the MSI board has very low auto short and long duration limit values. Took me awhile to find those settings too. Bumped those values up to max and was able to push the chip a bit further.

Right now the limit seems to be 4.6Ghz. I can do 4.6Ghz stable with prime95 for 5 hours so far at 1.43v. If I drop the volts any it crashes. I can't get higher than 4.6Ghz though. Tried pushing the multi to 47 and increased the volts up to 1.48. Still was crashing at that much juice. Didn't want to push it further than that. So looks like it's a 4.6Ghz max chip. Which seems to to be the max of the crappier chips from the link I posted above. My temps weren't even all that bad. 78C max at load running prime95.

I did watch a couple of MSI OC vids and read some threads. This board is the cheaper board and much worse at overclocking than the better ones. It does make a difference. With better boards people with the same chips were hitting higher values with less volts. Meaning these were people with multiple boards to test with but were using the same chip to test the various boards. The better boards with more phases and better built for overclocking would push the same chip to higher multi's with less voltage. So the board does seem to be a pretty big influence here.

Thinking of ordering a better board then and passing this one on to someone else that isn't looking to OC that far.

I'll have to see what the other board+chip combo that is in my friend's comp can do now. I already gave him the other set and couldn't play with both sets for much longer than a day the other day.

Also from reading that thread, it seems pretty apart there is a fine line between the good, the average, and the bad chips using good boards.

Exceptional chips can do 4.8Ghz at 1.4v, good chips do 4.7Ghz, average do 4.6Ghz, and bad do 4.5Ghz at 1.4v. Right now mine is in the bad category but that since I'm not using the better OC'ing boards, that is an influence. Mine is probably an average chip on a good board.

 

[HumblePie]

I have just one more suggestion. Ever since I built my newest rig, I was never really able to run my TridentZ DDR4 3200 at full speed. Sure I could boot into windows and run a few benchmarks, but when it came to gaming, it was crash central. Also Windows would eventually BSOD or error out on me.

Lately though after updating my UEFI to the latest version, I tried once again to see if I could run DDR4 3200 speed using XMP. It turns out that I couldn't.....again.

But after doing some research, I had one more trick up my sleeve to play. I disabled XMP, and entered the settings manually. Lo and behold, I can now run at DDR4 3200! :thumbsup:

I don't know what it was about G.Skill's XMP settings, but after entering my memory sub timings and voltages manually, I can now run at DDR4 3200 with stability!

Here's my Aida64 memory and cache benchmark run at DDR4 3200, with the CPU at 4.4ghz and the uncore at 4ghz. So my point is, you might try disabling XMP and entering the settings manually. It might allow you to run your RAM at the frequency it's meant to run at..

The problem is that I tried running looser than the timings already.

MSI I can try the XMP setting or I can use the base 3200 setting for memory speed. If I use the base setup I end up with 18-18-18-48-2T timings where my memory is supposed to be at 16-18-18-38-2T. It couldn't handle 3200 speeds with the looser timings at all.

 

[Ketchup]

I know boards used to have a pretty good impact on what kind of overclock you could get. A couple examples are a Q6600 that would stop about 200 Mhz over on an Asus board and a full 1 GHz over on a Gigabyte board (actually 1.2), Going back a bit, there was an Epox board that got a 200 MHz better overclock on an updated BIOS.

And some chips just don't overclock well. The CPU in my sig is one of them. It can barely go over stock (4.4). Of course, I barely care about overclocking anymore, so it works out.

So, enjoy what you have, and enjoy the high performance and energy savings of stock clock at stock voltage.

 

[Majcric]

Op, I'm glad to hear you somewhat got your issue sorted out. I've had my hands on a few boards and i7 6700k's, Asus seems to squeeze out those extra MHz the best if your considering another board.


The folks running the higher clocks on Skylake aren't afraid of voltage that's for sure. Personally I'm not comfortable going more than 1.35v. With the rate of IPC improvements I may be hanging on to this chip a while, thus why I don't want to push too hard. I have to admit there is a curiousity of what my chip would do at 1.4v and over. My prediction is 4.7.

With all that said I have to ask at 1.43v, what kind of temps are you seeing in the latest version of prime 95 small FFT. Say at just a 30min to hour run?