• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

3800x vs 9900k

This is interesting, I really had to look for a video with result I could compare too. https://www.youtube.com/watch?v=DIhLmWMnAgA

Test Bench: Intel 9900k Processor Overclocked to 5.0ghz All Cores No AVX Offset (1.27V)
Corsair H150i PRO 47.3 CFM Liquid CPU Cooler (Fans @ 100%)
Asus ROG STRIX Z390-E GAMING ATX LGA1151 Motherboard
Corsair Dominator Platinum RGB 16 GB (2 x 8 GB) DDR4-3200 Memory (16-18-18-36?)
Samsung 970 Evo 250 GB
M.2-2280 SSD (Boot)
Samsung 850 EVO 250 GB 2.5" SSD (Games)
Asus ROG Strix RTX 2080 OC Edition (GPU Boost Clock @ 1970MHz)
Corsair RMx 650W Power Supply Corsair 750D Airflow Edition Case

Benchmark Tool. Results:
Average Framerate (99th percentile): 64.45 Max.
Framerate (99th percentile): 93.17 Min.
Framerate (99th percentile): 38.37

Settings: DirectX: DX12 Resolution: 1920x1080p Quality: Ultra Texture Filtering: AF 16X Motion Blur: Normal Tesselation: Full Advanced PhysX: On Hairworks: On Ray Trace: Ultra DLSS: Off Shading Rate: 100%

Using 100% the same settings on my system.
AMD Ryzen 7 3800x PBO auto OC +200MHz (PBO limits auto) Scalar x1
ASRock X570 Taichi
NVIDIA GeForce RTX 2080
Team Group Inc. 8192 MB (DDR4-2400) - XMP 2.0 - P/N: TEAMGROUP-UD4-3600
(16-15-15-31-46-1 (tCAS-tRC-tRP-tRAS-tCS-tCR) 1899.6 MHz (DDR4-3800) Uncore: 1899.6 MHz)
Full custom water loop 360 rad and full copper block. push/pull fans.
Corsair AX1000



Benchmark Tool. Results:
  • Average Framerate (99th percentile): 64.88
  • Max. Framerate (99th percentile): 89.26
  • Min. Framerate (99th percentile): 38.37
Options: Resolution: 1920 x 1080; DirectX: DirectX 12; Quality: Ultra; Texture filtering: AF 16X; Motion Blur: Normal; Tesselation: Full; Advanced PhysX: On; Ray Trace: Ultra; DLSS: Off; Hairworks: On; Shading Rate: 100;

Would appear to be equal or GPU limited.
 
Last edited:
METRO REDUX BENCHMARK RESULTS Source https://www.youtube.com/watch?v=4Zf5i-fgTK8
Results: Average Framerate: 192.96
Max. Framerate: 349.41 (Frame: 27063)
Min. Framerate: 33.52 (Frame: 8) Settings:

Resolution: 1920x1080
Quality: Very High
SSAA: Off
Texture filtering: AF 16X
Motion Blur: Off
Tesselation: Very High
VSync: Off
Advanced PhysX: On

Test Bench:
Intel 9900k Processor Overclocked to 5.0ghz All Cores No AVX Offset (1.27V)
Corsair H150i PRO 47.3 CFM Liquid CPU Cooler (Fans @ 100%)
Asus ROG STRIX Z390-E GAMING ATX LGA1151 Motherboard
Corsair Dominator Platinum RGB 16 GB (2 x 8 GB) DDR4-3200 Memory
Samsung 970 Evo 250 GB M.2-2280 SSD (Boot)
Samsung 850 EVO 250 GB 2.5" SSD (Games)
Asus ROG Strix RTX 2080 OC Edition (GPU Boost Clock @ 1970MHz)
Corsair RMx 650W Power Supply
Corsair 750D Airflow Edition Case
Windows 10 Home Nvidia Driver 430.39

Using 100% the same settings on my system.
AMD Ryzen 7 3800x PBO auto OC +200MHz (PBO limits auto) Scalar x1
ASRock X570 Taichi
NVIDIA GeForce RTX 2080
Team Group Inc. 8192 MB (DDR4-2400) - XMP 2.0 - P/N: TEAMGROUP-UD4-3600
(16-15-15-31-46-1 (tCAS-tRC-tRP-tRAS-tCS-tCR) 1899.6 MHz (DDR4-3800) Uncore: 1899.6 MHz)
Full custom water loop 360 rad and full copper block. push/pull fans.
Corsair AX1000

METRO REDUX BENCHMARK RESULTS
  • Average Framerate: 201.00
  • Max. Framerate: 379.94
  • Min. Framerate: 21.66


Could the 3800x be as fast as a 9900k @ 5GHz all cores?
 
