• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Review 6700k vs 3700x in 23games

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Head1985

Golden Member
Jul 8, 2014
1,760
496
136
Not entirely true. You can either do a lot of goofy stuff to try and amp up your boost clocks (use PBO, fiddle with LLC settings), or you can go for static OC which will get you some decent gains in games that spawn a fair number of threads. I think 4.2 GHz static is attainable on almost any 3700x in games. I know that my 3900x, despite being able to boost to 4.6 GHz, runs games faster with a static OC of 4.4 GHz (or even 4.375 GHz) since the boost duration is so short. Max boost for a 3700x is supposed to be 4.3 GHz and PBO + Auto OC is often quite useless. If I had a 3700x, 4.2 GHz static on all cores would be my target for games. Or at least 4.15 GHz.
All core oc is not worth becuase:
1-cpu wont downclock in iddle and is always on max voltage.
2-i can oc to 4250mhz, but performance is same as with all settings default:) so its not worth it because cpu will degrade very fast without downclocking/iddle voltage.
 

Gideon

Senior member
Nov 27, 2007
666
859
136
. If I had a 3700x, 4.2 GHz static on all cores would be my target for games. Or at least 4.15 GHz.
Having a 3700x, I wouldn't really recommend that ,at least not for all.

With everything at stock, all of my cores are usually at 4.275 GHz (fluctuating between 4.25 - 4.30 GHz). And yes, that is for the entire gaming session, in games like BF1, Metro Exodus. With PBO the clocks are fixed @ 3250Mhz and pretty much never move in any direction. Without PBO, most of the cores are over 4.25 Ghz, in some segments all are at 4.3GHz (none of that ever happens with PBO).

Some anectodal screenshots:
Without PBO (Stock):


Without PBO (stock) less taxing games:


With PBO:

As you can see, even in quite multi-threaded games the CPU load isn't all that great (at least with my 1070 GTX)

Now, of course Cinebench (R20 Improves from ~4750 to ~4950 points) and other loads, using near 100% of the CPU improve considerably with PBO (and manual overclocks), but AFAIK, games do not utilize the cores to the same degree. Also, bear in mind, that at stock the CPU limited to a nice, never-to-exceed 88W PPT limit, which is dead-silent on my Noctua NH-U12S. PBO draws 144W+ and is nowhere near silent, yet in most games the performance actually takes a hit.

This might be because of my relatively weak GPU (1070 GTX) but it shows that even PBO, much less manual overclock aren't always helping, rather the opposite
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,434
136
All core oc is not worth becuase:
1-cpu wont downclock in iddle and is always on max voltage.
Not true. At all. If you leave UEFI clockspeed/voltage at stock (except for LLC and other VRM settings) and use Ryzen Master to set your clocks + vcore, then you will not be pegged to max voltage all the time. Additionally, you will not see voltage spikes above the voltage you set. No more "why is my CPU at 1.5v wahhh" type of thing. The only time I've ever gotten better performance from boost was in something perversely single-threaded like SuperPi, and I had to tweak a bunch of LLC settings very carefully to get it to boost reliably. It was weird and totally not worth the effort.

2-i can oc to 4250mhz, but performance is same as with all settings default:) so its not worth it because cpu will degrade very fast without downclocking/iddle voltage.
I find that hard to believe, but hey, I guess your board likes to boost? Mine sure doesn't. No way it's hitting 4250 MHz on any significant number of cores with just boost. I think the best I ever did was 4180 MHz. Maybe if I stuck everything at defaults and moved LLC down to Low, I could get there.

Having a 3700x, I wouldn't really recommend that ,at least not for all.
Have you tried performance-based benching between boost behavior and static OC? I did on my setup, and PBO was the big loser with stock coming in at second place. First place was a conservative all-core OC to 4.3 GHz. Note this was without me trying any LLC hyjinx to try and fool the boost algorithm into changing its behavior. I found that boost was more aggressive with lower LLC settings, while PBO was more aggressive with higher LLC settings.
 

DrMrLordX

Lifer
Apr 27, 2000
14,457
3,434
136
I want to add an addendum to my above post: in order to make my motherboard/CPU combo (3900x + x570 Aorus Master, 8/2/2019 UEFI with 1.0.0.3ABB) boost to the highest possible all-core/multicore clocks, I had to:

1). Set CPU and SoC LLC to "Normal" which is basically, LLC OFF
2). Set a voltage offset of - .1000v

That got my boost clocks in games up into the 4250 MHz territory. It also got me a CBR20 score of just over 7300, which is the highest i've gotten under this particular UEFI. I think I could do a bit better if I lowered RAM clocks to DDR4-3200. Maybe. But who wants to do that?
 

Kyron71

Member
Sep 5, 2019
76
8
11
But remember the 6700 was 2nd from tops when it came out.

3700 is down the list a bit for AMD.
 

StrangerGuy

Diamond Member
May 9, 2004
8,418
118
106
For those who want to see what single thread CPU-limited gaming really is like, try CS:GO with 31 bots: My 3600 gets 63% more avg FPS here than my old 4790K.
 

kasakka

Senior member
Mar 16, 2013
334
1
81
For reference, these are the scores I get with Shadow of the Tomb Raider benchmark with a 6600K @ 4.5 GHz and 2080 Ti at 1080p. Raytracing was off, ultra settings, DX12. Windows 10 64-bit version 1809.

