• 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

3800x @ stock 12700 cpu score lol https://www.3dmark.com/spy/8538557

Operating system 64-bit Windows 10 (10.0.18970)
The RAM is 3000.
Motherboard Micro-Star International Co., Ltd. B450 GAMING PRO CARBON AC (MS-7B85)

Current OS
Operating system 64-bit Windows 10 (10.0.18363)

That could mean big performance uplife will Windows 10 Insider Preview Build 18970 (20H1)
https://blogs.windows.com/windowsex...ncing-windows-10-insider-preview-build-18970/

No luck updating to it, latest build Windows 10 20H1 Insider Preview build 19008. does nothing to my performance but remove 20fps. https://www.askvg.com/new-update-build-of-windows-10-technical-preview-now-available-for-download/

Got a copy of 18990 insider preview iso to test.
 
Last edited:
ComputerBase had the chance to test the recently released Beta version of the AMD AGESA 1.0.0.4 for the ASRock X570 Taichi, the X470 Taichi, the X370 Taichi, and the X570 Phantom Gaming X. With these tests there was an improvement of 80MHz on the all-core clock speed, from 4245MHz to 4325MHz, using an ASROCK X470 Taichi paired with Ryzen 3800X. The clock comparison was made by siding the AGESA 1.0.0.3ABBA (4245MHz) and AGESA 1.0.0.4 (4325MHz). Although the stable all-core clock improved, the peak clock of 4550MHz was the same between both AGESA versions.



Another improvement that’s coming with the new AGESA 1.0.0.4 is the faster boot times, as MSI claims and HKEPC tested. As the table above and HKEPC’s video shows, the boot times are improved by over 20% in three different test scenarios. The system specifications for this test were an MSI MEG X570 GODLIKE with Ryzen5 3600 and 2×8GB memory. This update will come later, in November, for the X470 and X370 motherboards.

https://www.kitguru.net/components/...ncreases-all-core-boost-decreases-boot-times/

 
9900k if you ever want to play old games


He states only one game.

I can remember two.
  • Destiny 2
  • Max Payne
Destiny 2 got fixed with a bios patch.

Fixing Max Payne is a big issue if you want to play it these days. Windows 10 does not support DX8 for example and thus not Max Payne. Basically the game is unplayable on every system without patches. https://steamcommunity.com/sharedfiles/filedetails/?id=1633394421

The patch is being worked on here for the jpeg issue,
https://github.com/luigoalma/maxpayne-grphmfc-jpg-patcher
 
Last edited:
With Max Payne no one knows what is causing the issue as far as I can tell. It could be an old optimisation technique or cpu instructions that are no longer supported. A bug within Zen 2 like with destiny 2. As just one game is reporting an issue so far, it looks like Max Payne is doing something the Zen 2 no longer supports. It could be a assembly code optimisation in the loader. Remember Zen 2 CPU's support Windows 10 only and not anything before Windows 10. So there is no support for Max Payne or expectation that it should run normally. I would try a VM and see if I could get it to run there on windows 95. Yeah it sucks, as I have lots of old games I love playing from time to time.
 
It would certainly be worth a try, but the hypervisor just passes the capabilities of the CPU straight through to the VM for speed. So if Max Payne is reliant on a specific Intel CPU feature, it still isn’t going to be there.

I don't know what will happen. Will allow installing windows 95 on a ryzen 3000 cpu.
 
5.3ghz no avx offset dialed in. Auto llc with +0.190v offset. I'll make this my daily during winter and use my 5.2ghz profile for summer.

unknown.png

1.4 volts vcore? Also why is it suck at 150.00 watts? Did you limit it? That will kill performance which would explain the low 57c and not thermal death. My 4930k would hit 180 watts and hit 82c in prime95 small fft with close to 1.4 volts vcore.
 
Last edited:
OCCT is reading off the superio chip which is wildly inaccurate on GB z390 boards. I’m giving feedback to the dev on getting updated libraries to get the isl69138 and ir35201 sensors into the list so you don’t need hwinfo64 running also.

1.29V is good, 1.35V is still OK, anything at 1.4V and over is a risk.

Also prime95 fft 12k is 195 watts @ 1.290 volts vcore @ 5GHz. This is a version of prime95 without avx.

