cancel
Showing results for 
Search instead for 
Did you mean: 

AMD Ryzen 9 3900X vs. Rysen 7 3700X and Intel i9-9900K, i7-9700K, i7-8700K and i7-7700K

RuneSR2
Grand Champion
9u9h75uy94yb.jpg

Feels like hard work checking out latest cpus, because it doesn't interest me much. And why? - Because using the high-end gaming cpus released during the last 2 years or more VR games should work great and not be cpu bound (bottlenecked). Then again AMD seems on to great stuff with their new Ryzen 9 3900X cpu and let's see how it goes. 

Note that a common mistake is to test cpus using high res in games - doing that you risk transforming what should be a cpu test into a gpu test. Imagine testing new cpus in 4K with 8xMSAA in modern games using a GTX 1050 - then when could show that a cpu made 10 years ago is just a fast as an i9-9900K, because they result in exactly the same slideshow 😉 

It's important to find benchmarks that aren't gpu bound - or rather gpu bound as little as possible. Anandtech does offer tests done in 720/768p, which is nice - but in theory I'd like gaming results from even lower res, but many games may no longer support lower res than 720p. Remember that increasing the res only affects the GPU, the cpu calculates the same amount of polys in 640x480 as in 7360x4360 (8K), but the GPU does not at all have same workload when res increases. Thus res should always be as low as possible when testing cpus - and the video card should be as fast as possible to make it possible for us to see real cpu differences. 

Let's start with Tech Power Up - 720p using a 2080 Ti:

vgveeb1d5w5a.png
Source: https://www.techpowerup.com/review/amd-ryzen-9-3900x/

The results do indicate that the Ryzen 9 3900X is performing great - there's close to no difference between high end cpus, at least not something easily spotted with the naked eye. But the cpus could all be bottlenecked by the 2080 Ti even in 1280x720. So for gaming just buy the cheapest solution? Unfortunately Tech Power Up didn't include the i7-7700K, but Anandtech did. Let's check up and try to verify (or falsify) TechPowrUp's results:

World of Tanks
19i3hcugsswh.png
Setting 3900X to 100% (like TechPowerUp) we see:

i9-9900K = 109 %
i7-9700K = 105 %
i7-8700K = 102 %
Ryzen 9 3900X = 100 %
Ryzen 7 3700X = 99 %
i7-7700K = 93 %

You could gain 17 % going from 7700K to 9900K, but it'll probably translate to much less in the real world and in VR games, which usually are GPU limited. But interesting that high-end cpus can push more than 600 fps in that game. 

Shadow of War
k42ax04npkxf.png
Setting 3900X to 100% we see:

i7-8700K = 103 %
i9-9900K = 103 %
i7-9700K = 101 %
Ryzen 9 3900X = 100 %
Ryzen 7 3700X = 99 %
i7-7700K = 82 %

Here the i7-7700K is about 25% slower than the fastest cpus, so of course something could be gained... 

Ashes of the Singularity
 b6y57ye3wd2r.png
Setting 3900X to 100% we see:

i9-9900K = 105 %
i7-9700K = 103 %
Ryzen 9 3900X = 100 %
7-8700K = 99 %
Ryzen 7 3700X = 98 %
i7-7700K = 84 %

Going from 7700K to 9900K 25% may be gained in this RTS game. 

Strange Brigade (Vulkan)
ihy4nwktlbkl.png
Setting 3900X to 100% we see:

i9-9900K = 104 %
i7-9700K = 102 %
Ryzen 9 3900X = 100 %
Ryzen 7 3700X = 100 %
7-8700K = 98 %
i7-7700K = 85 %

About 22% can be gained going from 7700K to 9900K. 

GTA-V
8hxapxljqwta.png
Setting 3900X to 100% we see:

i7-9700K = 102 %
Ryzen 7 3700X = 101 %
i9-9900K = 101 %
Ryzen 9 3900X = 100 %
7-8700K = 97 %
i7-7700K = 88 %

About 15% can be gained going from 7700K to 9900K. 

