I'm upgrading from 3570K and looking mainly at these two options right now. Intel seems to have a bit better game performance at stock and also some room for overclocking. Both CPUs cost the same where I live ($330+tax) and I already have a decent air cooler (TRUE Spirit 140) which hopefully is going to be enough for 9700K, so there're no extra expenses for that. AMD could be a more future-proof platform, but to take advantage of that I would have to get a good MB and those aren't going to be cheaper than a decent Z390 board, so the total cost of MB+CPU is probably going to be about the same. I'm not really interested in any "productivity" stuff as the main things I do on this PC is gaming (which is simracing in Assetto Corsa for ~80% of the time) and web surfing. The current GPU is 1070 which roughly matches the CPU (as both tend to be just not fast enough for real 90 fps in VR with medium-size grids), but I'm going to upgrade to 2070s or 2080 later this year. So far, I'm leaning towards the Intel, but I have a feeling that I might be making a mistake as there's basically no upgrade path there, though a similar situation worked more or less OK for my 3570K build which I've had for 6+ years and only recently (after getting a VR headset) it started to show its age for me. What do you guys think?
If you’re only gaming then stick to Intel as they still have the higher IPC (and I say that as an Ryzen 5 2600 user), especially if you overclock them. If you’re also into streaming, working in 3D or with Adobe programs for example, then Ryzen is the way to go.
If you really want only simracing and especially VR, go with the i7. The 3700x is the best allround CPU now. It's got a slight advantage for games that use all cores like the latest Assassin's Creed, F1 2019 etc but it's only a slight advantage and only for some reviewers. The 9700k can go the highest without throttling and is just a beast. The 9900k got hyperthreading but HT can cause microstuttering etc in some games and not really gives you a big boost for gaming. Without a massive cooling and a manual overwrite of the maximum power usage, it will run slower compared to the 9700k. In VR you mostly only have 2-3 game threads running on the CPU, you want the highest single thread performance you can get with as many real cores you can get. The i7 9700k beats them all. Only my a margin but it might be the crucial margin to get your 90 fps in VR! If you would play with gsync/freesync I'd always get the 3700x now as it's just a better overall CPU and I think we should support the competition. But not for VR+Simracing! I saw 9700k cinebench single thread results of up to 225 points. That's nuts!
Good advice from both men. It'll cost you less as well with their advice, based on what you currently own...cooler etc...
It used to, but gen 2 Ryzens now have a better IPC. Unfortunately they can't go close to 5.0Ghz that Intel typically can do with good enough cooling, so the overall per-core performance seems to slightly on the Intel side still
Again, the cost right now is the virtually same (the cost of bumping the system up in a few years down the road is another and complicated question as we don't know what future games (and sims) will require). I already have DDR4 RAM kit lying in my desk drawer, so I only need to get a MB+CPU and probably a cooler for AMD as the stock one doesn't look good enough for me and my current cooler doesn't have AM4 mount (but maybe I can the mount for a few $).
Thank you. This is actually the much needed piece of information here. There're no good VR-specific tests out yet and I had no idea whether the lack of cores or lack of per-core performance was the reason for my current system struggling in VR because it clearly lacks both. I absolutely agree that even 5-10% could be a "make it or break it" case for VR, seen that myself already where adding just a few more opponents to the grid can send me into permanent reprojection mode because the system can't do consistent 11.1 ms frame times anymore
Normally you should be able to get the mount cheaply Since you would go 9700k max OC or 3700x with hopefully highest clock I guess you should get a good motherboard with nice VRMs etc. I've just watched a few videos about x470 vs x570 and z390 boards. x470 is the way to go if you don't need the absolute best nvme pci-e ssd speed. PCI-E 4.0 won't be needed anytime soon. Even PCI-E 3.0 is barely needed and only for rF2 apparently but not really proven... x470 doesn't have a damn cooler, is cheaper, will continue to be supported and the 3700x clocks just as fine on them. Here are the winner boards: Z390: Gigabyte Aorus Master Gigabyte Aorus Elite (cheaper and still good VRMs etc) x470: AsRock Taichi (ultimate only got the extra LAN thingy). The AsRock apparently has the best VRMs, same level as the most expensive Gigabytes but costs a lot less, has a debug code display. There's also the z390 Taichi which also is advertised with the 12 phase VRMs but I didn't see it in the videos. And it's not cheaper than the aorus ones with this chipset... Hope that gives you some starting point To make it a bit more clear: Assetto Corsa could be labeled a "2.5 thread application". When you look into the running threads via "process explorer", you'll see when a thread runs into its limit. For my 4c/8t i7 2600k it's 12.5%. AC has one thread running between 11-12.5% if I unlock my fps and another thread running between 10-12%. And a third one a lot lower... And a lot of tiny threads without real load. So the real test would be to run Cinebench with 2 and then 3 threads and compare results but absolutely nobody is doing it... Actually I did with some friends of mine. In VR, you get a higher CPU load because of VR being VR. But it's not like you'd get a "VR thread", no, the VR load is just added to the already limiting 11-12.5% thread. Windows is pretty good at spreading single threaded loads across all cores (basically one core per splitsecond running at 100%, then the next one that already cached a little bit of data). It can increase the overall CPU load quite a lot so a 2.5 thread application will almost max out a 4 core CPU! (which is why the 9600k is basically fine with 6 cores. The 9700k only adds a few fps due to the better spreading with the same IPC and clock). Here's my custom Cinebench table across the amount of threads. It's a mix of tests from the internet and runs from friends and myself. I estimated and filled in the missing data from the data I had available and also calculated the results going from my own CPU being able to run 50 fps for direct comparison: And for 3 threads (closest to most simracing titles right now, going from 60 fps): i7 2600k @ 4.4 GHz: 60 i5 3570k @ 4.2 GHz: 58 i7 8700k @ 4.8 GHz: 85 i7 4790k @ 4.5 GHz: 70 Ryzen 2700x OC (whatever was possible with the AMD boosting): 74 i3 8350k @ 4.9 GHz: 83 i9 9900k @ 4.7 GHz: 84 However if you take AC:Odyssey, which runs on 7 main threads basically, the results would be: i7 2600k @ 4.4 GHz: 60 i5 3570k @ 4.2 GHz: 45 i7 8700k @ 4.8 GHz: 103 i7 4790k @ 4.5 GHz: 68 Ryzen 2700x OC (whatever was possible with the AMD boosting): 99 i3 8350k @ 4.9 GHz: 66 i9 9900k @ 4.7 GHz: 111 Here is the big graph:
