It was in stock on Newegg for like 2 minutes. It's been going in and out of stock there and best buy for the past few days.

I was wondering if this I9 was worth it (atleast for TF2)

Is this a real upgrade compare to an I7 7700K ?

No it's not really worth it for just TF2. An overclocked 7700k probably won't be much slower than a 9900k in tf2. I wanted to have a decent amount of cores/threads for VMs for work/school while not sacrificing gaming performance so it was worth it for me in that context.

alright ty for the answer
I'll keep my money ;)

2639 frames 13.692 seconds 192.74 fps ( 5.19 ms/f) 17.625 fps variability
i7-7700k @4.5GHz
GTX1060 6Gb
16 Gigs DDR4 3000MHz

"i9 9900k @ 5ghz all cores"

What kind of cooling do you have on your PC, and what kind of temps do you have during full load?

Ryzen 7 2700x stock
16gb DDR4 3000mhz
GTX 970

2639 frames 13.710 seconds 192.49 fps ( 5.20 ms/f) 13.829 fps variability

Mastercomfig

Model Detail: Low
World Detail: Low
Shader Quality: Medium
Textures: Ultra
Shadows: Off
Effects: Low
Water: Very Low
Post Processing: Off
Pyrovision: Low
Motion Blur: Off
Anti-Aliasing: MSAA 4x
Character Detail: Low
Sprays: Off
Gibs: Off
Ragdolls: Off
Outlines: High
Sound Quality: High
Footsteps: On
Ambient Noise: Off

I5-8600k @ 5.2 ghz
16 gb 3200mhz 15 16 16 32 timing (2800 mhz rated) dual channel ddr4
vega 64

2639 frames 9.695 seconds 272.20 fps ( 3.67 ms/f) 17.228 fps variability

mastercomfig medium low
dx95

Now for something different

5.2 ghz
1333 mhz: 174.46 avg, low 90
2133 mhz: 218.83 avg, low 130
3000 mhz: 236.56 avg, low 150

3 ghz
1333 mhz: 131.68 avg, low 70
2133 mhz: 147.19 avg, low 90

The lows are not exact. they were all above the number but never below.

these are some more tests I did to test the effects of memory speed. I did my best to adjust timing to be in line for the averages of those speeds. I used my own demo on upward for this.I would love to see how even faster memory can effect fps but i can only overclock mine to 3000. the 3 ghz test is to try and mimic an ivy bridge or haswell cpu running ddr3. each test was ran 3 times. I did the log off and cancel trick to close all unnecessary tasks

I have a Be Quiet Dark Rock Pro 4. Under full load in synthetic benchmarks I'm at around 80C at 5.1GHz. I think my chip is fairly decent though because I was able to hit 5.1 stable at only 1.3V. I could do 5.2 at over 1.4V, but I was thermal throttling.

TF2 only really scales up to 3 threads and the 9900k and 7700k are basically the same architecture. If you can clock your 7700k up to about the same speeds then it should be just about the same speed in TF2.

Low:
4812 frames 17.886 seconds 269.04 fps ( 3.72 ms/f) 37.508 fps variability
4812 frames 17.800 seconds 270.34 fps ( 3.70 ms/f) 33.420 fps variability

Medium Low:
4812 frames 17.559 seconds 274.05 fps ( 3.65 ms/f) 36.882 fps variability
4812 frames 17.292 seconds 278.28 fps ( 3.59 ms/f) 34.368 fps variability

Medium:
4812 frames 18.160 seconds 264.98 fps ( 3.77 ms/f) 31.954 fps variability
4812 frames 18.222 seconds 264.08 fps ( 3.79 ms/f) 32.423 fps variability

Low w/ 3 Chrome tabs open (YT and Twitch simultaneously playing in background) + discord + spotify
4812 frames 17.924 seconds 268.47 fps ( 3.72 ms/f) 34.512 fps variability
4812 frames 17.829 seconds 269.90 fps ( 3.71 ms/f) 34.167 fps variability

Low @1920x1080:
4812 frames 18.322 seconds 262.63 fps ( 3.81 ms/f) 35.843 fps variability

CPU: Ryzen R5 2600 @3.4GHz (Stock cooler, no OC)
GPU: Radeon RX580 4g (No OC)
RAM: 16GB DDR4 @3000MHz
GPU Driver Version 18.11.1
Resolution: 1024x768 Stretched
Dxlevel: 81
OS: Windows 10 Education (tweaked and bloat-free)
Config: Mastercomfig 6.10.0 with minor changes

SMT On vs Off made no difference in TF2. I'll update with OC benchmarks once I buy a decent Noctua cooler. Also this seems like a major deviation from other Ryzen benchmarks here. Am I just lucky or are others doing something seriously wrong?

what demo did you use?

