Has anyone tried the 10900k yet?

...It will perform just about identically in gaming as any other processor going 4.7 GHz, including ones that cost over a thousand bucks.

 

...and it's only because of the higher maximum overclocks, not the 67 extra cores :)

 

If you just want to burn lots of money, get an RTX Titan. It actually *will* make an improvement.

 

AD

This isn't an accurate statement. Same clockspeeds across 7700K and more recent chips will be similar due to the same micro architecture and IPC inherent in the design of the chip. So all "Skylake" chips running at 5.0 GHz will perform similarly in single core metrics all else equal. Previous gens of chips will have lower IPC even when running at higher clockspeeds. I'd recommend anyone without at least a 7700K or newer that it's time to upgrade.

Based on benchmark data the 10600K is the sweetspot for price performance. After that it's your GPU holding you back on most games. Get the 10600K, and use the savings for a better GPU or a good cooler and then OC it and you can closely match 9900K/10900K framerates.

And the $2500 Titan will only get you a few frames over a typical 2080 Ti. DCS and other games don't use all that VRAM. It's a developer/AI card.

At some point in the future, more games will start to use 8 CPU cores but very few including DCS do right now. Tasks like production rendering and encoding are relatively easy to split up among cores/threads and you will see measurable gains with more cores. Rendering graphics in real time for games doesn't benefit from more cores with most current graphic engines.

does ram speed make any real differnce? i

yes, YES, YEEEES. For DCS it might even make more difference than the CPU! I wouldnt have believed that myself, until I "accidentally" tested it.

Last year I upgraded my old i5 to a ryzen 2600x. I didnt plan the upgrade, my motherboard just died, so I did not upgrade anything I did not need to, including my ram, which was pretty old DDR4-2133 that I could run at 2400.

If you look at any hardware review / benchmark suite, it will tell you that, yes, faster ram will improve gaming performance, particularly on Ryzens, but we are usually talking less than 10% even with super duper high end ram. Ram prices being excessive back then, I didnt bother.

Until I did, and considered adding the new ram to my old to get to 32 GB, but that would require running all my ram at low 2133 or at most 2400 speeds, so I benched the difference in DCS. And fell off my chair. The frame rate increase was perfectly linear. Comparing my old DDR4 at stock 2133 to my new DDR4 at 3200 I saw a 50% frame rate increase in my recorded test track (Su25 caucasus track with medium settings at 1080p, removing my 1070 as bottleneck), correlating almost perfectly with the increased ddr throughput. I dont even think synthetic benchmarks will show you that.

Now I dont remember the CAS latency of my old ram. Im sure it sucked, it was old valueram that dated back to when DDR4 was first available, so it wouldnt be only the higher throughput that helped me, latency may play a big role too, but the two are related and the difference was enormous. Think about it, even if I could overclock my cpu to 8 GHz, I probably wouldnt see such a boost. Did I feel stupid for not spending a little more on DDR4-3600.

In most reviews, you will see that gaming performance on Intel cpu's scales a little less with ram speed than on AMD cpu's (especially gen 1 and 2 ryzens), so the impact might be smaller on Intel, possibly due to its different cache or whatever, but I dare anyone to show me a review of even a ryzen that shows anything like the linear performance boost I saw on DCS. So dont rely on hardware reviews or cinebench to judge DCS performance.

To anyone saying that these newest intel cpu's are faster or not than the old 47xx chips; I really dont think the difference is in the chip or its IPC. Let alone the clockspeed. I bet by far the largest difference is due to the fact that the more modern chips can use faster DDR4. That may not matter much in Tomb Raider, so toms hardware guide may tell you its barely worth upgrading, but from my testing, ram speed seems to matter a hell of a lot in DCS.

Dont believe me? Try it. Underclock your memory and see if or how the performance drop correlates with your decrease in ram speed. If its anywhere near linear, you may assume that increasing ram speed will also lead to similar linear performance increases. And I would be very curious to see results on intel platforms.

(BTW on a general note, under clocking is a good and simple way to help you identify performance bottlenecks. You can easily and significantly under-clock your ram, cpu and gpu and then measure the result. If there is no or almost no difference, then upgrading that component isnt going to help you much or at all either. If you see a significant or ideally close to a linear decrease, you almost certainly identified your main performance bottleneck)

Well after reading almost every thread, I bought a i5-10600K + Gigabyte Z490 Gaming X + Gskill TridentZ DDR4 32gb 3600 CL17 to match my RTX 2070 Super.

I'm still waiting for some parts. Let's hope VR performance improves a bit!

I saw a significant VR performance increase going from my 4690K(4.2GHz) to a 10600K, not so much in terms of maximum framerate but reduced stuttering and increased lows.

Unfortunately no frametime numbers due to the OTT overlay not working.

However, the YAAB: Yet Another Arma Benchmark saw an 85% performance increase.

This is with a 2080Super, 3600MHz CL16 and after overclocking, 4.9GHz all core/4.8GHz cache.

