Groove_C

Hope that the additional L3 cache will also be available for 5800X and 5600X and not just for 5900X (demonstrated by AMD) and 5950X.

Wanted to buy 11900K today and also today was announced Zen3+, still on AM4 with DDR4 and PCI-E 4.0, but with 2/3 L3 cache more by the end of this year or the beginning of the next year. So it's 96 MB L3 instead of 32 MB L3 for 16 threads. What a good timing! So my i7-5775C will have to work until then.

If it wouldn't have been for OC, I no doubts would have bought R7 5800X, since at stock it performs better then 11700K/11900K. Even for CPU only OC I would have bought R7 5800X, since it's not complicated at all. Only RAM OC is much more complicated on Ryzen than on Intel. RAM OC in general, at all times has been the most complicated (vs. CPU and GPU) on all platforms and continues to get more and more complicated/sofisticated.

On Intel I will simply increase all the voltages, look which settings it can do and after testing for stability, will simply start to increase each voltage, one after another and recheck for stability to find min. stable voltages. It's still a lot of work, but much less than with Ryzen and at least I already know how to do it. So it will again cost me less time.

Yes, too many settings, software and time. But the thing is that tuning RAM from stock increases FPS considerably. But time needed to do this is also considerable. Having almost 100 FPS avg in YAAB 1080p standard thanks to well tuned RAM is a lot. More than 20 FPS more is a lot, in Arma. It's not like going from 100 to 120 or from 220 to 240 FPS in other games. Because having lowest FPS in Arma being not 25-30, but 45-50 is a big difference. Tuning RAM on Ryzen 5000 from like 3600 CL16 to like 3800/3866 CL14/15 is at least 10 FPS more, for free, if you don't count your time. Min FPS on my 2014 system is 45, in YAAB 1080p standard, which is nothing, because I don't play at 1080p, but at 1440p and not on standard, but at ultra. So my min FPS, on settings I really use to play daily, is much lower and thus such an upgrade is not even questioned - no hesitation at all. I will still spend a lot of money, but at least the FPS difference will finally be noticeable and not just measurable. Because with a well tuned 10th gen Intel and good RAM, FPS increase would have been like 10 FPS only (vs. my well tuned 2014 system) and that's for like 1200€. And it was too little for too high cost.

The only reason I'm still going with Intel, despite it costing me 200€ more for same performance, once/when both OC'ed, despite higher power draw, is because I'm used to how OC'ing CPU and RAM on Intel works and it will be much easier for me to OC it than to learn a lot of new stuff to OC AMD, spending too much time doing so. With Ryzen 5000 there are a lot more different voltages than on Intel and additionally also a lot of different resistances in Ohm, which Intel doesn't have. + unlike on Intel, you can't simply increase all of this and when you find your stable settings, simply begin to lower all the voltages/resistances. Voltages/resistances have to be no lower than, like on Intel, but also not higher than, which makes it more complicated for me and woud require a lot of time and learning process that I don't want.

Raptor Lake (13th gen in 2022) will be a refesh of Alder Lake with higher performance and larger cache. It's like 7700K vs. 6700K. Meteor Lake (14th gen in 2023) will be another approach than Alder and Raptor. But it will be the first Intel try in 7nm, which won't be very good, because it will be the first and it will be the last on LGA 1700. So I wouldn't go Intel again before late 2023 or 2024, when they release Lunar Lake (15th gen), their 2nd gen in 7 nm and a new socket that will stay for at least few generations. Or go AMD.

With my 11900K and fastest possible DDR4 I will wait until DDR5 gets faster and new CPU architecture matures after few generations.

Alder Lake will have very expensive and very slow (very high timings) DDR5. Because higher RAM frequency is only higher bandwidth - not better responsiveness (in nanosecs in Aida64). RAM bandwidth doesn't increase FPS in games - only RAM responsiveness (in nanosecs in Aida64) does. Before we see DDR5 with even higher frequencies than the stock 4800 MHz, with adequate/low timings and at an acceptable price, at least few years will pass. I'm also very skeptical what concerns Windows tasks scheduler in combination with Alder Lake, because it will have more and less powerful cores at the same time and running at different frequencies. So I prefer to buy a CPU with a proven design, in combination with the fastest DDR4.

Getting 11900K and Asus Apex Z590 in 1 week.

