Groove_C

R7 3700X ~4.2 GHz (SMT on) (Auto -> no PBO) 2x8 GB DDR4 3600 16-16-16-36 1.35 V @ 14-15-15-35 1.35 V via Ryzen DRAM Calculator X570 Aorus Elite GTX 1080 SSD Windows 10 + BIOS up to date Launcher mods + browser, Discord, Skype, Teamspeak etc + all known background programs/processes off YAAB 1080p standard Run 1: min 43 avg 65,4 (4219 MHz avg CPU frequency all cores combined) Run 2: min 42 avg 63,8 (4222 MHz avg CPU frequency all cores combined) Run 3: min 43 avg 63,5 (4206 MHz avg CPU frequency all cores combined) YAAB 1080p ultra Run 1: min 33 avg 50,8 (4191 MHz avg CPU frequency all cores combined) Run 2: min 36 avg 50,1 (4189 MHz avg CPU frequency all cores combined) Run 3: min 35 avg 48,7 (4202 MHz avg CPU frequency all cores combined) YAAB 1440p standard Run 1: min 41 avg 63,1 (4206 MHz avg CPU frequency all cores combined) Run 2: min 38 avg 64,9 (4226 MHz avg CPU frequency all cores combined) Run 3: min 43 avg 63,1 (4214 MHz avg CPU frequency all cores combined) YAAB 1440p ultra Run 1: min 31 avg 48,1 (4206 MHz avg CPU frequency all cores combined) Run 2: min 34 avg 47,1 (4196 MHz avg CPU frequency all cores combined) Run 3: min 31 avg 41,8 (42017 MHz avg CPU frequency all cores combined)

R7 3700X ~4.2 GHz (SMT on) (Auto -> no PBO) 2x8 GB DDR4 3600 16-16-16-36 1.35 V @ 14-15-15-35 1.35 V via Ryzen DRAM Calculator X570 Aorus Elite GTX 1080 SSD Windows 10 + BIOS up to date Launcher mods + browser, Discord, Skype, Teamspeak etc + all known background programs/processes off YAAB 1080p standard Run 1: min 43 avg 65,4 (4219 MHz avg CPU frequency all cores combined) Run 2: min 42 avg 63,8 (4222 MHz avg CPU frequency all cores combined) Run 3: min 43 avg 63,5 (4206 MHz avg CPU frequency all cores combined) YAAB 1080p ultra Run 1: min 33 avg 50,8 (4191 MHz avg CPU frequency all cores combined) Run 2: min 36 avg 50,1 (4189 MHz avg CPU frequency all cores combined) Run 3: min 35 avg 48,7 (4202 MHz avg CPU frequency all cores combined) YAAB 1440p standard Run 1: min 41 avg 63,1 (4206 MHz avg CPU frequency all cores combined) Run 2: min 38 avg 64,9 (4226 MHz avg CPU frequency all cores combined) Run 3: min 43 avg 63,1 (4214 MHz avg CPU frequency all cores combined) YAAB 1440p ultra Run 1: min 31 avg 48,1 (4206 MHz avg CPU frequency all cores combined) Run 2: min 34 avg 47,1 (4196 MHz avg CPU frequency all cores combined) Run 3: min 31 avg 41,8 (42017 MHz avg CPU frequency all cores combined)

i7-5775C 4.1/3.6/2.2 GHz core/cache/eDRAM (4/8 cores/threads) 2x8 GB DDR3 1600 MHz 9-9-9-24-1 2x8 GB DDR3 2400 MHz 12-13-14-32-2 GTX 1080 Ti 1920x1080 monitor resolution (test from May 2018 with Arma 3 patch 1.82) From this test, one can clearly draw the following conclusions: - RAM with higher frequency and lower timings really benefits to Arma, only on CPUs, that have not that much L3 cache - with Hyper-Threading on, FPS with 1600/2400 MHz RAM is exactly same as with Hyper-Threading off - enabling 128 MB eDRAM/L4 cache, eliminates the need to send the frame data to the intermediary in form of RAM, by proceeding frame data inside the 128 MB cache, without it leaving the CPU, thus leveling out the performance difference between RAM of different frequency and timings - enabling 128 MB eDRAM/L4 cache, boosts the min FPS from 40-42 to 52 My i7-5775C operates at 4.3/3.8/2.2 GHz with 4x8 GB DDR3 2400 MHz 10-11-12-18-1 😏 Source: https://www.ixbt.com/live/platform/igrovaya-proizvoditelnost-core-i7-5775c-v-razlichnyh-rezhimah-chast-2.html Other games were also tested for difference with HT on/off, different RAM frequency/timings, 128 MB L4 cache on/off and DX11 vs. DX 12. The test is in russian, but the pictures are self explanatory.