Yeah, I get 45 fps locked on the ground single player F-14. 50/60 at altitude.

Gonna try the carrier tomorrow

 

All of the reviews overly focus on how 'good' amd are as they have so many cores but realistically, to get the best in our chosen software, we need IPC, not 100s of weaker cores. So the info albeit useful is somewhat biased towards the minority that actually use the CPUs in production software and the games tested are typically not so CPU bound as DCS.

It isn't even IPC. We need the very best single-thread performance for DCS, and often gaming in general--but especially DCS.

Higher IPC but lower clock speeds give no performance increase, and likewise, higher clock speeds and lower IPC performance gives no improvement either.

We're at the stage now where improving single-thread performance for consumer grade computers has become extremely difficult and exorbitantly expensive. The sweet spot is supposed to the i5-10600K, which offers great single-thread performance when overclocked, but isn't outrageously expensive like the higher-up ones.

It's still expensive as all hell though--and here in Canada it's impossible to get, as well as Z490 motherboards. Neither are ever in stock. A 10600K costs about $450-$500 here, if you can get one, and that's a lot of dough for a very low-performance paperweight. Which is exactly what it is with no motherboard!

AD

It isn't even IPC. We need the very best single-thread performance for DCS, and often gaming in general--but especially DCS.

As I wrote above, in my testing the most crucial bottleneck is actually IO: Memory access. Newer cpus tend to support faster ram (and be paired with newer/faster ram modules). They tend to have larger, faster and more efficient l2/l3 caches. Higher clocked cpu's also have higher clocked memory controllers and l2/l3 caches. I wonder if we are not focusing on the wrong thing here, at least for DCS and in its current form. CPU clockspeed let alone IPC may not matter nearly as much as is generally assumed.

If anyone has access to a threadripper or an intel x series on LGA 2066, I would love to see some results. If my thesis is correct, their large caches should help anyhow; whether or not quad memory channel also helps would depend if DCS IO is more latency than bandwidth bound, but I wouldnt be surprised if these chips outperformed their desktop counterparts in DCS despite lower clocks and basically identical IPC. Maybe even by a significant margin.

yes, YES, YEEEES. For DCS it might even make more difference than the CPU! I wouldnt have believed that myself, until I "accidentally" tested it.

 

Last year I upgraded my old i5 to a ryzen 2600x. I didnt plan the upgrade, my motherboard just died, so I did not upgrade anything I did not need to, including my ram, which was pretty old DDR4-2133 that I could run at 2400.

 

If you look at any hardware review / benchmark suite, it will tell you that, yes, faster ram will improve gaming performance, particularly on Ryzens, but we are usually talking less than 10% even with super duper high end ram. Ram prices being excessive back then, I didnt bother.

 

Until I did, and considered adding the new ram to my old to get to 32 GB, but that would require running all my ram at low 2133 or at most 2400 speeds, so I benched the difference in DCS. And fell off my chair. The frame rate increase was perfectly linear. Comparing my old DDR4 at stock 2133 to my new DDR4 at 3200 I saw a 50% frame rate increase in my recorded test track (Su25 caucasus track with medium settings at 1080p, removing my 1070 as bottleneck), correlating almost perfectly with the increased ddr throughput. I dont even think synthetic benchmarks will show you that.

 

Now I dont remember the CAS latency of my old ram. Im sure it sucked, it was old valueram that dated back to when DDR4 was first available, so it wouldnt be only the higher throughput that helped me, latency may play a big role too, but the two are related and the difference was enormous. Think about it, even if I could overclock my cpu to 8 GHz, I probably wouldnt see such a boost. Did I feel stupid for not spending a little more on DDR4-3600.

 

In most reviews, you will see that gaming performance on Intel cpu's scales a little less with ram speed than on AMD cpu's (especially gen 1 and 2 ryzens), so the impact might be smaller on Intel, possibly due to its different cache or whatever, but I dare anyone to show me a review of even a ryzen that shows anything like the linear performance boost I saw on DCS. So dont rely on hardware reviews or cinebench to judge DCS performance.

 

To anyone saying that these newest intel cpu's are faster or not than the old 47xx chips; I really dont think the difference is in the chip or its IPC. Let alone the clockspeed. I bet by far the largest difference is due to the fact that the more modern chips can use faster DDR4. That may not matter much in Tomb Raider, so toms hardware guide may tell you its barely worth upgrading, but from my testing, ram speed seems to matter a hell of a lot in DCS.

 

Dont believe me? Try it. Underclock your memory and see if or how the performance drop correlates with your decrease in ram speed. If its anywhere near linear, you may assume that increasing ram speed will also lead to similar linear performance increases. And I would be very curious to see results on intel platforms.