Faster RAM let's the CPU and GPU better show what they're capable of. Faster RAM bottlenecks CPU and GPU less. With faster RAM, data for the CPU is prepared faster, thus the CPU spends less time doing nothing and is fed with more data or is fed with data faster, which in turn makes the GPU also idle less, waiting for data from the CPU. So with faster RAM, the CPU itself is also better loaded/used in %, transmitting more data or transmitting data faster to the GPU, which then also waits/idles less and thus is also better loaded/used.

Depends on how you look at it. With 3600 CL14 at stock (XMP), you know it was preselected/binned at factory, to some extent, which of course requires additional time and manpower and better memory chips for sure. But numbers not worse than what stated on the box are at least guaranteed, in form of specs and aren't just a pure luck. And that's what you're paying for. Have already often enough seen enormous discrepancies past guaranteed stock specs on same model products, just different samples. While 3600 16-16-16-36 is certainly better value, the chance to push it to 3600 CL15/14 or higher, due to usual quality discrepancies within different samples of same model, is much slimmer. The additional performance you potentionally could squeeze out from it, is not guaranteed by the manufacturer, as it's past the specs on the box and is rather a nice bonus, in addition to what it's capable of at stock and could potentially apply only to some of the samples, more for some and less for the others - it's a lottery. It's like if you buy several G.Skill Trident Z (Neo/RGB/Royal) 3200 14-14-14-34 model samples. Some will do even 4600 CL16, while some will already max out at only something like 3600 CL16.

It's OK, but it could be cooled better, for higher frequencies at same or lower voltages and higher RAM frequency with lower timings. Noctua NH-U12A would have been a better option for sure. Or NH-D15S Chromax Black.

No, not Yet.

@rowdied I myself have G.Skill Trident Z Neo 32GB 3800 MHz 14-16-16-36 (F4-3800C14D-32GTZN). But 32 GB 3600 MHz CL14 is already very good. Actually the second best <4000 MHz, after mine.

@rowdied G.Skill Trident Z Neo 32GB 3600 MHz 14-15-15-35 (F4-3600C14D-32GTZN). 16 GB now is already almost like 8 GB several years ago. So it should be not less than 32 GB. And RAM will be even more expensive soon. Already known/announced information.

@rowdied MSI Tomahawk is much better than the Gigabyte Elite.

If there will be no XT refreshed CPU models, I still will be able to buy current 5800X for even less than now.

X570S motherboards will come soon. Maybe we will see Ryzen 5000 XT CPUs as well. I will wait for sure.

@Chimere seems like I will buy a 5800X then, hoping it will be able to do -30 on all cores. I have a Noctua NH-D15S that must be on par with your Noctua NH-U12A. I have no problems to let it run @ 100% while stress testing. Runs @ 1000 rpm in Arma and on desktop. I don't pay that much attention to the max voltage, since it's possible it will be like this just for few seconds or even less. I take the average voltage as "real" voltage as it's the voltage at which the CPU will spend 99% of the time.

@Chimere was 1.35 V the max voltage, the lowest or average/current voltage and which utility do you use to monitor the voltage? I ask to try this exact Prime95 version: ftp://ftp.mersenne.org/gimps/p95v298b6.win64.zip Select "custom" instead of "blend". Select "Run FFTs in-place" and unselect AVX2/FMA3 and AVX and let it run like 10 minutes and please report the frequency of all cores, max temp and voltage you observe, thx. It will help me a lot with my decision what to buy.

@Chimere what Prime95 version? So all cores were constantly @ 4.9 GHz in Prime95? What was the voltage? Which settings have you changed in the BIOS and to what/which values?

Corsair have a small cheap weak plastic pump and thin radiator. You pay only for Corsair brand, RGB and useless software. Arctic has a massive metallic pump, massive radiator and their warranty was recently upped to 6 years instead of only 2 before. No RGB bling bling, no software needed and only cable to power it all up. Cools better than Corsair.

@ValkenArctic Liquid Freezer II 240 or better 280 mm don't cost 200 USD.

11400F frequency all cores too low, power draw too high, RAM it can accept is max 3600 MHz. 10400F even slower. B450 boards are limited to less than 3000 MHz RAM. 10600K is good. Would need a Z490 board for RAM to be able to operate at more than 3000 MHz. At least Asus TUF Z490 I would say.