Hyperthreading off works great!

[Cruachan 705]

1 hour ago, Mace said:

I'd rather have 8 physical cores than six though.  :)

Hi Rhett, 

I do, and those 2 spare cores/4 LPs are dedicated to running some 3rd Party Addons thereby avoiding potential contentions with the main App.

Regards,

Mike

[glider1 90]

Does anyone know whether there is a trade-off between the rate of scenery loading and traffic levels depending on how many cores you want P3D4.4 to use?

The main thread off-loads jobs to the other cores. The more cores there are the busier the main thread is doing the offloading. I've seen this firsthand when I ran P3D4.4 on 4 cores verses 8 cores (HT off HT on). With HT off P3D4.4 sees only 4 cores and the main thread is about 10% less busy.

If the main thread is less busy managing less cores, does that mean that you can run more traffic without smoothness degradation at the expense of slower loading scenery?

Thanks

[Bert Pieke 7,585]

Good question.. but who knows exactly.. and each new P3D release is different in this regard..

As you provide more cores to P3D, there is always the risk of the "too many cooks in the kitchen" syndrome.. test it out on your own system!

As the above discussion shows, at least some of us have found that 6 cores is a sweet spot.. but YMMV

Edited January 3, 2019 by Bert Pieke

[fghdgdfdfgfgf 111]

9 hours ago, Cruachan said:

Hi Ken,

However, the story is somewhat different for Single Thread Performance, which, as we know, is important for Prepar3D as applied to Core 0:

8600K (2,517/3.6) x 5.3 = 3705

5960X (1,996/3.0) x 4.6 = 3060

I have gone to 4.7Ghz stable for a couple of years but ‘only’ on 4 cores/8 LPs, remaining cores at 4.6Ghz. I note that the 8600K does not support hyperthreading, but it has 6 Cores which, as I’ve demonstrated, seems to be the optimal count for Prepar3D.

Regards,

Mike

Here you go Mike  ...  big smile!

...and in case were wondering what 4266 looks like:

 

and a bonus 4300 mhz pict: ...might try it later. :-)

 

-Cheers

 

And I dare anyone to tell me mem doesnt make a difference on a already fine running machine. :-)

@Mace No stutters .....and check out my fan speeds in the above graphic .....the cpu fans are Noctua nf-s212a pushing air through the radiator ...yup ....no noise. :-)

[vgbaron 3,668]

1 hour ago, FunknNasty said:

And I dare anyone to tell me mem doesnt make a difference on a already fine running machine. :-)

Well, since you asked.......   

Actually I agree but the important part is already fine running machine. The memory should be the last item on your clocking list - get everything else stable and smoking and then start with the memory.

i think many people scan through clocking guides and just latch on to a single item rather than following the whole process. 

Vic

[newtie 533]

10 hours ago, Bert Pieke said:

Which one more time supports the notion that for those of us who are still happily flying with quad Haswell processors, like my 4770K@4.3, an affinity mask that provides 6 cores for P3D to run on, is indeed a good choice 

4770k @ 4.4, AM=250 here. Works perfect.

We might be related 😁

[lownslo 592]

2 hours ago, FunknNasty said:

 

Hiya Ken... just curious why you didn't show the most important CPU-Z page... the page under the "BENCH" tab.  But seeing as how you threw the $100 challenge out there, perhaps you shouldn't.

Greg

[fghdgdfdfgfgf 111]

10 hours ago, lownslo said:

Hiya Ken... just curious why you didn't show the most important CPU-Z page... the page under the "BENCH" tab.  But seeing as how you threw the $100 challenge out there, perhaps you shouldn't.

Greg

Tough crowd! :-)

 

 

[fghdgdfdfgfgf 111]

22 hours ago, Cruachan said:

http://www.cpu-world.com/Compare/108/Intel_Core_i5_i5-8600K_vs_Intel_Core_i7_Extreme_Edition_i7-5960X.html

https://m.cpubenchmark.net/

Benchmarks (PassMark - CPU Mark)

8600K (12,775/3.6 x 5.3 = 18,807

5960X (15,948/3.0 x 4.6 = 24,453

However, the story is somewhat different for Single Thread Performance, which, as we know, is important for Prepar3D as applied to Core 0:

8600K (2,517/3.6) x 5.3 = 3705

5960X (1,996/3.0) x 4.6 = 3060

I have gone to 4.7Ghz stable for a couple of years but ‘only’ on 4 cores/8 LPs, remaining cores at 4.6Ghz. I note that the 8600K does not support hyperthreading, but it has 6 Cores which, as I’ve demonstrated, seems to be the optimal count for Prepar3D.

Regards,

Mike

Mike, not that we can measure P3Dv4.4 in any meaningful way, but I'd place a fun bet that my setup is 50% faster than what you have.  Oh, and those Passmark scores drive me crazy ----I do like the cpu-world link. ...thanks  

Oh, what cache speed do those things (5960x) run at?  I was shocked at what Rob and his 79xxxX was running at.

