13900K RTX4090 ultimate work/gaming build for 1080p (reply to Wendell)

Hello everyone
in a level1techs video Wendell asked who and why would someone buy a RX7900 or RTX4090 to play in 1080p and to explain it in forums so here goes.

WHY 1080P :
I have a very special 10y old 27" 1080p TN monitor
an old 3d model where the lightboost hack works LightBoost HOWTO | Blur Busters
this gives me zero ghosting nor motion blur, not less…NONE
I can read the street names on this and they look as sharp as a still picture TestUFO: Photo
“working” black frame insertion has no competition, sadly it’s apparently very hard to make, as usual a hack worked better than what manufacturers do :confused:
playing on anything else feels like going from 120fps to 30fps you can’t go back from this
oled is close but not quite the same, I’m placing my hopes into recently announced LG 27" 1440p oled monitor
the monitor flickering nonstop isn’t going to last forever one day it will break and then I’ll move to oled, no VA or IPS for me

THE BUILD :
this is the 25th iteration of my personal PC and has seen 6 different computers inside it
9900k z390
3960x trx40
5950x x570
12900k z690
13900k z790
it started as an all air cooled and ended in a triple rad custom loop with everything in-between 1,2,3 rads cpu,cpu+gpu, cpu+gpu+vrm waterblocks, trying various models and brands of fans the lot…
a lot of work and testing to reach this design so I can say with little doubt you cannot make better at this size



SYSTEM PARTS :

  • Case : Lian Li o11 Dynamic XL
  • Motherboard : MSI MAG Z790 TOMAHAWK WIFI DDR4
  • Memory : 4x8Gbs Corsair Dominator DDR4-3600 18-19-19-39 2T
  • Videocard : Inno3d iChill Frostbite RTX 4090
  • Soundcard : Creative Sound Blaster ZxR
  • Network card : Intel X550-T1 10Gbps
  • Storage : 4x nvme + 6x ssd

COOLING :

  • Pump : Hydro X Series XD5 RGB Pump/Reservoir Combo
  • Radiators : 2x Corsair Hydro X Series XR7 360mm (55mm) + 1x Corsair Hydro X Series XR7 360mm (30mm)
  • Cpu waterblock : EKWB EK-Quantum Velocity² D-RGB – 1700
  • Gpu waterblock : alphacool Eisblock Aurora Acryl GPX-N RTX 4090 Reference mit Backplate
  • Fans : Corsair QL rgb 120mm fans
  • Fan Controller : 2x corsair commander core xt

COMPROMISES :

  • rear fan should be an intake for optimal cooling, but it gets accelerated by the top ones and starts whining, also it’s already an imbalanced 6vs4 intake exhaust 1 more intake is too many
  • QL fans ? rated worst static pressure ! good enough and they have the softest air sound of all brands, yes better than noctuas who whine hard when they start actually pushing air, QL make a woosh even at max speed and while not silent you can absolutely work all day in the same room with no earplugs
  • RGB ? completely uneccessary but can be turned off with 1 key press
  • Fan placement ? the big compromise, fans work better in push but for the rgb I have the bottom and vertical fans set in pull, sucking air through a rad isn’t good but worse it changes the way the air exists the fan, in pull mode air exits in a 45° crown shape with almost no flow in front of the fan unlike push mode creating a straight hairdryer like jet, if you don’t want rgb and want maximum cooling place fans in push

DDR4 ? :
ddr5 is beta testing, I hate it only 2 sticks work at acceptable speeds and even like that it’s very finicky only g.skill worked on z690 corsair didn’t boot and on x670e it’s the opposite corsair boots but g.skill doesn’t
I couldn’t make my very expensive msi meg ace x670e ddr5 work and so it’s on a shelf for now :confused:
I wasted days and money with ddr5 but gave up went ddr4 with a 3600 kit I had and don’t regret it one bit “it just works”
I feel sad for the many topics I see in level1techs and elsewhere I tried to warn people since beginning 2022 that ddr5 didn’t work and all I got for it was shadow-banning, don’t ever ask a tech youtuber why he talks about 4 dimms support but reviews with only 2 dimms while having 20 sticks behind him, they don’t like when their dishonesty is exposed

