AMD 3700x vs Intel 10700k if priced the same

[rfxcasey]

Building a new PC, I'm torn between an AMD 3700x build or a Intel 10700k build. With previous pricing, it was kind of a no brainer with the scales tipped in AMD's favor, but just today, I'm seeing 10700K on sale for 320 USD which is 10 dollars cheaper than the 3700x. All 5000 series AMD chips are out of stock everywhere so that's not really an option right now.

 

AMD's socket AM4 has been out since 2016, though it's promised that the higher end chipsets, while supporting PCI_E 4.0, will support Zen 3 5000 series CPUs. Intel's higher end chips boast Gen 11 compatibility but not PCI_E 4.0. Early test and analysis say PCI_E 4.0 won't be a factor for the foreseeable future, however M2 SSDs will benefit from PCI_E 4.0 currently, the trick will be keeping them cool enough to not throttle down and kill performance.

 

EDIT: It appears that, while not listed in specs, the Intel z490 chipsets may support PCI_E 4.0 with a BIOS update.

 

Intel's 10700K seems to have a distinct advantage over the 3700x in many games and even some, but not all, productivity tests. DCS however, is not really a traditional 'game' and it's been my experience, that Intel CPUs, traditionally, show a rather strong performance advantage in their given market category. Don't really plan on flying MS Flight Simulator 2020 but some early benchmarks suggest a clear advantage for Intel.

 

If the AMD 5900x or 5950x will be sub or around 400 USD in the next 4 years, going with a 3700x and z570 chipset motherboard now might make sense, to sell the 3700x on and shell out the extra cash for the 5000 series. It remains to be seen how Intel's 11th gen will stack up in both price and performance.

 

I've got two builds on the way now, an AMD 3700x with z570 MSI Tomahawk WiFi and an Intel 10700k with z490 Tomahawk non Wifi on the way. Either build will receive a Noctua NH-U14S heat solution with a 140mm fan, which will be carried over to any future processor upgrades. The system will have 32GB of 3200Mhz RAM. Video card is an RTX 2060 Super, monitor is 2560x1080 plus possibly X2 1400x900 side monitors (yes a 2080 ti or 3080 would make more sense for triple display). 75Hz max with G-Sync, it's perfectly fine for me, I don't own a 120Hz or higher refresh monitor.

 

I'm feeling the Intel will be stronger for DCS overall, but the 3700x may actually outperform in the overall picture when running TrackIR, HOTAS software, voice communication software, hosting and or recording or live streaming. Please reply with your thoughts. 

Edited December 23, 2020 by rfxcasey

[LucShep]

I'm obviously biased (see my sig) but all I can say is that I'm extremely satisfied with my i7 10700K(F), couldn't be happier.

The R7 3700X is still an excelent processor, no doubt, but the i7 10700K is the better processor of the two.

At 320,00 USD, and even cheaper than the 3700X ? ....man, that's a no brainer.

 

The MSI Z490 Tomahawk and Gigabyte Z490 VisionG are definitely superb choices, well featured and worth the ~190,00 USD price tag.
But if you'd prefer to get a more affordable Z490 mobo for the 10700K, then you also have two excelent choices in the form of the MSI Z490-A Pro (what I got) and the Asus Z490-P, either works perfect with the i7 10700K and are priced at ~150,00 USD. 

Getting a decent cooler is par for the course and, for sure, a good air cooler like the Noctua NH-U14S will do mighty fine (got a Scythe Fuma2 myself). 

 

The i7 10700K is more or less an improved i9 9900K (slightly cooler and faster), so perhaps the following vid can add more bits of useful info on your choice.
It's comparing the i9 9900K vs R5 5600X vs R7 3700X:
 

 

Edited December 23, 2020 by LucShep
...spelling (?)

[rfxcasey]

Thanks for the input. One thing I forgot to mention, thermals. So far, from what I've seen, the 3700x seems to run cooler, though results might have lot to do with the motherboard VRM. I'm old skool. Been building PCs since probably '91 and while many concepts and lessons have remained the same overall, many things have changed. With the newer more 'intelligent' processors and even motherboards, controlling heat seems to be a bigger factor in maximizing performance then ever. Back in the day, you just set it and let it ride, especially when overclocking. I suppose you can still do that overriding the automation of the system but it's surely not the best way to go about it. I read that MSI had a problem with some of the VRMs on their previous motherboards so they completely overcompensated on the x570 Tomahawk, giving it a VRM that surpasses anything even in their highest gaming motherboard options. Thermal tests on the z570 Tomahawks VRM vs the Asus TUF mobo for instance so just how efficient the Tomahawk's module is. I don't plan on overclocking, but the cooler the better which will allow the CPU to do it's thing and boost as high as possible. I do not believe the VRM on the z490 Tomahawk is even close to being as good.

 

https://youtu.be/O2KbpmMg44M

 

 

Edited December 23, 2020 by rfxcasey

[LucShep]

On the subject of Z490 VRM quality, I suggest watching Hardware Unboxed on youtube (or Techspot, where they place those episodes in text), they extensively tested this aspect on Z490 motherboards across different segments, with stock and OC'ed i9 10900K as well (so, most extreme scenario). MSI ones have excelent VRM solutions in all price segments.

This summer I assisted building a rig with a friend, he's got an i9 10850KA and a Z490 Tomahawk, runs it overclocked at 5.1 all core (on a 360AIO) and seems really happy with his system. That board got doubler phases (I believe) and in practice equivalent to a good 12+phase VRM solution, very robust and kind of proven among OC'ers if you search around. It's a safe bet, that and the VisionG.