Code:
CPU g = CPU game
CPU r = CPU render

Fps  | CPU g | CPU r | GPU
Min  | 110   |  72   | 129
Max  | 248   | 393   | 348
Avg  | 169   | 127   | 172
95%  | 124   |  75   | 139
This is almost entirely CPU bound, 7% GPU bound according to the stats. Going up to 4K makes it 79% GPU bound.
Upgraded to 3700X on Gigabyte X470 Aorus Gaming 7 and 3200 MHz 16 GB. Stock settings so far, just early testing. Anyway, the same test was run and the results were the following:

Code:
CPU g = CPU game
CPU r = CPU render

Fps  | CPU g | CPU r | GPU
Min  | 85   |  135   | 130
Max  | 156   | 335   | 252
Avg  | 120   | 194   | 162
95%  | 89   |  143   | 134
Interestingly I see drops in some but the CPU render values are up a lot. Perceptually the game runs a lot smoother at the usual 4K DLSS + RTX setting I use.
 
  • Like
Reactions: moinmoin

Head1985

Golden Member
Jul 8, 2014
1,760
496
136
I retested some games with ABBA bios + with 3600Mhz ram running at 1:1:1.In some games there is huge perf increase.Also i am not sure with war thunder cpu test.There was update since i tested it.But i dont think performance increase was because update.
Here are results:
2019-10-01 (2).pngac oddy.pngcars2.pngfallout.pnghitman.pngking.pngmafia.pngrise.pngshadow.pngwar.png
 
Last edited:

besset

Junior Member
Oct 26, 2019
2
0
6
So does anyone know if the ryzen 3000 series perform worse in other Bethesda games like skyrim SE? I play a lot of modded fallout 4 and skyrim SE and im planning on getting a ryzen 3600 but after seeing the poor ryzen performance I might go for a intel CPU. And since we can play fo4 and skyrim in high refresh with the havok fps fix mod it might be worth it for those extra frames if I'm gonna heavily mod. The ryzen 3000 series perform well in almost all other games though...🤔
 

thesmokingman

Platinum Member
May 6, 2010
2,168
88
91
All core oc is not worth becuase:
1-cpu wont downclock in iddle and is always on max voltage.
2-i can oc to 4250mhz, but performance is same as with all settings default:) so its not worth it because cpu will degrade very fast without downclocking/iddle voltage.
That is so wrong. You can all core oc while keeping power savings. The advantage with all core vs letting the cpu do its own thing ie. stock/pbo/etc etc is for the multicore IPC. All core is faster in highly threaded and sustained workloads. If all you do is play games it's probably better off w/o an all core oc and just running stock.
 

Elfear

Diamond Member
May 30, 2004
6,904
315
126
So does anyone know if the ryzen 3000 series perform worse in other Bethesda games like skyrim SE? I play a lot of modded fallout 4 and skyrim SE and im planning on getting a ryzen 3600 but after seeing the poor ryzen performance I might go for a intel CPU. And since we can play fo4 and skyrim in high refresh with the havok fps fix mod it might be worth it for those extra frames if I'm gonna heavily mod. The ryzen 3000 series perform well in almost all other games though...🤔
Good question. I'd like to hear the answer to this too. The Ryzen 3000 series still isn't as good as Coffee Lake with drawcalls but how often do drawcalls cause the performance bottleneck while playing modded Skyrim SE and FO4?
 

thesmokingman

Platinum Member
May 6, 2010
2,168
88
91
So does anyone know if the ryzen 3000 series perform worse in other Bethesda games like skyrim SE? I play a lot of modded fallout 4 and skyrim SE and im planning on getting a ryzen 3600 but after seeing the poor ryzen performance I might go for a intel CPU. And since we can play fo4 and skyrim in high refresh with the havok fps fix mod it might be worth it for those extra frames if I'm gonna heavily mod. The ryzen 3000 series perform well in almost all other games though...🤔
I can't imagine that it performs worse. I'd recommend not using an all core overclock instead running stock so you get your max single core boost which will pay dividends in Skyrim, which is really a terribad single thread turd. The IPC difference between Intel and AMD is rather small today, so its up to you.
 

Arkaign

Lifer
Oct 27, 2006
20,365
752
126
So does anyone know if the ryzen 3000 series perform worse in other Bethesda games like skyrim SE? I play a lot of modded fallout 4 and skyrim SE and im planning on getting a ryzen 3600 but after seeing the poor ryzen performance I might go for a intel CPU. And since we can play fo4 and skyrim in high refresh with the havok fps fix mod it might be worth it for those extra frames if I'm gonna heavily mod. The ryzen 3000 series perform well in almost all other games though...🤔
Do you do much encoding or multimedia production?

If yes, get a 3600 or better yet a 3700X.

It will be very competent in gaming, and superior if you run a lot of other heavy MT apps that you find yourself waiting on.

If gaming is your primary use, get a 9700K.

You can run it at 4.8-5Ghz under a good air cooler, quietly and with nice temps, and have the best gaming performance available.
 

Guru

Senior member
May 5, 2017
741
267
106
It figures that older games with ancient engines like Crysis 3 and Fallout 4 perform better on the 6700k, nothing to see here. Fallout 4 engine is the same engine from Fallout 3, meaning 10 years old and it figures with Fallout 4 being extremely buggy and Fallout 76 being extremely buggy as well.

Crysis 3 came out in 2013, it was in development for 1.5 years on their Cryengine, which was only slightly updated from their Crysis 2 engine which came out in 2011.

So yeah, 4 cores at 4.5GHz and lower latency in 2 and 4 core optimized engines does perform better.

It's only been about 3 years until games started using more than 6 cores, even Windows was so behind the time that it could only run 4 cores decent. So yeah, it wasn't until recently that Win10 was updated to better handle more than 4 cores. Before that it was pathetic.
 
  • Like
Reactions: Tarkin77

ASK THE COMMUNITY