COOLING DESIGN :


this is not based on theory but reality
I tried all the combinations possible and I mean all…even inverting fans 1 by 1 on rads to see if chaotic airflow would somehow work
this gave me the best result, minus using the rear fan as intake which gives you better temps but make an annoying whining noise because the top-rear fans accelerate each other going above spec
the rear fans is mostly there to exhaust gpu backplate heat but on a watercooled 4090 even with no active backplate it’s quite tame, mildly warm air is all I feel
this can render process videos or whatever for 8hrs straight with barely an impact on your room temperature
that’s what the 3rd rad is for, when it comes to liquid cooling if your liquid temp goes above 45°C it’s a fail > 45°C liquid temp has a noticeable impact on room temperature you want to stay below
this config with the 13900k and 4090 both very good at idle as seen in below picture 8hrs of idling (network disconnected) is I dare say, cold !

air cooled and with older cpu and gpu I had no heating and was sweating in briefs at night in winter, now I’m cold, the room temp dropped at least 10°C which is good because my air cooled 3960x+1080ti running 100% cpu for 9hrs gave me 48°C/118°F room temp in summer, that’s what triggered the move to liquid cooling

I think that covers the build itself
the 13900k and 4090 both need a different message because I have a lot to talk about them, everything reviewers don’t tell you (sadly)

EDIT : when I said ultimate I meant ultimate compromise which might not be everyone’s definition of ultimate ><

  • it’s not the ultimate cooling (use non rgb fans and set them all in push mode)
  • it’s not the ultimate performance DDR4 is a limit factor for sure
    but if you want performance & temperatures & noise levels & rgb it has all
    I just demoed to a friend and his comment at max fan speed (1700ish) “wouldn’t bother me while working”
1 Like

OPINION ON 13900K

  • I do not understand this cpu at all

  • in most games it behaves like an 8 cores cpu

  • in multi-tasking I barely see efficiency cores do anything but the results are there “it works” for a price…watts lots of watts, this is how I monitor what it’s doing (when applicable) custom made rivatuner overlay (aida64 only reports 16 frequencies so it’s missing the last 8 e-cores)

  • I do a lot of things at the same time, watching youtube while encoding videos and testing games it doesn’t seem to care the ryzen 16 cores I understand but not this as it feels as fast or faster than amd with only 8 “real” cores I even forgot I had kombustor stress testing the gpu for 3hrs in the background only realized it while trying to run borderlands 3 benchmarks (still ran 144fps lol)

  • it’s the hottest cpu I have ever tested to this day (and I had an oc 3960x that reached 430w)

  • it requires at least an AIO but really you want a custom loop at this stage

  • a custom loop won’t spare you from going to 95-100°C but the temps go down really quickly, this is single-thread cinebenchR23 run followed by 3 multi-thread runs it takes 2s to go from 99°C to 38°C
    (true for gpus too especially the super efficient rtx4000)
    image
    below are multiple cpus with the explained triple rad ql fans setup


    the 13900k you cannot properly cool 384w on 2cm of die >< (24x11mm) it’s hopeless
    the 5950x cooling slope is slow but I ran a multi-stage load dependent voltages and clocks using 1usmus ClockTuner v2.1 for Ryzen (I think i described it 1-2 years ago here)

OPINION ON RTX 4090

  • this gpu it reminds me of when I got 10Gbit internet and slammed onto a wall of bottlenecks everywhere
  • this gpu 180° the cpu/gpu balance, cpus have become too slow even a 13900k or 7950x unless you have ultra fast DDR5
    it seems to help as proven by Wendell in this video 2 G.Skill DIMMs on a MEG z790 Godlike Build! What Difference Does it Make? - YouTube
  • mine is a 450w inno3d ichill frostbite which came with bad contact between gpu and block (112°C hotspot :hot_face:) I repasted it with thermal grizzly kryonaut which gave me the best temperatures I have ever seen on a gpu, I have no aircon my room is at 23-25°C I kept the overheating results for a reason which i will explain below

3DMARK RESULTS overheated/repasted/repasted [email protected] 3090MHz/24000MHz :