I was thinking about getting Aorus Pro Z390. It has the same VRMs compared to Elite but got an extra m.2 heatsink (and I'm gonna use 2 NVME drives so it will come handy) for just a dollar more at a local store. Master and Ultra versions are too expensive and I don't think I really need those extra bells and whistles they come with. And yes, x570 is expensive right now without any immediate benefits and I'm not a fan of the damn fan (pun intended) either, so I would use 3700X with X470 or even B450 board with a good VRM instead. Thank you for your advice! I think I'm gonna wait a few weeks to see if there's any progress with new Ryzens. Maybe updated BIOSes will improve their overclocking or performance (unlikely) and maybe someone will eventually review them in simracing titles (F1 2018 and PCars 2 don't count ) compared to Intel CPUs
@demetri I forgot to mention: I'm also a VR user and I can run most sims just fine with the Ryzen 5 2600x. My 2080 might also help in this case. However I'd say that the last Ryzen gen and the current are good enough for actual sims in VR. Later this year I'll probably upgrade to a 3700 or 3900x.
I totally believe you that you're totally fine with your 2600x, not going against it. But "most sims just fine" and "good enough" aren't very accurate statements. Do you have some statistics of how much reprojection % you experiencing with some example-situations? Like maybe playing a full grid replay in Assetto Corsa that you could share for us for comparison? (Since a replay has a slightly lower load, maybe run a mod track with 30+ cars?) And of course what apps you're running in Assetto Corsa. My 2600k is dropping down to sub 60 fps during 24 car multiplayer race starts but I do have quite some apps active!
The damn fan alone is reason enough to avoid X570, I agree. It looks like the chipset is "only" consuming up to 9 W according to this chart, which raises the question why the fan was needed. Chipsets like P45 from 10 years ago consumed more than that due things like memory controllers not being integrated on the CPU, yet motherboards had some clever passive cooling solutions. IMO the 3700X is a winner in every category except gaming performance. The frequencies are not great, but a lower frequency doesn't matter when IPC was improved by over 10%. Even the single core performance is great now in almost all non-gaming benchmarks and matching Intel's. For whatever the reason, certain games still perform around 5-10% better on Intel, maybe due to the lower core-to-core latency of their architecture. I haven't yet seen benchmarks from sim racing titles, so hard to say what the difference will be in those.
RAM latency is significantly lower on Intel and most games are RAM-bound I believe as they don't do massive computations but rather interact with multiple complex objects representing the game world which typically means lots of RAM access in the process. I could clearly see that being the case for AC because just bumping up RAM clocks from 1600 to 2000 (with the same timings) gave me 10% fps increase which is no small thing. 2nd gen Ryzen got a beefier L3 cache and I guess that also helped with the games, but still not enough to match the leader.
Unless they've added some cool new efficiencies in each core it doesn't look like the 10th gen Intel CPU's are going to add much. They natively run 3200MHz DRAM, which means they "may" handle memory at a higher clock speed better. I'm currently running my DRAM at 3200MHz on my i9-9900K. The i9-9900K equivalent is still limited to 4.8 boost CPU clock speed with all CPU's running. They have new 10/20 core chips, but they are limiting the clock speeds quite a bit such that they will probably run games slower than the 8/16 chips that have a higher per core speed.
I just went to the 3600X yesterday for my sims and WMR. It is so fast, I just couldn't justify the additional cost of the 3700, 3800 or 3900 series. Coming from the 2600X... which I was already satisfied with, this was just a bonus. I think all the talk about additional fps, is okay...but of more concern to me is absolute smoothness. I am quite happy with framerate in all my sims. I loved the smoothness in VR...even with the 2600. It was just a great experience. The 3600 test I ran last night confirms it for me. Great CPU...and that's on an X470 board to boot.
That one is gonna need a new MB anyway, so I don't even care about nextgen Intel CPUs. My plan for the Intel option is to build a system that will last me ~5 years with only a GPU upgrade in the middle of the road. This approach worked just fine with 3570K. Also, DDR4 is likely to be superseded by DDR5 in a few years, so investing into an upgradable DDR4-based platform could be a moot point. What good would a hypothetical socket 1155 8-core 5GHz CPU do me if DDR3 would still be constraining its performance
If I already had an AM4 system that would be a no-brainer for me as these new chips made a big jump compared to 2000 series and you only need to get a new CPU. In my case there's also an MB and RAM added to the total cost of the upgrade, so yeah, I wanna to get the best bang for my buck here
Their 9700K usage never went above 49% in that PC2 test, so I guess it was not the extra HT cores that made the i9 faster there. Could be higher boost clock or 33% larger L3 cache that made the difference
In real VR test cases I can tell you that with all the i9-9900K cores running 4.8GHz wirh 4 x 8Gb 3200MHz DDR4 memory I can max out one or two of it's cores easily just by tweaking the super sampling. The key problem is still lack of parallel processing by the game engines because there is plenty of extra CPU horsepower being wasted.