F1 2018
4c272cerw653.png
Setting 3900X to 100% we see:

Ryzen 9 3900X = 100 %
i9-9900K = 98 %
Ryzen 7 3700X = 95 %
i7-9700K = 94 %
7-8700K = 87 %
i7-7700K = 78 %

About 26% can be gained going from 7700K to 9900K. 

These were the 6 games benchmarked by Anandtech. Averaging all Anandtech's results we get:

i9-9900K = 103 % (range 98 - 109 %)
i7-9700K = 101 % (range 94 - 105 %)
Ryzen 9 3900X = 100 % (range 100 - 100 %)
Ryzen 7 3700X = 99 % (range 95 - 101 %)
i7-8700K = 98 %  (range 87 - 103 %)
i7-7700K = 85 % (range 78 - 93 %)

Here compared to TechPowerUp's average results (from 10 games):

i9-9900K = 107 % 
i7-9700K = 106 %
i7-8700K = 103 % 
Ryzen 9 3900X = 100 % 
Ryzen 7 3700X = 96% 
i7-7700K = not tested

If we now combine TechPowerUp's results with Anantech's results we get these final results:

i9-9900K = 105 % 
i7-9700K = 104 % 
i7-8700K = 101 %  
Ryzen 9 3900X = 100 % 
Ryzen 7 3700X = 98 % 
i7-7700K = 85 % (only tested by Anandtech, probably will result in less than 3% error margin)

Thus going from 7700K to 9900K may result in about 25 % performance in cpu bound games - but since nearly all current VR games are gpu bound - and even an i7-7700K may be able to produce way more than 90 (and 144) fps, benefits of upgrading cpu to 9900K or Ryzen 9 3900X may be very limited or non-existent if you've got a 7700K, similar or faster cpu. 

For now I'll continue not to care about cpus 😉 

BTW - power consumption reveal something interesting:

fqawmia3s0ce.png
Suddenly 7700K still shines 😉 And 3700X too. Setting 3900X to 100% we see (lower is now best):

Ryzen 7 3700X = 63 %
i7-7700K = 67 %
i7-9700K = 87 %
Ryzen 9 3900X = 100 %
i7-8700K = 106 %
i9-9900K = 118 %

Suddenly Ryzen 3700X looks like king of the hill - it's performing so close to 9900K in games that you'll probably never know the difference, but the 9900K uses 87% more power than 3700X, quite amazing that AMD could achieve such amazing performance per watt with the 3700X!

Source (Anandtech): https://www.anandtech.com/show/14605/the-and-ryzen-3700x-3900x-review-raising-the-bar/

caoio83hyqbr.jpg

Oculus Rift CV1, Valve Index & PSVR2, Asus Strix OC RTX™ 3090, i9-10900K (5.3Ghz), 32GB 3200MHz, 16TB SSD
"Ask not what VR can do for you, but what you can do for VR"

17 REPLIES 17

RedRizla
Honored Visionary


When the next gen gaming consoles come out, i suspect cpus will matter much more. For now a 4 core will be just fine. But in 2021 your 4 core i7 7700k maybe due for an upgrade.


Why? What game do you think will use all 4 cores?

cleanupdisc
Adventurer

RedRizla said:



When the next gen gaming consoles come out, i suspect cpus will matter much more. For now a 4 core will be just fine. But in 2021 your 4 core i7 7700k maybe due for an upgrade.


Why? What game do you think will use all 4 cores?


Games that are multiplatform that come out with the ps vr 2. With the new cpu architecture of zen 2 we can see intel is being very competitive also. We had 4 cores 8 threads for like 7 years straight.
https://youtu.be/ozM8p6c-Ljc

Now we are seeing advancements with competition. Your 4 cores are fine right now but when intel is releasing 6 core 12thread i5s, your 4 cores turn into an i3.