API Overhead feature test ovhe DX11MT/DX11ST/DX12/Vulkan : 7327728/5013254/60264420/47706078
API Overhead feature test repa DX11MT/DX11ST/DX12/Vulkan : 7422714/5170316/61871838/47946294
API Overhead feature test rpoc DX11MT/DX11ST/DX12/Vulkan : 7443677/4906467/61707078/47921568

CloudGate ovhe score/graph/physi : 87210/193408/29848
CloudGate repa score/graph/physi : 83843/180398/29180
CloudGate rpoc score/graph/physi : 82848/180732/28612

CPU Profile MaT/16T/8T/4T/2T/1T ovhe : 17157/12294/9039/4786/2427/1225
CPU Profile MaT/16T/8T/4T/2T/1T repa : 17172/12238/9059/4793/2420/1221
CPU Profile MaT/16T/8T/4T/2T/1T rpoc : 17140/12295/9049/4785/2428/1225

DirectX Raytracing feature ovhe : 143.78
DirectX Raytracing feature repa : 145.73
DirectX Raytracing feature rpoc : 154.02

Firestrike ovhe score/graph/physi/combi : 47405/64620/50901/15287
Firestrike repa score/graph/physi/combi : 46510/64090/50854/14602
Firestrike rpoc score/graph/physi/combi : 47808/67286/50711/14679

Firestrike Extreme ovhe score/graph/physi/combi : 38630/46167/50615/14976
Firestrike Extreme repa score/graph/physi/combi : 39067/46406/50664/15446
Firestrike Extreme rpoc score/graph/physi/combi : 40441/48698/50897/15677

Firestrike Ultra ovhe score/graph/physi/combi : 25123/25180/50740/14161
Firestrike Ultra repa score/graph/physi/combi : 25252/25378/50890/14086
Firestrike Ultra rpoc score/graph/physi/combi : 26579/27119/50699/14269

Ice Storm ovhe score/graph/physi : 256816/684944/80565
Ice Storm repa score/graph/physi : 259693/708313/80731
Ice Storm rpoc score/graph/physi : 260169/708712/80920

Ice Storm Extreme ovhe score/graph/physi : 254709/661989/80775
Ice Storm Extreme repa score/graph/physi : 255526/680916/80189
Ice Storm Extreme rpoc score/graph/physi : 257865/687121/80924

Intel XeSS feature test 4K Balanced ovhe Off/On/Diff : 56.10/101.48/80.9
Intel XeSS feature test 4K Balanced repa Off/On/Diff : 56.93/102.48/80.0
Intel XeSS feature test 4K Balanced rpoc Off/On/Diff : 60.07/107.41/78.8

Intel XeSS feature test 4K Performance ovhe Off/On/Diff : 56.19/122.78/118.5
Intel XeSS feature test 4K Performance repa Off/On/Diff : 57.02/122.13/114.2
Intel XeSS feature test 4K Performance rpoc Off/On/Diff : failed

Intel XeSS feature test 4K Quality ovhe Off/On/Diff : 56.06/93.12/66.1
Intel XeSS feature test 4K Quality repa Off/On/Diff : 56.97/94.57/66.0
Intel XeSS feature test 4K Quality rpoc Off/On/Diff : failed

Intel XeSS feature test 4K Ultra Quality ovhe Off/On/Diff : 56.09/71.61/27.7
Intel XeSS feature test 4K Ultra Quality repa Off/On/Diff : 56.93/72.61/27.6
Intel XeSS feature test 4K Ultra Quality rpoc Off/On/Diff : failed

Mesh Shader feature test ovhe Off/On/Diff : 351.86/1109.96/215.5
Mesh Shader feature test repa Off/On/Diff : 352.30/1122.26/218.6
Mesh Shader feature test rpoc Off/On/Diff : 361.50/1149.00/217.8

Night Raid ovhe score/graph/physi : 94986/207335/23335
Night Raid repa score/graph/physi : 94759/204541/23447
Night Raid rpoc score/graph/physi : 96845/217666/23362

Nvidia DLSS2 feature test 4K Performance ovhe Off/On : 59.89/143.11
Nvidia DLSS2 feature test 4K Performance repa Off/On : 60.61/144.14
Nvidia DLSS2 feature test 4K Performance rpoc Off/On : 63.91/151.39

