For those you who are not into the tech guru joke stash, it's Coffee Lake.

6 hours ago, joemiller said:

Very well, Bob. Thanks for the clarification. Hopefully, we can overclock all cores on the 9900k or 9700k to near 6.0Ghz. I guess we won't need HT on as that would create more heat.

Are there any official info of when they will be available for purchase? 

The 9th Gen CPUs release from embargo on Oct 19th (this Friday). 

Motherboards featuring the new Z390 chipset have been showing up on the shelves for a couple weeks now.  I've got a Z390 Asus Maximus XI Hero and 32 GB of GSkill 3600MHz CAS 15 RAM already on the way...hope to see the i9-9900K CPU ship out Friday.  I have high hopes of seeing what doubling my core count to 8 physical cores while running in the 5.3-5.5 GHz range might do for P3Dv4.  I have a fairly aggressive water loop (two externally-mounted 360mm rads just for the CPU) in a basement man cave that stays at ~68 deg F (20 deg C) year round.  Might actually find cause to dial up the fans and the pump speed from their lowest settings for this one.

@Rob--do you know if the new Covfefe Lake R CPUs 🙂 can have their cores independently clocked like the -X SKUs do?

Regards

6 hours ago, Rob Ainscough said:

.

For P3D performance I've seen ZERO benefit from running ALL cores as high as they can go, but I have seen a benefit from running 1 core (the main sync thread) higher while the other cores are still high but about 200-300Mhz lower ... this increases stability, reduces heat, and doesn't have any detriment to P3D performance.

P3D does use all cores but in most of my testing I see about 60-80% utilization across the other cores while the primary main thread is around 90-98%, but the main sync thread will always without except bare the grunt of work as it has to assemble the final frame render in a virtual world that is "bound to time" ... and that's very important distinction "bound to time"

However, seeing as the Z390 chipset and CPU are relatively new, there will most likely be several EFI versions that need to happen before all the bugs are worked out.

NOTE: this is still NOT the processor that has hardware fixes for Spectre and Meltdown, I believe those CPUs are a couple of generations away.

Cheers, Rob.

Ok, if I understand correctly, for P3D purposes, there is no need to OC all cores to the highest possible? Meaning if P3D mainly uses the first core (core 1- or core 0-  not sure which is the main one P3D uses) , it's best to OC cores individually by OC the main core, as high as possible, while the others can remain a bit lower? Or would you recommend to OC all cores equally for sim purposes? 

56 minutes ago, joemiller said:

Ok, if I understand correctly, for P3D purposes, there is no need to OC all cores to the highest possible? Meaning if P3D mainly uses the first core (core 1- or core 0-  not sure which is the main one P3D uses) , it's best to OC cores individually by OC the main core, as high as possible, while the others can remain a bit lower? Or would you recommend to OC all cores equally for sim purposes? 

I think, though, that prosumer-level non Extreme-edition CPUs (e.g. below the $1000 price point) can't independently manage different clock multipliers for each of the cores.  As I read it, you can set different multipliers for each level of intel's turbo-boost (v2 for a prosumer chip, v3 for the extreme editions).  In the case of my i7-7700K the boost thresholds are simply for one, two, three, or four cores under load...in the new 8-core chips I think it's probably set for two, four, six or eight loaded cores.  But I interpret the tech documentation to say that the mobo chipset sets only one multiplier used across the entire CPU.  So if I set the multiplier to be 50 (x100 MHZ = 5.0 GHz) for one CPU, 49 for two, 48 for three, and 47 for four, if three of the four CPUs are under load the controller sets the clock mult to 48 (per the three-core load threshold) across the entire CPU, and not 50 for core 0, 49 for core 1, etc...

I think the extreme edition processors, which have ITBM v3.0, might be able to set separate clock mults for their cores, but I'm not certain about that.

Also, in the interest of avoiding oversimplification, there are other other considerations beyond the number of active CPUs that are used to determine processor speed, such as current to the CPU, instantaneous and average temperature, etc.  But assuming good cooling, a stable power supply with appropriately balanced voltages set, etc, we manage the overclock with the multiplier settings.

So the bottom line is that if I want 5.3 GHz on the main thread, then I have to run the entire chip with a mult of 53 set for all core loading scenarios in ITBM v2.0

Regards

 

Quote

I think, though, that prosumer-level non Extreme-edition CPUs (e.g. below the $1000 price point) can't independently manage different clock multipliers for each of the cores.  

 

You can for me. 8700K Maximus X Code. The first core must be the highest frequency though. The other cores I can set as desired.

Although Windows 10 then balances the load. I believe the workaround is to set the affinity in Windows.

Max or Westman will probably have more knowledge re this as I don't usually OC per core.

Edited October 17, 20187 yr by martin-w

23 hours ago, martin-w said:

 

You can for me. 8700K Maximus X Code. The first core must be the highest frequency though. The other cores I can set as desired.

Although Windows 10 then balances the load. I believe the workaround is to set the affinity in Windows.

Max or Westman will probably have more knowledge re this as I don't usually OC per core.

I've never bothered with the feature, though flight sim is pretty much the perfect application for it.  Maybe I'll give it a shot when I can get my hands on the Z390 board I really want (EVGA Dark).