[westman 257]

3 hours ago, FunknNasty said:

Mike, not that we can measure P3Dv4.4 in any meaningful way, but I'd place a fun bet that my setup is 50% faster than what you have.  Oh, and those Passmark scores drive me crazy ----I do like the cpu-world link. ...thanks  

Oh, what cache speed do those things (5960x) run at?  I was shocked at what Rob and his 79xxxX was running at.

not read the tread only the last pages , but in my world the 8600k is superior to the old tractor 5960x have one i know , have  a good 8700k @5.5 ht off 4600mhz c17  but run my sim with HT off with a 7920x @5.1 ghz 4x8gb 4000mhz c16 32gb  disabled 2 cores to get more shared cashe 10 cores active, prefer this over the 5.5ghz 8700k 

passmark  is trashmark not a really  good one 

Edited January 3, 2019 by westman

[lownslo 592]

12 hours ago, FunknNasty said:

Tough crowd! :-)

Naw, just bustin' your chops (you gave yourself plenty of wiggle room by claiming only "99%" 🍻).  Thanks for posting your "Bench" screen... looks like you've got it dialed in nicely.

Seriously, though, while many "real" benchers may not put much value in CPU-Z, it could be useful to us simmers.  It would remove the variable of a complicated and demanding flight simulator, and instead just show our system's single and multi-thread performance.  Since many of us use the same class of CPU's this could be helpful to those troubleshooting their overclock as well as system config.  Also, if a system posted a good CPU-Z score those folks trying to help someone with a problem could rule out major issues with the hardware/OS and move to troubleshooting the sim.  Would save a lot of time knowing how a system benches in such a simple and easy-to-use program.

Cheers,

Greg

[Guest]

On 1/1/2019 at 5:11 AM, Ron Attwood said:

It says on the front of my computer 'Antec' and I have an Acer monitor. But I can't see BIOS anywhere.

They're called UEFI these day ... most PCs built within the last 5-7 years will most likely be UEFI (Apple's desktop line was early adopter of EFI).  A PC that comes from a vendor like Dell, HP and is not a custom build, may not have access to the UEFI ... OR there might be a special key sequence you need when booting the PC.

In your situation, "Antec" sounds like the case manufacturer not the actual "vendor" ... so I'm guessing you have some type of custom build or a supplier that makes custom builds?

For most custom PC's, during the boot process hit the DEL key or the F2 key (you may need to hit this key repeatedly) ... this will load up your UEFI/BIOS and you can then turn HT On/Off from there.

Hyper-threading is Intel's approach to increasing the number of independent instructions in the pipeline and REQUIRES OS support.  It's important to note that HT does NOT duplicate main execution resources ... HT increases system latency (not good), can increase cache thrashing, and increases heat and power consumption.  How effective HT is varies significantly by application and overall performance of all the sub-systems from RAM, disk I/O, cache sizes, CPU frequency etc.

For all the processors I've used HT OFF and higher CPU/cache frequency produces better results relative to P3D.  But it really depends on your "Complete" set of hardware (MB, GPU, DDRx, CPU, etc.) which is why there are variances in results of HT On/Off.  I do feel it's a valid summation that if one plans to Overclock their CPU, then turn HT OFF.

Cheers, Rob.

Edited January 4, 2019 by Guest

[westman 257]

1 hour ago, Rob Ainscough said:

 

Hyper-threading is Intel's approach to increasing the number of independent instructions in the pipeline and REQUIRES OS support.  It's important to note that HT does NOT duplicate main execution resources ... HT increases system latency (not good), can increase cache thrashing, and increases heat and power consumption.  How effective HT is varies significantly by application and overall performance of all the sub-systems from RAM, disk I/O, cache sizes, CPU frequency etc.

For all the processors I've used HT OFF and higher CPU/cache frequency produces better results relative to P3D.  But it really depends on your "Complete" set of hardware (MB, GPU, DDRx, CPU, etc.) which is why there are variances in results of HT On/Off.  I do feel it's a valid summation that if one plans to Overclock their CPU, then turn HT OFF.

Cheers, Rob.

Agree to 100%, done the same for ages

[ckyliu 1,749]

Having read all this I've never felt more pleased to be simply running a bog standard quad core with no HT functionality 🤣 I've always regarded FSX/P3D affinity masking with some suspicion, I certainly don't think it's worth all the time spent tweaking and experimenting for what seem to be marginal gains, except perhaps if you have chronic micro stutters. The only thing that seems to be a certainty is that turning HT off typically allows for a faster stable overclock and anything ESP based benefits from a higher clock more than most tweaks.

Edited January 4, 2019 by ckyliu

[Ron Attwood 3,467]

I WAS kidding Rob. Unfortunately it wasn't recognised as such. My bad.

Cheers, Ron