 

(BTW on a general note, under clocking is a good and simple way to help you identify performance bottlenecks. You can easily and significantly under-clock your ram, cpu and gpu and then measure the result. If there is no or almost no difference, then upgrading that component isnt going to help you much or at all either. If you see a significant or ideally close to a linear decrease, you almost certainly identified your main performance bottleneck)

DCS may or may not be RAM bound. Depends. Since your system dates back to 2011 I would assume what your observing could be alleviating a bottleneck on your particular system but may have diminishing returns for more modern systems which very likely will benefit more from a faster graphics card. I made this observation because I had opportunity to test changing graphics card (from a GTX 970 to a 1080Ti) on my Ryzen machine and I saw huge improvements from that but less so from processor (2700X to a 3900X) or switching from 3200 to 3600Mhz RAM frequency.

While not a 10900k I have just ordered the following this evening:

Intel Core i7 10700K, S 1200, Comet Lake, 8 Cores, 16 Threads, 3.8GHz, 5.1GHz Turbo

ASUS ROG STRIX Z490-F GAMING, Intel Z490, S 1200 motherboard

64GB (2x32GB) Corsair DDR4 Vengeance RGB PRO Black, PC4-25600 (3200) memory

should all be here on Friday, going from this trusty 4790k setup to that should be a nice upgrade, will be using the 2080RTX and SSDs etc and carrying them over to this.

Since your system dates back to 2011

My Ryzen 2600X is not THAT old :)

I saw huge improvements from that but less so from processor (2700X to a 3900X) or switching from 3200 to 3600Mhz RAM frequency.

Note that I explicitly tried to reduce the GPU as bottleneck in my tests. No doubt, if you run at high res and turn up the settings, and especially in VR, its easy to make the GPU become your main bottleneck and then obviously you can no longer expect linear scaling by adjusting other components, like CPU or ram.

Id still be interested in seeing your results at 1080p low or medium with varying dram clockspeeds. Ryzen 3xxx performance is said to be less dependent on ram speed than my generation, but Id still be surprised if you wouldnt see similar, near linear scaling, especially if you can keep CAS latency constant.

Got the 10700k system running last night (no OC, 64 GB CL14 3200, fast NVME storage, 2070 Super Gaming X factory OC).

I'm no longer CPU bound on the Channel map in either 1440p or 4K. I'm not in front of the data so won't give numbers but lots of headroom. Runs smooth as glass everywhere EXCEPT an occasional stutter when low over the town of Dover and screwing around zooming buildings in the Spit. I think this is a code problem and not hardware - like a block of data is loaded or swapped out and is slow for some reason. It doesnt happen consistently. Could also just be Windows being Windows (POS).

Runs cool - CPU max is 60 degrees, GPU is 70. Case is a 'be quiet 500DX' - really nice case but I dont have much experience with these things.

GPU is maxed at 100% in both resolutions but you'd never know without the data. MSAA x4, Ani x16, high most things. I turned on SSLR in 4K and couldn't tell a difference in performance by seat of the pants.

Happy with the machine, it's a bit of a beast. I probably won't bother OC at this point and will add a new 3xxx series GPU in the fall. Really need a better screen.

Is it posible to aply DLSS 2.0 on DCS?

Don´t know if that could help with that Dover stuttering problem.

Anyway: stuttering with a fresh install on that machine is telling that the map needs improvements yet.

Which Mobo did you get reece146, I am going for the 10700k but the mobo choice is the only thing stopping me from upgrading at the moment. I was looking at the Gigabyte Aorus Elite without the wifi.

I did the Gigabyte Aorus Pro AX.

Depending on your intentions you might want to take a look at the mobo manual before buying and make sure you have the CPU lanes going to the devices you want them to go to.

I wanted the first two PCIex16 to split to x8 + x8 with two cards just in case - now that I have it built I'm not so certain that is a use case I'll ever use.

This board has all the overclocking stuff that people like but now that I'm using it I'm not so sure I'll bother overclocking. CPU is NOT stressed at all in single player - haven't tested in multi player but it's got so much head room doubt it'll be an issue.

For cooler I went with the 'be quiet' Dark Rock Pro 4. Really nice piece of kit. It is kind of low hanging though - I can just get the first RAM stick in and out from under it (G.Skill Trident Z DDR4-3200MHz CL14-14-14-34 - F4-3200C14D-32GTZ). You can probably use that info to get the RAM physical height if you are curious about that with that cooler. I got lucky it works - did not design around the RAM to cooler height. I have to remove the cooler's exterior fan and all the other sticks to get the #1 slot stick in and out though. Not a big deal - how often will you do that really...

This is the first PC I've built in like 15 years - things have moved on quite nicely. I'm really impressed with the quality of all the parts I bought - not the expensive junk we were buying/building in the 80s-90-00s.

You have sold me on it mate, the thing that was stopping me from buying the elite board was the realtec 2.5 GB LAN, as I would prefer the Intel Lan and the PRO AX has the Intel GB Lan. I am sorted for the cooler as I have been advised by Arctic that my ARCTIC Liquid Freezer 240 will fit the 1200 socket.

Thanks for the info, much appreciated.

i7 10700k good to go :)

https://valid.x86.fr/kab7p1

https://www.userbenchmark.com/UserRun/30230608