Nvidia DLSS2 feature test 4K Quality ovhe Off/On : 59.78/103.52
Nvidia DLSS2 feature test 4K Quality repa Off/On : 60.59/104.92
Nvidia DLSS2 feature test 4K Quality rpoc Off/On : 63.89/110.57

Nvidia DLSS3 feature test 4K Performance ovhe Off/On : 58.52/219.89
Nvidia DLSS3 feature test 4K Performance repa Off/On : 59.33/220.74
Nvidia DLSS3 feature test 4K Performance rpoc Off/On : 62.73/231.91

Nvidia DLSS3 feature test 4K Quality ovhe Off/On : 58.53/157.63
Nvidia DLSS3 feature test 4K Quality repa Off/On : 59.37/159.16
Nvidia DLSS3 feature test 4K Quality rpoc Off/On : 62.70/169.13

PCI Express feature test ovhe : 24.28
PCI Express feature test repa : 24.26
PCI Express feature test rpoc : 24.27

Port Royal ovhe : 26410
Port Royal repa : 26766
Port Royal rpoc : 28105

Sampler Feedback feature test ovhe Off/on/Diff : 1201.83/1316.73/9.6
Sampler Feedback feature test repa Off/on/Diff : 1203.05/1319.12/9.6
Sampler Feedback feature test rpoc Off/on/Diff : 1311.24/1435.23/9.5

Sky Diver ovhe score/graph/cpu : 135562/248187/54896/61470
Sky Diver repa score/graph/cpu : 135226/247678/55037/60923
Sky Diver rpoc score/graph/cpu : 135583/250388/54344/61730

Speed Way ovhe : norun
Speed Way repa : 10269
Speed Way rpoc : 10979

Time Spy ovhe score/graph/cpu : 32714/37426/19093
Time Spy repa score/graph/cpu : 32435/37396/18516
Time Spy rpoc score/graph/cpu : 33420/38511/19108

Time Extreme Spy ovhe score/graph/cpu : 18440/19881/13072
Time Extreme Spy repa score/graph/cpu : 18699/20241/13061
Time Extreme Spy rpoc score/graph/cpu : 19392/21171/13138

VRS feature test Tier1 ovhe Off/on/Diff : 3113.08/4202.12/35.0
VRS feature test Tier1 repa Off/on/Diff : 3330.49/4145.45/24.5
VRS feature test Tier1 rpoc Off/on/Diff : 3432.33/4095.15/19.3

VRS feature test Tier2 ovhe Off/on/Diff : 360.60/550.88/52.8
VRS feature test Tier2 repa Off/on/Diff : 380.66/578.71/52.0
VRS feature test Tier2 rpoc Off/on/Diff : 391.36/593.12/51.6

Wildlife ovhe : 162084
Wildlife repa : 162168
Wildlife rpoc : 164149

Wildlife Extreme ovhe : 90341
Wildlife Extreme repa : 90717
Wildlife Extreme rpoc : 94531

Borderlands 3 1080p Badass DX11 ovhe : 189.53
Borderlands 3 1080p Badass DX11 repa : 188.44
Borderlands 3 1080p Badass DX11 rpoc : 188.28
Borderlands 3 1080p Badass DX12 ovhe : 274.68
Borderlands 3 1080p Badass DX12 repa : 283.69
Borderlands 3 1080p Badass DX12 rpoc : 284.69
Borderlands 3 1080p Ultra DX11 ovhe : 190.43
Borderlands 3 1080p Ultra DX11 repa : 190.39
Borderlands 3 1080p Ultra DX11 rpoc : 190.25
Borderlands 3 1080p Ultra DX12 ovhe : 281.78
Borderlands 3 1080p Ultra DX12 repa : 282.21
Borderlands 3 1080p Ultra DX12 rpoc : 288.83

Shadow of the Tomb Raider 1080p Highest DLSS Quality ovhe AvgFPS/GpuBound% : 274/0
Shadow of the Tomb Raider 1080p Highest DLSS Quality repa AvgFPS/GpuBound% : 270/0
Shadow of the Tomb Raider 1080p Highest DLSS Quality rpoc AvgFPS/GpuBound% : 269/0
Shadow of the Tomb Raider 1080p Highest DLSS none ovhe AvgFPS/GpuBound% : 274/0
Shadow of the Tomb Raider 1080p Highest DLSS none repa AvgFPS/GpuBound% : 273/0
Shadow of the Tomb Raider 1080p Highest DLSS none rpoc AvgFPS/GpuBound% : 265/0