*

#522
How would the 3 GHz test simulate IVB/HSW and DDR3?
What timings did you use?

This doesn't add up.

The rest seems more or less in line with the behaviour I saw years ago, so I guess my advice is still valid, which is good to know.

I used 9-9-9-28 for 1333 mhz. I used 10-11-11-30 for 2133. I choose 3 ghz as a middle ground between SB and haswell. A haswell i7 would be quite a bit faster as they run close to or over 4 ghz. Also my cpu has about a 10-15% IPC advantage. It would be cool to see someone with a 4790k and 2133 or 2400 mhz ddr3 test my results. It would also be cool for someone with high speed memory like 3600 can check to see if it scales that high. I'm getting about 8 fps per 200 mhz all the way up to 3200.

I used my own demo because the download for the demo in the beginning of the thread is sketch AF. I ended up getting the demo from a friend after i did all the testing and didn't want to redo it.

I'm still confused on the 3000 vs 3200 MHz part. Are those typos?

So you're not really simulating IVB (or SB as you've now mentioned) or HSW, just lower single threaded performance in general.
Keep in mind that SB/IVB/HSW all don't have 6 cores on LGA115x. There's 6/8 cores (and even 4) on LGA2011, but that's with Quad Channel and 2.5 MB L3 per core so it's still a bit different.
Sadly DDR4 is quite a bit different from DDR3 so you can't truly simulate it and no one got a DDR3 mobo for Skylake so those timings are as close as you can get.
Ignoring all the internal changes on the caches and the different L2 from SKL onward it would be nice to test L3 impact someday. I never got around to doing it and an i7-4790K at the same clockrate as the Pentium G3258 with the same memory speed/timings got similar enough results that I didn't bother at the time but it could be done via CAT.
Just to see if there'd be a difference between i3/i5/i9 now. Before it the Pentium G3258 wasn't going to be able to compete with an i5/i7 purely based on threads, but overclockable 4C/4T i3 vs 8C/16T i9 would be interesting to see if 8 vs 16 MB L3 makes any difference at all.

The demo isn't sketchy, it's just broken like all old demos.

that one usually gives more frames than the one provided in first post (dustbowl pub)

for me it was 40-50 frame difference lol

The new demo has the new particles for the dragon's fury and shit, I think that's why it's preferred to use.

My memory speed was at 3000 for a long time and I only recently got it running at 3200. Also the site that demo is on is sketchy AF. it has nothing to do with being old.

I know the cpus are quite a bit different, but in the end it really doesnt matter. I damn near matched my friends 4770 with in a few % in terms of fps. This started with me watching him on twitch and noticing he was getting pretty bad frames, and I was like, huh, the 4770 doesn't have the bad of IPC, why are his frames so bad?

I started by using passmark to match my single threaded performance to his. And after doing that, I still got pretty good frames, way better than what he had. So then I was like, huh, the cpu speed should be the same, but I never touched my ram, and thats a lot faster still. So I dropped my ram speed to 1600, adjusted the timings, and suddenly I lose a bunch of fps and I'm getting more or less the same fps he his. I probably lost 30+ fps on average just from adjusting the memory speed.

a follow up on my original benchmark post after having sometime and getting some courage to overclock

4 ghz - 65ºc peak

i didn't want to reboot just to bm back to stock but it was like 220 fps lol

4.5 ghz - 70ºc peak

4812 frames 19.545 seconds 246.20 fps ( 4.06 ms/f) 30.738 fps variability

both of these used the auto voltage

tried getting to 4.8ghz with both 1.250v and auto voltage and everytime i ran the stress test in the intel tuning utility it crashed so it was def not stable
might try to get to 4.8 in the future but i'm happy with my 4.5 oc atm and don't want to risk it rn with 4.8, let alone 5.0 or above
also there is a chance of my case and everything inside it being dusty and i only have an outtake fan so that might've affected my thermals

specs below

Ok so the 3000 benchmarks were before you got it to 3200. What were the timings? Just to complete the list.