In April one will be able to decide whether to buy an i5-10600K (6/12 cores/threads) or an i7-9700K (8 cores/threads) for lower price than now.

I know for sure right now, that there will be an i5-10600K with 6/12 cores/threads. It's already confirmed. Coming in April.

🤕 https://www.zdnet.com/article/new-plundervolt-attack-impacts-intel-cpus/

If AMD won't significantly improve RAM latency on Ryzen 4xxx and won't make it possible to go 4000-4200 MHz RAM in 1:1 mode, I might buy Intel again, despite PCI-E 3.0, 14 nm and much smaller cache.

Ryzens handicaped by 3200 MHz RAM (because of Infinty Fabric) vs. no handicap for i7-9700K with same RAM (because one chip design -> no Infinity Fabric). So if you consider that one needs to buy a 3600 MHz good performance RAM for Ryzen to only get close to Intel, it impacts a lot the total price, as only a fraction of users know which specific RAM models can be bought and how to OC them to 3600/3800 MHz CL14/CL15, to save money. While the average FPS from i7-9700K (5.0 GHz) is only slightly higher than for 3700X (4.3 GHz) and 3800X (4.5 GHz), the 1% low and 0.1% low FPS is an other story, really. And this is with no cache OC on i7-9700K + it's not even a i9-9900K(S).

Is it with PBO, CCX OC, overvoltage or simply latest BIOS? 3600X with 4.4 GHz or 3800X with 4.5 GHz is certainly better, for Arma.

@oldbear and average for each core?

On Ryzen CPUs, to have an idea of the approximate frequency, since it's different on each core and fluctuates during tests on same core, one has to reset the values with application/game launched, when the load starts to be applied. After this, one can observe average frequecy of each thread and calculate an average frequency of threads combined.

@oldbear I don't think that min frequency in Arma drops down to 2.2 GHz and the average is 3.45 GHz. It's most probably the min on desktop and this is why the average is 3.45. Could you please press on the "clock" button in HWiNFO to reset all the values when you run YAAB? To have accurate picture of actual frequency per core. Thx in advance.

@oldbear what is the CPU you have, a 3600 or 3600X? What's its average frequency in Arma, per thread and all threads?

Furthermore, all 2020 Intel CPUs will suffer from same performance loses, right out of the box, since they will be based on current architecture. Also a well know fact, for people that researched a lot on this specific matter, is that Intel knew right from the start about a lot of flaws in the architecture, but was hoping that nobody will discover them. But this way, they could squeeze out more performance, spend less time on the architecture and invest significantly less. So basically, they could have avoided a lot of these security holes right from the start, by taking more time and investing more into the architecture.

New Ryzen CPUs were reviewed vs. Intel in July, whereas 77 new security holes in Intel CPUs were made public only in November, despite them being provided to Intel by different research universities a long time ago. Intel had to make them public, because of pressure from different research universities, that threatened Intel to make them all public themselves. So Intel took the initiative, to limit its image damage. Not all of them could be fixed up to date + previous BIOS/Windows patches, that already cost performance, were not able to fix completely the holes they were released for. A bunch of new holes were already discovered by these same research universities, which will additionally cost performance to Intel in the near feature, when they will be made public, once Intel gets fixes (for the majority of them). Not everything can be fixed. Some holes require a whole new CPU architecture in order to fix them. AMD/Microsoft patches were releases also past July reviews. So the majority of reviews don't reflect the current state of performance of both brands' CPUs. Considering all of the above, I can't say I'm really a happy Intel user/owner.

2020 Intel CPUs won't be much different from now. Only 100 MHz more at stock and lower prices, to try to counter AMD Ryzen 4xxx, partially. They will still require watercooling and a decent mainboard, because of all the heat and energy, since they will still use same old technology from 2015.

@Tankbuster you know very well that I proposed to you to wait a little bit to see what to buy, since your PC was still working and its performance wasn't bed at all. But you decided otherwise. Sure, 24 threads and 70 MB cache is better than 16 threads and 16 MB cache, especially considering +- same price (depending on where you live). + No need for watercooling and expensive mainboard. + AMD/Microsoft are still working to further improve BIOSes and Windows, for even more performance and even better power and load/threads management.