Unigine Heaven 4.0 Extreme ovhe fps/score : 485.5/12230
Unigine Heaven 4.0 Extreme repa fps/score : 487.7/12286
Unigine Heaven 4.0 Extreme rpoc fps/score : 500.7/12612

Unigine Superposition 720pLow ovhe : 53275
Unigine Superposition 720pLow repa : 53085
Unigine Superposition 720pLow rpoc : 52563

Unigine Superposition 1080pExtreme ovhe : 22387
Unigine Superposition 1080pExtreme repa : 22552
Unigine Superposition 1080pExtreme rpoc : 23546

Unigine Superposition 4KOptimized ovhe : 34741
Unigine Superposition 4KOptimized repa : 35229
Unigine Superposition 4KOptimized rpoc : 37301

SUMMARY :
I wish I could have colored the text would have made it easier on the eye so I’ll point what you should look for :

  • I kept the overheating gpu results as they reinforce what other tests show
  • in lower resolutions or when the gpu is underused in older benchs for example you lose performance as the cpu demand from the gpu is straining it even more for no real gain
  • shadow of the tomb raider should be retired it might be more acceptable in 4K rather than 1080p but it’s clear we reached the limits of the game, 4K has a few % left but still
    imageimage!
  • I’m sad to say after watching Wendell video above that Borderlands 3 in 1080p which was my go to for gpu comparison has also reached it’s limits (I get 280ish in dx12 with ddr 3600 when wendell gets 300 with ddr 7200 proving)
  • I almost couldn’t reach 450w even repasted I think two 3dmark bench were close but watts only came after ocing roughly +120 on gpu (variable because of cpu limitation) and +1500 on memory
  • ocing for my use has a negative impact in 1080p 120fps or unlimited or 4k/60fps the gpu is underused but the increased strain on cpu from the oc seems to stay I played several hours or forza horizon 5 1080p max settings+RT and it’s clearly smoother without the oc also that overclock makes my gpu watts go from 330 to 440 for no positive result except in benchmarks
  • haven’t tried undervolting or underclocking no reason to do it for 1080p the gpu load is so low it does it by itself
  • DX11 or older games are unable to use those monster gpus 8 cores is not enough just check borderlands 3 results, I had almost the same ones with a 3090 while the DX12 ones are drastically different with +100fps

Youtubers from what I’ve seen often talk about the efficiency of the card
this is especially true in 1080p where the 4090 sleeps in most games, this is a “benefit” of running lower than 4K resolutions if you want to see it like that.
Usual forum and video comments “too many watts electricity isn’t cheap” “too many watts it’ll be as hot as the sun”
Well in 1080p I can run max details+RT (which I couldn’t) 1080p @120fps and the entire pc uses barely more watts than my 1080ti alone (310w oc) was using
it shows in room and liquid temps, I’ll have to slow down my fans as I setup my icue “gaming” profile for 800w@psu and I’m mostly running <400w as long as I’m in 1080p
I have no picture but just to give you an idea the ascent 1080p max RT in the main town hub which has extreme RT effects
12900k + 3090 830w@psu
13900k + 4090 550w@psu
yeah it’s that dramatic and that’s all on the gpu, the 13900k uses more watts
My 1200w psu has become overkill I didn’t expect that, I bought it when the 3960x + 1080ti reached 1100w, it’s crazy how cpus and gpus have advanced in 2-3 years

I’m sorry but I tried reading the post over and over again, but I don’t get why you would buy top tier hardware to still use a 1080p display.
To me the two things are not correlated: that display is gonna perform no matter what PC you put behind it. If you wanted to take full advantage of your 144Hz (or 120Hz) display you could’ve just kept using your old GPU and CPU, like the i9 9900K. Maybe pair it with a 3080, but even that might be overkill.
If you’re playing at 300fps or 6000fps is not gonna make a difference.

Still a sick build, ain’t gonna question that!