5.3GHz @ 1.4 volts is basically going to be higher than 57c on epic custom water with avx instructions. Both the temps and load watts are wrong. CPU-z shows a load voltage of 1.4volts which can't be right for an avx load temp of 57c. With full custom water cooling. "5.2Ghz is stable on p95 non avx @ 1.43 with a max temp of 89c and a "constant" temp of around 82c." https://www.reddit.com/r/intel/comments/9v569b/safe_247_voltage_i99900k/

nyjets239
1 point · 1 year ago
I settled on 1.345 @ 5.1 with no offset. Not worth the extra voltage for 100mhz.

level 2
mrbubbles02
1 point · 1 year ago
What are your clocks & temps under AVX workloads? From what I've read you're going to have a hard time at 5GHz AVX without an offset.

My i7-9700k has a hard time as-is on stock speeds, without the hyperthreading, on air cooling at 4600MHz AVX workloads.

level 3
nyjets239
1 point · 1 year ago · edited 1 year ago
Around 75-80c in realbench, cinebench, and aida64. Prime95 gets 95-100c but it doesnt crash so i guess im good. That's the most extreme case though. Im using custom watercooling loop so my temps are pretty good I assume.

MSI abvice for stress testing...
- AIDA64 or Prime95 v26.6 (non-AVX) / Prime95 v27.9 (AVX) for a stress test
If the temperature is above 90° C, you should lower the Core Voltage.

Please run prime95 avx and post the temps.


3D Mark time spy cpu G.SKILL Trident Z RGB Series 32 GB (4 x 8GB) CL16 3600 MHz
https://www.3dmark.com/3dm/40710090? and https://www.3dmark.com/spy/9194514 and https://www.3dmark.com/spy/9195676

https://cdn.wccftech.com/wp-content/uploads/2019/10/3DMark-Timespy-1480x1138.png
Power
https://cdn.wccftech.com/wp-content/uploads/2019/10/Power-1-1480x1143.png
ASUS Ryujin 240 AIO liquid cooler
https://cdn.wccftech.com/wp-content/uploads/2019/10/Temps-1-1480x1140.png

9900ks = 3800x OC with same RAM.
 
Last edited:
5.3ghz benchmarks. A lot of the KS units should be able to do this or close.

CPU-Z and V RAY:

R20:

5.3GHz all cores, most won't reach that or won't be able to cool it. 9900k/kf/ks its all about cooling, people can bench for awhile but the temps are too high. If you have 5.3GHz on a 9900k rejoice, as very very few people will also have that type of overclock. It's very rare. Most people will get 5.1GHz all cores on the 9900ks and a few will hit 5.2GHz but 5.3GHz will be very rare. People run AVX offsets to keep things cool.

Take the 9900k time spy cpu score, the average is 11k. https://www.3dmark.com/newsearch#advanced?test=spy P&cpuId=2402&gpuId=&gpuCount=0&deviceType=ALL&memoryChannels=0&country=&scoreType=physicsScore&hofMode=false&showInvalidResults=false

The 9900ks average is 11713 https://www.3dmark.com/newsearch#advanced?test=spy P&cpuId=2544&gpuId=&gpuCount=0&deviceType=ALL&memoryChannels=0&country=&scoreType=physicsScore&hofMode=false&showInvalidResults=false
The highest speed is 5,404 MHz. https://www.3dmark.com/spy/9118396 Most are topping out a 5,204 MHz in the top scores.

A 5.3GHz all core 9900k is so very rare and this is bench runs on 3d mark time spy not 24/7 overclocks.

In Cinebench R20 I get MT 5200 approx. on a good day. This is 3800x pbo and ram OC. That is enough to bet the 9900ks stock with a decent RAM kit.

There are lots of crazy 3800x scores, https://www.3dmark.com/spy/8538557
A true 12730 time spy score would produce a 3800x that would destroy all before it. Given the RAM is 16,384 MB Corsair DDR4 @ 2,998 MHz and the core is 4,349 MHz, there is lots of room for improvement. It's pointless stating that these outliers are indictive of the whole distribution. They are not.
 
Last edited:
Timespy scales well with memory also not just cores. Example; I'm timespy bottlenecked because my mobo is really bad at frequency scaling on 2 dimms. It hits at wall at 3600mhz. So going from 5.2 to 5.3ghz is only worth a ~ hundred points.