If you're not as well informed, it's ok. But majority of known tests were published before so many new security holes on Intel CPUs were made public and Windows/BIOS patches came out as a result. Same goes for tests of AMD's current CPUs. When majority of tests were done, they couldn't perform at advertised frequencies and there were troubles with Win 10. So Intel CPUs performed better in tests, because none of latest security patches were applied to them, which degrade their performance up to 10%. AMD CPUs performed worse, because lower frequency as advertised and prior to BIOS/Windows updates, which fixed most of this. Just FYI. And I repeat once again, I myself have 2 Intel CPUs, but I'm not a blind fanboy. If AMD gets better, then I will buy AMD. I won't buy a product of a certain brand, because it's this brand or because I had only positive personal long term experience with it. Things change and one should track market/changes, to make a wise choice, not a blind choice.

I see you constantly liking positive comments only about Intel CPUs, which is pure non-sense. You know, sometimes manufacturers change the lead. It happens. Nothing dramatic. It's actually beneficial to consumers to have competition between manufacturers, for better price, performance and progress.

@Tankbuster you better attentively read what I write, before you comment. After latest Windows and BIOS updates to patch most of 77 security holes (not even all), Intel CPUs lost up to 10% performance because of this, when the perf difference vs. AMD was already sub 10%, despite much higher frequency on Intel CPUs. + AMD has released several BIOS updates to further increase CPUs' frequency + all the Windows updates, which favor a lot AMD CPUs. So the difference at the moment is only marginal/measurable, but no real difference. And this is at much higher frequency. Intel CPUs at same or lower frequency than current gen AMD CPUs, perform worse. It's a fact. I myself have 2 Intel CPUs, but I'm not a blind fanboy. I buy what is best at a given moment, regardless of brand and regardless of (long term) personal experience with a certain brand.

@SearchAndRescue no, it's not correct at all. It has stock frequency of 2.9 GHz, max boost frequency of 4.1, but it's not for all cores. All cores frequency is 3.7 GHz only. R5 3600 has stock frequency of 3.6 GHz, max boost frequency of 4.2, but it's not all cores. All cores frequency should be around 4.0 GHz or slightly more. 4.6 GHz boost is for i5-9600K and again, it's not for all cores. One can probably OC the i5-9400F to have all cores boost of 4.1 GHz, but depends on motherboard + AMD at same frequency will be faster, because more cache and also has double the threads. Intel can only marginally overtake AMD only by it having much higher frequency than AMD.

@adamsean01 what's the resolution of the monitor? 100% R5 3600, since it has higher frequency all cores than i5-9400F and more threads and cache.

My mainboard is also gradually/progressively dying. I hope it will last me until Intel 10th gen or Ryzen 4xxx. Otherwise it will also be an i7-9700K. On Intel CPUs one can also manually overclock each core separately, since like I said before, they're not all of same quality. So some cores can handle 100-200 MHz than others. It will look odd for sure, but it's a good approach, despite it being more time consuming than traditional OC with same frequency for all cores.

Its VRMs are lower quality than on Elite + VRMs cooling is also not great at all (very basic). So VRMs can get pretty warm and thus CPU's frequency will/can be lowered, even before CPU itself reaches 100°C, which it won't. Also it's good to know that such VRMs can allow to overclock the CPU later, when it will start to be a bottleneck, to like 5.0 or 5.1 GHz, with another cooler.

@Boomerang Trotter if you don't simply choose the ultra preset in video settings, but make adjustments to it, you won't see any visual difference when moving/running, driving or flying. Ultra settings are FPS-hungry, but to see the minimal difference (vs. very high), you have to make screenshots and search for differences and know where to look for. It doesn't make any sense to buy R7 3800X for same or higher price than i7-9700K. If you go this route, it's to later directly upgrade to new CPU, RAM and mainboard. If only 3200 MHz CL16 RAM instead of 3600 MHz CL16, Intel delivers more performance with slower RAM as it's stock frequency all cores under load is 4.6 GHz. Thermalright Macho HR-02 should be enough if you use the CPU stock. I wouldn't recommend Gigabyte Z390 UD at all - to avoid! Gigabyte Z390 Elite is what you should buy. It has better onboard audio, better and cooler power delivery elements for the CPU, has a cooler for M.2 NVME SSD, has USB 3.1 ports. Arma also prefers more real cores vs. just threads. 1 thread is like ~25% of 1 real core. So 4 threads is like 1 real core.