3 Likes

I was about to say the same thing - awesome build, custom loop looks pretty great. If you ever empty it to add/change things, maybe move the line between the reservoir and GPU to the bottom side of the card and put a 45° fitting out the top towards the radiator to help with that bend in the tubing towards the rear of the case, but that’s just aesthetic.

But, I still don’t get the justification for the 4090… Don’t get me wrong, I have a 3090 that does nothing but play Dwarf Fortress these days too, so if you’ve got the money and you just want a 4090, that’s all the justification you need. :rofl:

It just isn’t going to help you at 1080p regardless of what monitor you’ve got hooked up, and things are going to be largely CPU-bound as the refresh rate goes up. And at 10 years old, I assume your monitor isn’t 240Hz. So I’m not trying to discourage the GPU choice, just confused by the “why 1080p” premise lol

a working BFI is like VR explained to someone who hasn’t seen it…you cannot understand
and I wish you never do because you cannot unsee this, I will not play moving games on anything else, I have a 4K oled tv next to it I don’t play it “it’s blurry as hell” as far as I’m concerned…yes those 5/5 stars oleds are unwatchable when BFI stopped working for a few weeks I started reading wh40k books instead
you’re better off not knowing, it’s a curse

this is a video I made years ago, yes I know video compression is better on a still image than in movement but it represents perfectly what BFI is compared to everything else, what other monitors or TV are to me is like this video in movement, what you see with a proper BFI are the still frames…all the time even when you move
no monitor response can ever compare to it because the problem comes from our eyes&brain, bfi “erases” the image in your brain every black frame
I totally forgot to say it but you need 100-120fps constant for it to be at it’s best otherwise you start seeing the flickering or “stuttering”

other than that
I don’t play with AA or FXAA at worse sharpness is all I care (related to bfi because I can have almost still picture quality in movement I want it, no blur !)
I play heavily modded skyrim and fallouts which give 13fps or even 3 to 99% of computers both for cpu (it’s running like 2000 scripts in the background) and gpu my “needs” threshold in 1080p far surpasses retail games it’s unrelatable to non-modding gamers
this is what decides which computer I keep not standard games

you will understand why it’s still actually a good choice for 1080p when I populate the 4090 part, it has benefits

last thing I have a 4k tv so it needs to be able to drive 4k too and it’s not my fault if I can’t find a better monitor it’s because they are all trash, told you it’s a curse I bought 3…and went back to my grey-blacks 10y old TN I even fetched my CRT from the attic in desperation but you can’t go back to this, it’s all blur ><
I hope this 27" 1440p 240Hz oled will save me lol

1 Like

Adding to the choir, here. You really do need a good monitor to go with that PC, otherwise you’re buying a Ferrari only to use it as a horse carriage… I mean you could, no doubt, but sense makes none :person_shrugging:

As for monitors, I can suggest digging around the HUB offshoot Monitors Unboxed:

updated the 4090 “review” with too many benchmarks

sorry I was afk I share hobbies with Krista (hiking) and Ryan (sportscars)
this is what happens when you play too much gran turismo/forza and skyrim you end up rallying up mountain roads to go “explore” :wink:


you know you play too much when you gaze at the scenery and think “that texture looks great” :rofl:

CASES WHERE A 112°C HOTSPOT OVERHEATED CPU DOES BETTER THAN A 3GHZ 24GHZ OC HALF THE TEMP

CloudGate ovhe score/graph/physi : 87210/193408/29848 (best)
CloudGate rpoc score/graph/physi : 82848/180732/28612 (worst)

Borderlands 3 1080p Badass DX11 ovhe : 189.53 (best)
Borderlands 3 1080p Badass DX11 rpoc : 188.28 (worst)
Borderlands 3 1080p Ultra DX11 ovhe : 190.43 (best)
Borderlands 3 1080p Ultra DX11 rpoc : 190.25 (worst)
both ranking are reversed and logical in DX12 with the faster gpu gaining fps (a little not much)

Shadow of the Tomb Raider 1080p Highest DLSS Quality ovhe AvgFPS/GpuBound% : 274/0 (best)
Shadow of the Tomb Raider 1080p Highest DLSS Quality rpoc AvgFPS/GpuBound% : 269/0 (worst)
Shadow of the Tomb Raider 1080p Highest DLSS none ovhe AvgFPS/GpuBound% : 274/0 (best)
Shadow of the Tomb Raider 1080p Highest DLSS none rpoc AvgFPS/GpuBound% : 265/0 (worst)
here we have proof from the game engine itself that it’s 0% gpu bound and again the best scores come from a slower gpu
also note that DLSS brings nothing as we are cpu limited