Most of the people running 9900k's stock and with xmp ram will perform to the average. If I could get CL16/4133mhz which is standard for a high end z390 board, I'd be in the mid 13k range instead of low 13k. https://www.3dmark.com/3dm/40886806

Thats what I have found as well. The RAM speed and timings are worth more than the core clock speed. So I have focused on the RAM. You could find that if you lower the clock speed, then you "may" get better RAM overclocking. Could be that the RAM is Dual rank or in the wrong slots. Could be you are T-topology (Gigabyte?) and not daisy chained. Just sheer bad luck etc.

The clock speed increase is just a few points, even for the 3800x. Most people won't even tune the RAM, just install it and leave it at the kits settings. A stock 9900ks is 10700 cpu points in time spy. Add a 3600 RAM kit and you hit approx 11400. 5.3GHz and you hit 12k time spy cpu. You can tighten the timing after that.

Note the stock RAM for the 9900k is DDR4-2666 anything above that is overclocking. You are not bottlenecked by not getting above 3600 (1800MHz). You have simply reached the limit of the IMC/RAM overclock.

There is something wrong with many of the time spy scores. There is a 3800x that gets 12700 cpu points with crap RAM and underclocked cores, 3000 RAM not the stock 3200. AgentNardz then calls it "Stock". https://www.3dmark.com/spy/8538557

This is the top 9900ks https://www.3dmark.com/spy/9118396
CPU Score 13950 5,404 MHz Patriot Memory DDR4 @ 4,400 MHz

9900ks 13213 https://www.3dmark.com/spy/9132256
The 13200 you gave above. What would it take?
CPU Score 13213 5,303 MHz Kingston DDR4 @ 3,998 MHz

These are both in the top 100 9900ks cpu's. Almost everyone is not getting this level of performance in time spy cpu. Average is 11700.
 
Last edited:
That's my score from this morning. 52core/47uncore and 15-15-15/3600mhz ram. My aorus pro is a 4 layer pcb with t-top so with 2 dimm's it hit a wall at 3600mhz. I have it tuned as tight as it'll get (b-die) with 255 trfc and manual tuning of all tertiary timings.

Sounds like that is as far as you will go. The RAM timings are already tightened and you are likely already 1.45Volts+ on the RAM. With a motherboard that uses T-topology will normally overclock the best with 4 sticks. With two sticks I would have got a daisy chain motherboard. b-die as far as I know is single ranked.

Even if you cant overclock more, its still a sweet ride. 99th percentile and all.
 
Last edited:
nm wrong info

https://www.reddit.com/r/intel/comments/cfb12g/gigabyte_z390_aorus_pro_wifi_daisy_chain_or/eu8qnco/

yep its completely stable using patriot 4400 19 19 19 39. is there max frequency difference between 2,4 and 1,3 slots for ram?

https://www.amazon.co.uk/Viper-Steel-4400MHz-Performance-Memory/dp/B07KXLFDL6

On the motherboards memory QVL you should reach 3800-3866 at least. Just using the benchmark info you provided.

https://download.gigabyte.com/FileList/Memory/mb_memory_z390-aorus-pro-wifi_190415.pdf

Try changing the RAM slots, should not work but worth a try.
Seems to work well with patriot 4400 19 19 19 39. The QVL loves Samsung b-die.

"Besides the first two models, the range looks very promising and overclocking-friendly because, according to the manufacturer’s info, it’s supposed to be based on Samsung’s B-Die memory chips."
https://www.guru3d.com/articles-pages/patriot-viper-steel-4000-mhz-cl19-ddr4-review,1.html

Looks like Samsung B-die. What settings are you using when you increase the frequency? Auto and increased DRAM voltage should be the answer.
 
Last edited:
Core i9 9900KS @ 5.1GHz HEVC Power Consumption 365W.
Core i9 9900KS @ 5.0GHz HEVC Power Consumption 355W
Core i9 9900KS @ 5.0GHz (MCE off) HEVC Power Consumption 254W
Core i9 9900K @ 4.7GHz HEVC Power Consumption 289W
Ryzen 9 3900X HEVC Power Consumption 228W
 
Back
Top Bottom