A small thing but I thought interesting enough to share.
I am still waiting for some hardware bits to arrive in the mail, but I have had the i9 9900K, new motherboard and some new DDR4 for a couple of days now. I’m still waiting for the 2080 GPU plus some new cooling bits, so I can really break this stuff too. One of the DDR4 3200 sticks was dead on arrival sadly, but the nice people at NewEgg were pretty responsive and shipped me out a replacement set next day while I RMA’d the bust ones.
Because they come in a pair, and it is ok to run a single DDR4 stick, I ran yesterday XP11 on 16GB (as the other 16GB was dead). I was keen to check out the i9, even at stock speeds and the old 1070 GPU. Today I put in the new set for the full 32 GB and was mildly surprised to see how much it helped X-Plane 11.
I do use a lot of ‘stuff’ with XP, so that’s a mitigating factor. I’m using:
Active Sky XP (a separate process that comes up in its own window).
Pilot2ATC (another separate app that talks to X-Plane)
World Traffic 3. Using the real-world routes/aircraft packs. A plug-in that loads up in XP11.
XP11 latest beta, in VR at 200% SS.
Zibo 737 700 using Avitab in-built.
What I found with the above today was that I using about 20GB of RAM, once all things are up and ready to fly. As previously using 16GB, it wasn’t like I ‘ran out’, just more that it was swapping to the pagefile or not caching extra stuff. Things do seem a lot smoother, just from the extra stick of RAM alone, in that it’s an obvious thing to notice even with nothing else on the hardware changing.
I think we’re at the case where the sims aren’t using anything like the full 32GB, but they are starting to use more than 16GB, so it’s at the inflection point of if it is worth it or not. For the extra money (2 x 8GB vs 2 x 16GB) I think it is now worth it.
I’ll check and see how much mine is using…does the 2080 offload any of that RAM usage or do the processes have distinct, dedicated allocations that don’t do any load sharing? (You might not know that…by that I mean like is flight modeling CPU computations and graphics GPU?)
Pretty much separate, in VRAM for textures and buffers is on the GPU side, and will use some regular RAM if exhausted, but the method of moving it is a magnitude slower than what the GPU bus uses.
Sadly my 1070 used for this test has 8GB, and the 2080 I’m getting only has 8GB, while your 2080ti has 11GB, so it might be slightly better for the super high resolutions we’re pushing through on this thing.
Hmm…just booted up after a couple days away and see that SteamVR has updated…and oddly enough, my Global Resolution Setting (Automatic) has been set to 150% whereas before it was auto set to 200%… Wonder what is going on there?
Automatic Application Resolution scalar is now clamped at 1.5x (was 2.0x) and resulting resolution is now aligned to 4 pixels. Users may still override this value with a custom setting under the Video section of the desktop SteamVR settings or in per-Application settings.
So with SkyMaxx Pro and Real Weather Connector running (but no clouds in this area of the world), X-IvAP, and VR set to 150% SS (I need to change that back to 200%) and most settings medium to high, and a moderately dense scenery airport (Key West payware) I’m getting RAM usage of 18.2 out of 31.9GB and GPU 10.3GB out of 11GB…
No, just set the XMP profile on the memory, but all vanilla. It’s killing me.
I’m using an old H60 water cooler block on it, but have a huge new chunk of metal and fan coming in the mail. Check out this puppy - https://noctua.at/en/nh-d15
It is worth pointing out that for the i9 9900K stock everything, it will just overclock itself using the Intel Turboboost anyway. You should be seeing 4.7GHz across all the cores when it gets exercised. It has the ability to go to 5GHz automatically only one or two cores, but what tends to happen is that Windows is always doing stuff, so the 4.7 is most often seen.
If you are every curious on what your individual cores are doing then I do recommend HWINFO64, it’s a great sensor viewer package:
Just use the ‘Sensors’ option and you’ll get lots of info on practically everything going on, including the automatic multiplier speed changes that the i9 is doing for you.
My plan with the new cooler is to play around with making all the cores 5.0 or maybe a bit higher. With new stuff it usually takes me a week or two to get up a head of steam to try to ruin something that works fine already.
I think you just overclocked it all to 5GHz using that tool then - cool!
The only real difference between doing it manually and with a util is that you can sometimes managed to get a little lower temperature, just because you can lock in what sort of voltage you need for stability a little more precisely. It is a pain in the butt to do, as you have to run stability tests longer and reboot a lot, but it allows you either to safely go a bit higher or just run at 5Ghz (chip lottery dependent) and with lower overall temps. The other difference is that turning off the speed state stuff means the speed tends to stay the same but you don’t underclock automatically when it is quiet. This is because it is usually the voltage changes that break an overclock, so a manual overclock tends to be a little less efficient in electricity as it is using more constant juice.
Beach is now overclocking eh? You’ll be off down to the aquarium to pick up that water cooler units soon enough…
I have that cooler, and I can confirm it is AWESOME! I’ve never seen temps over 50 deg C or so, although I am still running an old 2700k overclocked @ 4.6 GHz. I’ve never managed to push the system hard enough where water cooling would make a big difference, I think.
Good to hear. It’s a pretty large unit, so on measuring the space in the case I only have about 2mm free, so going to be a squeeze.
I’ve only used water cooled self contained units for years now, but more for the quietness and ok for the size cooling ability. I was going to go that way again, but wanted a 280mm dual radiator set-up, and realized I would need to change cases (my existing case has an enormous slow moving fan at the top, that I didn’t want to hack around with) and sort of lost interest. The noctua looks great, so looking forward to trying it out.
Absolutely. It’s an unlocked processor, meaning the multiplier can be changed and was designed as such (plus was an extra feature that increased its price). It does automatically change speed from the default 3.8GHz to 4.4GHz by itself already, as part of the built in Intel turboboost.
It really depends on your motherboard, RAM and your cooling for how much is possible in terms of increasing over stock speeds. The unlock processors (X or K ending specs) usually have motherboard utilities that run in Windows to overclock. They change the voltages and clock speeds and work well, although can be a little coarse in their changes, leading to using more voltages (i.e. heat) than needed.
Yep, give us a shout if questions or do let us know how you get on. I’m no expert at it but do like finding out what the various options do. Chipsets in the last few years have really robust thermal protection, so it is actually pretty hard to do anything like damage - a chip will just underclock itself as a safety feature, and with unlocked multipliers, XMP profiles and the like is pretty straightforward compared to what it used to be.