Unigine Superposition 720pLow ovhe : 53275 (best)
Unigine Superposition 720pLow rpoc : 52563 (worst)
same scenario very low res, mostly cpu limited and slower gpu wins

I haven’t tested older gpus with the 13900k but we have reached some kind of tipping point at which the ultra fast 4090 demand on the cpu is too much for no useable benefit
Some older benchmarks are at 4000+fps which is probably the problem lol
4000fps in 1080p = 1000fps in 4K and I guess, a guess, the physics calculation of 4000fps are heavier on the cpu than on the graphical side

the upside is if slowing it down which I will try gives me better fps in 1080p the power usage is going to lower even more and it’s already very low
I went from regularly having 600-800w@psu 2 years ago (thus the overkill custom loop I designed for a 1100w@psu threadripper pc) to 350w with a much smoother experience

At 1080p this build is a waste of money.

But since your not actually asking for advice, just validation I will simply say you do you. You will be getting more frames with HDR/RTX on than your monitor can display for any game, and your 1% lows should hit your screens refresh rate most of the time.

you are half right I expect no advice unless you’re builzoid or tested much more hardware than me but I don’t care about validation I’m giving my results which you can take or leave
a spark of motivation or helping a decision in one way or another is all I hope out of it
I’m doing all this for myself already I’m just trying to share it here
I’m also wasing a lot of time because I have more than most and hope to find something interesting, if you can take something out if it then great (case in point my wall of benchmarks that almost no one will read)

I dislike HDR but with no experience in gaming I can only talk about movies
on 600+ side-by-side HD vs 4K HDR the HD version wins 9/10 for me
most 4k is really upscaled 2k often over-compressed on top of that
a sky is supposed to be blue, most HDR skies are too bright sometimes pure white it’s not how it’s supposed to look and does not in the original SDR version, I don’t have the latest “ultimate HDR” so it might be why but it also mean I have the “most customers” experience
unfair example as it streaming and not a full size bluray but watch netflix 2022 Troll in HD then 4K, if the 4K version looks better something is wrong with your tv settings >< the 4K version has a 4x larger filesize but somehow looks half the res of HD on a 4K tv :confused:

talking about wasting time, here are the results of my downclocked 4090 testing

  • when you reach the cpu bottleneck stage, background apps for example a web-browser make you lose more fps than whatever your gpu does so there’s no point trying to downclock it to reduce the strain on the cpu
  • downclocking gave me no good result other than lowering gpu watts which in this generation is low enough unless you play maxed RT 4K
  • I expected to have a systematic cpu watts increase with oc and show the “strain” I guessed but it wasn’t so
  • ocing those RTX 4000 at least in 1080p does not seem to be so great other than to get benchmarking scores, the watts increase is real but the performance gain not really worth it

let’s be real I’m not saying 1080p is the best use for this videocard okay lol I know it’s overkill for now
but the two upgrades I see are either the incoming 1440p 27" oled is 240Hz and the other would be a 4K 120Hz both tough goals to achieve

here are the obviously cpu limited benches updated with the watts and -502Mhz downclock on both the gpu and memory 2400/20000 scores
i removed the ovhe overheated results as they messed up the temps and in turn the watts

CloudGate repa score/graph/physi : 83843/180398/29180 *** watts cpu/gpu 201.21/268.32
CloudGate rpoc score/graph/physi : 82848/180732/28612 *** watts cpu/gpu 206.22/306.58
CloudGate dwcl score/graph/physi : 83533/181766/28889 *** watts cpu/gpu 205.41/240.79