Either way your 4 core will be fine for gaming for a good while. Just dont think its gonna be amazing for the next 6 years!



RuneSR2
Grand Champion
Just had some fun benchmarking a few rigs - using the good ol' Heaven benchmark:

https://youtu.be/uEXv8A0-_KU

Testing in 1920x1080, fullscreen, extreme tessellation, ultra setting and 8xAA, my 17-year old stepson's $2.5k MSI Leopard laptop (i7 9750H, RTX 2060, 16GB) got 82 fps. (I think he bought that laptop like 6 or 9 months ago - it may be cheaper now.) 

Now - just starting to build a rig for my 12-year old son, he's now using an i5 10400, 16GB and is using the integrated Intel UHD 630 gpu - and he's only using 1280x1024 (using an old Sony lcd from my *cough* storage, he's paying mostly himself for the rig). Even in 1280x1024 Heaven wasn't kind to him - he got 7.2 fps :# Then again, UHD 630 isn't exactly built for that kind of tessellation and 8xAA, lol. Minecraft runs like a dream on the 10400 😉 (He's waiting for me to upgrade so he can borrow my 1080)

Fun thing - thinking that the i7 9750H + RTX 2060 laptop in theory might be close to my VR rig bought in 2017 (oc'ed GTX 1080, i7 7700K, 16GB) - I also ran the Heaven benchmark in 1920x1080 using the same extreme settings - and I got 122 fps. That's about 50% faster the the MSI Leopard laptop (which even used high performance settings). 

In short, maybe there're some great laptops out there, but do be very careful what you buy! 

Another fun thing, Heaven eats gpu power like there's no tomorrow, but decreasing the res a lot will turn the benchmark into a cpu test. In the lowest res - like 640x480 - still using extreme settings - my i7 7700K managed to get 277 fps. 

Now I'm wondering if changing to RTX 3090 will make 277 fps possible - in 1920x1080, 4K will be something completely different, lol. (Some Reddit dude managed to get 207 fps from his 8700K using RTX 3080, so 277 fps might be too much even for RTX 3090, maybe... )

Oculus Rift CV1, Valve Index & PSVR2, Asus Strix OC RTX™ 3090, i9-10900K (5.3Ghz), 32GB 3200MHz, 16TB SSD
"Ask not what VR can do for you, but what you can do for VR"

TomCgcmfc
MVP
MVP
@RuneSR2  Best to stay away from so-called gaming laptops, esp. for VR imho (been there, done that btw).  For what your stepson paid for his fairly high end laptop he could currently build a much better gaming desktop PC.  Even with a i7 9700k, rtx 3070, 16-32 Gb 3000-3600 ram, and an inexpensive 24in 120-144hz gaming monitor he would have quite a few $'s left over.  Maybe even enough to by an inexpensive laptop for school use.
i9 13900K water cooled, RTX4090, Z790 MB w/wifi6e, 32Gb 6400 ram, 2x2TB SSD, 1000W PSU, Win 11, QPro, Q3, w/Link and Air Link, Vive Pro1 with Etsy lens mod and Index Controllers

RuneSR2
Grand Champion

TomCgcmfc said:

@RuneSR2  Best to stay away from so-called gaming laptops, esp. for VR imho (been there, done that btw).  For what your stepson paid for his fairly high end laptop he could currently build a much better gaming desktop PC.  Even with a i7 9700k, rtx 3070, 16-32 Gb 3000-3600 ram, and an inexpensive 24in 120-144hz gaming monitor he would have quite a few $'s left over.  Maybe even enough to by an inexpensive laptop for school use.


 Fully agreed, but he often plays with friends and wanted something easy to carry (it's quite light, guess that's expensive too). He was informed back then that it was his money and choice, I think he liked the design of the laptop a lot too 😉 

By the way, my 12 yo son got 1 x 16GB 3200 MHz ram in his rig - unfortunately that's a bad thing, because the UHD 630 needs dual channel ram to fly:

https://youtu.be/s_2Z208nDGE

Even using a Nvidia gpu dual channel ram may provide some extra performance - a lot of people seem unaware of this. 