Which site is the demo on now? The zippyshare and mega links are broken again.

Passmark isn't all that accurate, but yes, memory speed and timings matter a lot. The impact increases with higher core clockrates. From the limited testing I've done it seems to be that the timings are more important than speed. E.g. 1333 CL7 (10.5ns) gives you almost have the increase in fps over 1333 CL9 (13.5ns) that 2400 CL12 (10ns) gets you. In other words 25% lower timings should be roughly equivalent to doubling the frequency at the same latency.

.

https://www102.zippyshare.com/v/fvfg8CAP/file.html

=

Finally managed to get over 300fps in benchmark1.dem.

2639 frames 8.499 seconds 310.52 fps ( 3.22 ms/f) 24.430 fps variability

-dxlevel 81, mastercomfig v7 all low/off except shaders medium.

CPU: i3-8350K@5,000MHz core, 4,400MHz cache, 1.41V
RAM: 4,133MHz, 16-17-17-33, 1.5V

What got me over the line was messing with one of the secondary timings on the RAM, tREFI. I maxed it out to 65535 and got a massive performance boost in TF2, about +20 minimum FPS.

running i5-6600k, on z170x Gaming 3, 16GB 2400 MHz RAM:

mastcomf-med-high, 3.9 GHz (stock): 126 fps
mastcomf-med-high, 4.5 GHz: 151 fps

Oh hello again, i've upgraded my gpu/ram combo for 8GB ddr2 f2-6400cl5-*sumthingidr* and a 560ti overclocked to 980 core/ 2300 memory @1V vcore.

Same settings (highest, no config, r_lod -1, no vsync or blur) on Windows 7 Ice Extreme.
CPU: Intel core 2 quad Q9550 @4,2GHz (494 FSB * 8,5 multi) @1,45vcore bios (1,440 fixed)
MOBO: ASUS P5K Vanilla (ouchie my mobo is really troubling me out with past 494 FSB, I wish I could do 500 FSB)
RAM: 4*2GB G.Skill FC-6400CL5S-2GBPQ 5-5-5-18 800mhz @1,9V oc'd to 989mhz 5-6-5-10-50 @ 2,05V
GPU: MSI 560ti Twin Frozr II/OC @ oc'd to even higher stock oc to 980/2300 @1V

I don't think I can achieve even higher frames w/o overclocking or not changing my current config.

After a bit of research, ive been able to locate some of the RAM settings in my BIOS. Honestly you should try tweaking RAM as well as CPU clock speed if you want improved fps.

Before I had no access, I had 1600Mhz DDR3 at 11-11-11-28. All i did was bump up voltage from 1.35 to 1.5, increased frequency to 1866MHz and increased tREFI to the max. Wasn't able to get 2133 working with my limited knowledge though unfortunately no matter how much i loosened the primary and secondary timings.

CPU: Intel Core i7 5700HQ @ 3.6GHz (HT on) core affinities 2-7
RAM: 16GB DDR3 11-11-11-28 @1.5V and 1866MHz
GPU: NVIDIA GeForce GTX 960M 2GB stock with 417.35 drivers
CFG: Mastercomfig low v6.10 @ DXLEVEL 81
Res: 1024x768, Fullscreen

3.6GHz

As a comparison before at 3.7GHz

Launch options:
-novid -noipx -threads 4 -nod3dex -w 1024 -h 768 -softparticlesdefaultoff

Windows 10 1809 x64 Home
No fullscreen optimizations
One single display

I wasn't seeing 100% CPU utilization on any active cores relating to TF2 before. Now i see much closer to 100%, which i speculate is due to NVIDIA and their copy system to the system RAM (optimus bleh). I say this because before I was getting lower FPS the higher resolution I went. Before I was getting about 21GB/s in memory tests at 80ns. Now I can get 26GB/s at 69ns. May be i'm wrong and these are expected gains from RAM tuning idk.

More or less expected in TF2.
You should do it properly though and use a bit lower timings as well. TF2 cares about those too.
Also tREFI is not a performance setting. Do not mess with it unless you're actually doing proper long tests afterwards to verify that you're not losing data.

You sick man, 1.5V for a CPU that supports (on its IMC) max 1.35V (DDR3L/LPDDR3) is astonishingly dangerous and can cause serious electromigration.