Borderlands 3 1080p Badass DX11 repa : 188.44 *** watts cpu/gpu 143.74/401.32
Borderlands 3 1080p Badass DX11 rpoc : 188.28 *** watts cpu/gpu 133.58/451.00
Borderlands 3 1080p Badass DX11 dwcl : 188.59 *** watts cpu/gpu 126.45/349.49
Borderlands 3 1080p Badass DX12 repa : 283.69 *** watts cpu/gpu 174.93/386.00
Borderlands 3 1080p Badass DX12 rpoc : 284.69 *** watts cpu/gpu 160.79/436.72
Borderlands 3 1080p Ultra DX11 repa : 190.39 *** watts cpu/gpu 172.33/400.88
Borderlands 3 1080p Ultra DX11 rpoc : 190.25 *** watts cpu/gpu 131.03/444.74
Borderlands 3 1080p Ultra DX11 dwcl : 189.99 *** watts cpu/gpu 130.66/357.46
Borderlands 3 1080p Ultra DX12 repa : 282.21 *** watts cpu/gpu 195.94/393.21
Borderlands 3 1080p Ultra DX12 rpoc : 288.83 *** watts cpu/gpu 214.41/440.49
(i left the DX12 results for comparison, didn’t try downclocking by that stage it was clear it had no practical use, better use sdtock settings)

Shadow of the Tomb Raider 1080p Highest DLSS Quality repa AvgFPS/GpuBound% : 270/0 *** watts cpu/gpu 185.79/261.14
Shadow of the Tomb Raider 1080p Highest DLSS Quality rpoc AvgFPS/GpuBound% : 269/0 *** watts cpu/gpu 185.16/281.52
Shadow of the Tomb Raider 1080p Highest DLSS Quality dwcl AvgFPS/GpuBound% : 269/0 *** watts cpu/gpu 178.39/236.99
Shadow of the Tomb Raider 1080p Highest DLSS none repa AvgFPS/GpuBound% : 273/0 *** watts cpu/gpu 181.62/293.49
Shadow of the Tomb Raider 1080p Highest DLSS none rpoc AvgFPS/GpuBound% : 265/0 *** watts cpu/gpu 190.43/321.86
Shadow of the Tomb Raider 1080p Highest DLSS none dwcl AvgFPS/GpuBound% : 271/0 *** watts cpu/gpu 182.46/281.30

Unigine Superposition 720pLow repa : 53085 *** watts cpu/gpu 82.38/144.33
Unigine Superposition 720pLow rpoc : 52563 *** watts cpu/gpu 91.54/130.51
Unigine Superposition 720pLow dwcl : 52731 *** watts cpu/gpu 86.51/148.36

if you don’t want to read this mess I understand so here’s a single comparison that says all about the current cpu/gpu struggle
borderlands 3 1080p badasss
DX11 188.44fps for 401.32watts
DX12 283.69fps for 386.00watts
+100fps for -15watts on the gpu strange times

p.s. when you worked a long time in a technical domain you think in weird ways, finding problems is my work so I see everything in a negative way sorry :wink:

I will say down clocked the 4090 can have great performance/watt. So if you undervolt and down clock to 3090 performance you can run that card for 1/3 the power of stock.

in 1080p or 1440p if you use vsync or alternative it’s not necessary at all
the 4090 is more efficient than any other videocard
it comes in great part by being underused just like a car engine at 2000rpms uses less gas than at 8000
this a a pretty good demo but I notice it in all the games unless I unlock fps and then the card will just generate frames until you hit power limit


having upgraded from a 3090 I can tell you the difference in efficiency is HUGE for a 1 gen gap
the ascent a cyberpunk indy diablo like has a very pretty town hub with heavy RT effects on the 3090 I had 830w@psu and the heatwave coming out of the pc with it…same thing with a 4090…530w O_O it’s not only efficiency but the doubling of rt cores (I believe) because I don’t play
many RT games but I noticed the wattage is much less in 4000 series

as of overclocking the 4090 I’m now running 3000-3150MHz (whatever it decides really I added +120 in afterburner but real values change all the time) and 24Ghz memory (more = crashed)
I’m not exagerating it runs +100w 400 instead of 300, 300 instead of 200 and well never saw 100w under a certain load the card downlocks and downvolts almost to idle running 60w

I wonder what kind of wattage a viiista HCP has ? wink wink

Sure, I get your main argument, but here is a counter argument… If I save 100W and game, say, 50 hours every month, that is 5 kWh saved.

How many years will that be before I save the $1200 difference between an RTX 4090 and an RX 6700 XT?

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.