Actually I ordered 2 x 8GB 3000 MHz for the son's rig - the dudes assembling the rig probably thought they did me a favor by upgrading 2 x 8GB 3000 MHz to 1 x 16GB 3200 MHz, sigh. 

But then I got a great idea - I ordered another 16GB 3200 MHz stick - then my son gets my 2 x 8GB 3200 MHz, and I get 2 x 16GB 3200 MHz  B) He can keep using the 2 x 16GB until my RTX 3090 and 750w psu arrive, lol - not sure 32GB really is needed in my current games. 

Guess that's my upgrading goals for 2021 - RTX 3090, 32GB 3200 MHz and a new and shiny Corsair SF750 PSU, lol. i7 7700K still seems fine, haven't even tried to overclock it yet.

Oculus Rift CV1, Valve Index & PSVR2, Asus Strix OC RTX™ 3090, i9-10900K (5.3Ghz), 32GB 3200MHz, 16TB SSD
"Ask not what VR can do for you, but what you can do for VR"

RuneSR2
Grand Champion
This is bad - this is performance using single and dual channel mem even with a 2080 Ti (SC = Single Channel, DC = Dual Channel):



"As expected, Assassins’ Creed chokes really hard when using a 2400MHz, single-channel configuration. Even the mighty 3900X can’t hit an average of 60 FPS (with an RTX 2080 Ti) at 1080p in Origins, with lows of just 20. Switching to a dual-channel config, more than doubles the lows and pushes the average to over 80 FPS. It’s interesting to note that a dual-channel 2400MHz kit is better than a single channel 3600MHz config."

I wonder how many VR users are aware of this...

Source: 
https://www.hardwaretimes.com/dual-channel-vs-single-channel-ram-which-one-is-better/ ;

It may seem like a good idea getting like 1 x 16GB now and leaving a slot free for future upgrades - but not with the above results for sure!

Also the difference between 2400 MHz and 3600 MHz is kinda massive. I did measure 10% more cpu performance after changing from 2400 MHz to 3200 MHz. 

Comparing rigs isn't necessarily that easy - a lot of factors need to be included 😉

Oculus Rift CV1, Valve Index & PSVR2, Asus Strix OC RTX™ 3090, i9-10900K (5.3Ghz), 32GB 3200MHz, 16TB SSD
"Ask not what VR can do for you, but what you can do for VR"

OmegaM4N
Expert Trustee
Dual channel is at times almost doubling the avaible bandwidth so i can see why it would beat a single channel of even faster memory, but then you got stuff like latency and if the game itself has been optimized correctly, of have they just threw it out the door and allowed brute force to fix a lot of stuff for them, as they seem to do with most games these days. ha
CV1/Vive-knuckles)/Dell Vr Visor/Go/Quest II/ PSVR.

RuneSR2
Grand Champion

OmegaM4N said:

Dual channel is at times almost doubling the avaible bandwidth so i can see why it would beat a single channel of even faster memory, but then you got stuff like latency and if the game itself has been optimized correctly, of have they just threw it out the door and allowed brute force to fix a lot of stuff for them, as they seem to do with most games these days. ha



Added another 16GB stick to the son's rig today - so he's running 2 x 16GB 3200 MHz (same latency, identical sticks, dual channel confirmed by CPU-Z) with the very slow UHD 630, lol. Heaven Benchmark instantly went up 25%, but he says that Minecraft is now solid 60 fps, while it before some places went down to about 24 fps - of course that's just impressions and not a real benchmark. 

Oculus Rift CV1, Valve Index & PSVR2, Asus Strix OC RTX™ 3090, i9-10900K (5.3Ghz), 32GB 3200MHz, 16TB SSD
"Ask not what VR can do for you, but what you can do for VR"