K skus irrelevant?

[GloriousPain]

Was watching Linus' Skylake OCing Guide on vessel after coming home from a small vacation, and I thought back to Linus' 12 and 18 Core Xeon Videos, and thought if baseclock overclocking is more of a thing on skylake, at least for the 6700k and 6600k  if this carries over to say Xeons and Non-K skus won't it make the K skus much less relevant? I mean 6700k? OC using multiplier and base clock to hit 4.7Ghz at the $350ish price won't be terribly impressive if say a Xeon 1230 V5 for $250 using just baseclock could hit 4.5Ghz... Or is this likely to not carry over and Baseclocks are still not likely to break 105 or 110 for anything non K sku?

[DarkBlade2117]

Xeons donn't expect to go over 4.0 GHZ.

The highest i3 baseclock is 3.9Ghz ( i3-6320 ) and Xeons do not tend to go over anything an i3 has to offer in terms lf base clock.

[W-L]

Was watching Linus' Skylake OCing Guide on vessel after coming home from a small vacation, and I thought back to Linus' 12 and 18 Core Xeon Videos, and thought if baseclock overclocking is more of a thing on skylake, at least for the 6700k and 6600k  if this carries over to say Xeons and Non-K skus won't it make the K skus much less relevant? I mean 6700k? OC using multiplier and base clock to hit 4.7Ghz at the $350ish price won't be terribly impressive if say a Xeon 1230 V5 for $250 using just baseclock could hit 4.5Ghz... Or is this likely to not carry over and Baseclocks are still not likely to break 105 or 110 for anything non K sku?

 

Baseclock overclocking is sometimes a hit and miss since it changes the speed of everything on the system so it can possibly mess it up sometimes, usually you can't really go over a base clock of 105 giving only about a 5% boost where unlocked processors can do a lot more.

[GloriousPain]

Xeons donn't expect to go over 4.0 GHZ.

The highest i3 baseclock is 3.9Ghz ( i3-6320 ) and Xeons do not tend to go over anything an i3 has to offer in terms lf base clock.

Darn thanks; Figured I was missing something or being terribly optimistic. Guess I was. 

[TheProfosist]

That all depends on whether or not the CPU takes to different base clocks. The first one to do this well in a while has been Haswell-E prior to that your looking at first gen i7 enthusiast parts. At least now it a bit easier since the memeory clock is separate. That used to cause so many headaches.

[TheProfosist]

Darn thanks; Figured I was missing something or being terribly optimistic. Guess I was.

Other thing is that Xeon's are locked down pretty hard I dont even know if you can change the base clock at all on one.

[GloriousPain]

Baseclock overclocking is sometimes a hit and miss since it changes the speed of everything on the system so it can possibly mess it up sometimes, usually you can't really go over a base clock of 105 giving only about a 5% boost where unlocked processors can do a lot more.

Linus mentioned this in the video apperently on skylake somethings that were previously attached to CPU strap/Baseclock aren't anymore eliminating ALOT of those issues, Linus did get a BCLK of 200 Mhz in the video, but that was with the multiplier crippled to 23 achieving effectively 4.6Ghz

[GloriousPain]

That all depends on whether or not the CPU takes to different base clocks. The first one to do this well in a while has been Haswell-E prior to that your looking at first gen i7 enthusiast parts. At least now it a bit easier since the memeory clock is separate. That used to cause so many headaches.

Not sure how to respond to this

Other thing is that Xeon's are locked down pretty hard I dont even know if you can change the base clock at all on one.

Linus has done so in the past, which as I said in OP is what gave me the thought he did it to, 12 AND 18 Core xeons... if those aren't locked I don't see why a Quad Core would be...

[TheProfosist]

Not sure how to respond to this

Linus has done so in the past, which as I said in OP is what gave me the thought he did it to, 12 AND 18 Core xeons... if those aren't locked I don't see why a Quad Core would be...

Basically with multiplier overclocking its very easy to get a fairly decent stable OC. With BClock, especially prior to the independent memory bclock it is a pain to get something stable. The last CPU I did it on was the i7-3820 which you actually got a decent bit out of at its max or second highest multiplier.

yes on newer chips when you BCLock goes up the allowed multiplier goes down. This is needed especially since many kits of DDR4 ram prefer you run at a BClock of 125.

Also I think they are lock to preset intervals so like 100, 125, 150, etc

[TheProfosist]

I guess I could go into how you use BClock in OCing since the introduction of the the adjustable multiplier if you would like. P

Prior to that all you had was BClock.

[airdeano]

another item to watch out for is that the PCIe x16 is still controlled through the

CPU lanes and moving the BLCK clock forward also increase the PCIe x16/x8

frequency. so like moving the slider on GPU overclocking, moving the BLCK

freq multiples the bus freq the same way. and this is why most BLCK OC fail

boot/POST is the GPU might not handle the extra bus speed and GPU error

is not suspected. something you cannot fix until booted or GBIOS is tweaked

to handle the extra multiplier (125/150/175/200).

[TheProfosist]

another item to watch out for is that the PCIe x16 is still controlled through the

CPU lanes and moving the BLCK clock forward also increase the PCIe x16/x8

frequency. so like moving the slider on GPU overclocking, moving the BLCK

freq multiples the bus freq the same way. and this is why most BLCK OC fail

boot/POST is the GPU might not handle the extra bus speed and GPU error

is not suspected. something you cannot fix until booted or GBIOS is tweaked

to handle the extra multiplier (125/150/175/200).

I just imagine GPU's going through the DMI link and it gave me goosebumps. (the bad kind of goosebumps, not the good kind)

[W-L]

Linus mentioned this in the video apperently on skylake somethings that were previously attached to CPU strap/Baseclock aren't anymore eliminating ALOT of those issues, Linus did get a BCLK of 200 Mhz in the video, but that was with the multiplier crippled to 23 achieving effectively 4.6Ghz

 

hmm that's different just watched the video, if the base clocks have no effect on the PCI-E it would be a nice option but it would still affect other parts of the system just less of it. I see it as more of a secondary option for overclocking more than a sure fire way of going at it, one thing that would do is give larger intervals between frequency jumps instead of the regular 100Mhz which TBH I'm not really sure if that will have a positive or negative effect but I guess time will tell as more users try it.

[airdeano]

the data transfer/latency through the chipset at DMI 3.0 is equal to the

PCIe GEN III 8GT/s, but latency (chipset controller then to the CPU vs

CPU direct) will bottleneck the GPU. unless a pipeline to the CPU can be

tunneled, DMI 3.0 will be reduced to low-level transfers

(data storage/computations or short burst read/write).

[TheProfosist]

the data transfer/latency through the chipset at DMI 3.0 is equal to the

PCIe GEN III 8GT/s, but latency (chipset controller then to the CPU vs

CPU direct) will bottleneck the GPU. unless a pipeline to the CPU can be

tunneled, DMI 3.0 will be reduced to low-level transfers

(data storage/computations or short burst read/write).

Thing is you have all that other stuff including the SATA controller going through that DMI link.