Last edited:
Increasing the minimum fps is better for the overall experience but in benchmarks the minimum can be a single frame. It's more the change in frame rate that gets me.
 
Last edited:
I wouldn't even consider an Intel CPU until they bring out a new CPU architecture, the current ones are just too compromised on the security front. So on that point I would only go with an AMD CPU until this situation changes. I'll happily take a few % hit on gaming fps (which at the resolution and GPU I use there probably wouldn't even be a difference in fps)

And you also get the same or more cores for less with AMD, on top of all of the above.
 
I have found that RAM timing are what matter, IF increase does help but RAM timings as tight as possible are the way to get great performance out of Ryren 3000. 1866 ~ 1900 IF for me (3733 ~ 3800 RAM). Tighter RAM timings lead to better performance at 1866 IF which puts it about equal to 1900 IF RAM with tight timings. Getting RAM with as tight as timings as possible basically removes the need to overclock Ryren 3000 in any way. Overclocking the cores can still get you more but this is just a small increase. With good overclock 4.5GHz all cores and RAM timings I can get 11700 in time spy but the fastest RAM timings get me to 11500 points. In this way every could Ryzen 3800x can perform at approx. the same performance with a good cooler.

ABBA made IF 1900 unstable for me but reducing the SoC voltage from 1.050 to 1.040 so far fixes cracking with audio after running the V-Ray Benchmark and ends the reboots I was having. Too much SoC voltage can make SoC unstable. Hard drives start disappearing after realbench benchmarks and one or two applications can make IF unstable and cause audio to start cracking. This can even lead to restarts of the PC system.

The main priority for Ryzen 3800x CPU is to increase IF until you can reach a maximum IF that allows for the best possible tight RAM timings. Going above 1900 IF is very hard and not having a 1:1 between RAM and IF leads to big performance losses.

If you want to go to an IF of 1900, just remember that too much voltage can cause instability, as much as too little. There is not a need to pump lots of voltage into these chips. A good safe overclock can lead to great performance. Also pushing the RAM overclock gives the most in terms of performance, not core overclocking. This makes the 3800x very reliable for overclock performance, as RAM overclocking is very reliable. Each 3800x can reach 9900k 5GHz all cores (OC 3200 RAM) like performance in time spy but the 9900k can still pull ahead of a 3800x with even more expensive RAM giving massive boosts in performance. Basically the Ryzen 3000 by design hits a wall at approx. IF 1900 which means RAM running faster than 3800 causes high latency and low performance.

The the 3800x is better value for money than a 9900k, you will need to spend a lot more money on a 9900k build to have any chance of outperforming a safe overclocked 3800x. Buying a kit like https://www.gskill.com/community/15...Memory-Kit-for-AMD-Ryzen-3000-&-X570-Platform https://www.overclockers.co.uk/g.sk...channel-kit-f4-3600c16d-16gtzr-my-10l-gs.html https://www.overclockers.co.uk/g.sk...nnel-kit-f4-3600c16d-16gtzkw-my-102-gs.htmlor https://www.overclockers.co.uk/team...3600mhz-dual-channel-kit-black-my-08q-tg.html should net you better performance than almost all 9900k systems (most people don't overclock and very few 9900k's have RAM faster than 3200).
 
Thank you zx128k,

I'd never been further than loading XMP profile for RAM until I bought Ryzen 2.

So how do I go about finding if I can do 1900IF? I've read before that you need to run everything at stock - is this (default) in BIOS profile with RAM at 1200Mhz?

I have native 3600Mhz Dark Pro RAM, so that should happily OC easier to 3800 than 3200Mhz ram, shouldn't it?