As for the MSI Z490-A PRO that I went for (did my bit of research), and although I wouldn't imediately recommend it for an harcore OC'ed i9 10850K or 10900K, it certainly gets my vote of aproval for the i7 10700K even if overclocking (managed 5.0 all core at 1.3v).

 

About overclocking, the funny thing is, for the first time in about 20 years here I find that overclocking results, for real life PC usage, are not felt as much in practice (for me) as it used to be years ago with other CPUs, and at detriment of a lot more heat. 

It seems to me that these newer Intel CPUs are now better "tweaked", more pushed (correct term?) and fine balanced more than ever, so good already from factory, to a point that I feel OC'ing becomes more of a fun thing as hobby (gains are there, but not that big), whereas before it used to be about extracting substancial hidden speed.

 

Got my i7 10700KF back to stock (again) just slightly undervolt, as I found the factory optimal settings of BIOS has voltages way too high.

Right now I'm using Vcore at 1.27v, VCCIO at 1.15v and VCCSA 1.20v.

CPU idles at 23ºC(!!), games run at 45ºC~50ºC (usually around 25W~60W in gaming, depends on game optimization). Can run benchmarks and extreme stress test all day long, stable with same good scores, most that I've seen is 75ºC in CPU Package (that's with turbo all core kicking, near 200W ocasionally!) and VRMs sensors never gone over 70ºC.

More performance than I envisioned needing, installation was a breeze, no fuss, no issues. To say that I'm satisfied is an understatement.

It may be considered older architecture based (14nm+++++++) but it's a really good, well refined processor. With such prices, it's a great deal IMO.
 

Edited December 23, 2020 by LucShep
...spelling (?)

[rfxcasey]

Wow, interesting indeed, have you recorded any performance loss or does it actually help the CPU boost higher keeping the temps down. I haven't built a machine since my i5-3570k, every so many years I have to brush up on the new tech, I don't build them for a living anymore. Any info you can point me too on the methods used to determine the optimal power setting, is it just a matter of incremental changes and testing or is there a more scientific way? Can you give the specs on your machine, video setting and DCS performance? What resolution are you running?

[LucShep]

12 hours ago, rfxcasey said:

Wow, interesting indeed, have you recorded any performance loss or does it actually help the CPU boost higher keeping the temps down. I haven't built a machine since my i5-3570k, every so many years I have to brush up on the new tech, I don't build them for a living anymore. Any info you can point me too on the methods used to determine the optimal power setting, is it just a matter of incremental changes and testing or is there a more scientific way? Can you give the specs on your machine, video setting and DCS performance? What resolution are you running?

Recorded I don't have as I formated my PC recently, but I do recall geting slightly higher consistent scores across different benchmarks, compared to stock voltages (f.ex, CB R15 now 2115, before 2080, and CB R20 now get 5065, before 5040), though maybe I should consider that as margin of error.

Temperatures with heavy CPU usage is where it's really noticeable, benefits are clear and there are no disadvantages.

From memory, the stock voltages (on Auto) were Vcore at 1.34v, VCCIO at 1.24v and VCCSA 1.28v, and that with all core turbo would make CPU package get ~90ºC after a minute when stressed. That's stupid overvolted from factory, and it seems it's same across most motherboards.  Same thing for MCE, on boards that have this silly feature, to turn it off is a must do.

I suspect they make it so as to ensure performance+stability with different bin qualities of these processors (silicon lottery), but it's way overdone, really excessive.

 ....the "runs hot" fame of Intel is surely inflated by this.  

VCCIO at 1.15v and VCCSA 1.20v is the sweet spot and good for 99,9% of RAM, it seems (BTW, I got Crucial Ballistix 3200 CL16 bumped up at 3466Mhz and 1.40v Dram Voltage).

The info around is to never surpass 1.30v on either of those VCCIO and VCCSA settings (anything over 1.25v in those is excessive, IMO), to avoid long term CPU degradation.

 

As for Vcore, ~1.27v should work perfect for any 10700K at stock clocks (the 1.34v of Auto is outrageous!), as I'm also finding from others lurking on forums and Reddit.

With this done, temperatures (and wattage) then are more contained, actually surprisingly low, nothing like tech reviewers make you believe.

"Test and you'll see", is all I can recommend.

Edited December 23, 2020 by LucShep
...spelling (?)

[rfxcasey]

Well, I'm going to use your posted changes as a starting point, or maybe start just a little less aggressive, test for stability and then move it down to match you and re test. I tend to lean on the milder side of deviating from the norm, though I did OC my i5 3570K from 3.8Ghz to 4.4 out of the box with only the slightest V core bump and it's been rock stable since around 2012. Cool and stable while attaining best performance is how I like to run. My 3570K will be getting handed down, I hate to see it go, in reality it is still hanging in there. Heck, I'm temped to just build both rigs and do some testing but I don't know, I'll think about it. If I do, I'll do some comparisons and do a vid or write up. I'm still torn though, Passmark show scores the 3700x just above the 10700K, probably a little more efficient multicore handling, either one of these builds is going to get the job done. Nice thing about the 3700x is it's 65 watts. 

[nalpagutt]

As someone using AMD I would recommend intel.10700K Performance is really great.

Edited December 25, 2020 by nalpagutt