Thanks in advance
 
It will take a lot of testing to make sure IF is stable. You can set it in bios. The issue I had was after updating to ABBA bios. V Ray is great for IF testing. If you run lots of V ray bench marks and listen to music you can see if there is cracking in audo. If there is then IF is unstable. I had no issues with anything else. Reducing SoC from 1.050 to 1.040 which software shows as 1.045 fix it. All the other test program like aida64 memtest86 showed no issues.

It''s just important that you can get the tightest timings you can. If that's at IF 1866 RAM 3733 then that will be close to IF1900. I get the same performance for both atm but use IF1900 and 3800 RAM because I have reduced voltage at that speed. It takes 1.4volts on the ram @if1900 but 1.45volt @if1866 for the same performance.

Benchmarks with PDO disabled. IF 1900 RAM 3800 vDRAM 1.4 IF voltage 1.040

Metro exodus benchmark Nvidia drivers physx CPU


Time spy - 11408
https://www.3dmark.com/spy/8796206

Metro redux - Nvidia drivers physx CPU


Memory speed test


V ray score


DRAM Calculator for Ryzen


r15 & r20





9900k
https://www.youtube.com/watch?v=JZ0n9CkUkbE
https://www.youtube.com/watch?v=L54ppoyaTCQ
 
Last edited:
You don't need super ram on a 9900k to get good scores. That's a mid tier board so nothing exotic there either. It can help but it's not needed. Here's R15 and VRAY:

V-RAY:
unknown.png


R-15:
unknown.png
 
Last edited:
Try with the default RAM speed of 2666 or even an overclock to 3200. See what happens and who wins. V ray is owned by the 9900k, it loves faster clocks.

That is also a 5.1GHz which is a very rare overclock for a 9900k. Also that 3600 RAM has tight timings, far tighter than mine. 3600 RAM for the 3800x is like 3200 RAM for the 9900k. The clocks speeds are worth less, than the RAM speeds.

Even so at 5.1GHz you are not that much faster. My scores are for a stock 3800x with better ram timings. 3600 RAM is the recommend RAM for the 3800x by AMD. It's what the reviews use.

That is a top 5% 9900k cpu, almost all 9900k's (95%) are worse than yours for performance. So that cannot be a indication of 9900k performance.

The same overclock for the 3800x is a all core 4.5GHz with tight RAM timings i.e cl14 @ 3800. That would destroy your setup completely at any clock speed in R15 and is just as rare as a 5.1GHz 9900k cpu.

I know this because I can run r15 at 4.5GHz all cores. For a start a 4.4Ghz all cores 3800x is well above 2300 in r15. 4.5GHz is above 2400.

r15 4.4GHz all core tight ram (not as tight timing wise as yours) vcore 1.325v (voltage picked because its safe)


All Ryzen 3000 cpus are far faster in r15 and r20. I don't need much more than 4.4GHz all cores to beat almost all 9900k's. 4.5GHz all cores is just unnecessary. The same way the 9900k wins in vray benchmark, amd win in Cinebench.
 
Last edited:
If you're able to daily your 3800x at 4.5 all core, then yeah you certainly should! What's the point of buying a higher end sku and not running it to it's potential?

I posted my R20 score around 5500+ on here in the R20 thread.

I'm quite happy to recommend a 9900k to anyone who is willing to anyone with a relaxed budget but wants the absolute best 8 core/16 thread available today. The 3700x is my set and forget recommendation with a tighter budget.
 
Last edited:
No chance anyone should choose a 9900K over a 3800X or 3700X. The 14nm relic draws way too much power and generates way too much heat. It's also on a dead platform. Whereas AM4 will get Ryzen 4800X and 4700X next year. Easy choice.
 
Robert896r1 said:
If you're able to daily your 3800x at 4.5 all core, then yeah you certainly should! What's the point of buying a higher end sku and not running it to it's potential?

I posted my R20 score around 5500+ on here in the R20 thread.

I'm quite happy to recommend a 9900k to anyone who is willing to anyone with a relaxed budget but wants the absolute best 8 core/16 thread available today. The 3700x is my set and forget recommendation with a tighter budget.
Click to expand...

It's because with the 3800x I got things get very hot and you need a lot of vcore. I need to keep LLC at 3 and pick 1.45volts vcore. It droops into the safe zone vcore wise under heavy load.
 
You must log in or register to reply here.
Back
Top Bottom