• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

RTX 4070 - efficiency & undervolting

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.18/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
Cooling Alphacool Apex UV - Alphacool Eisblock XPX Aurora + EK Quantum ARGB 3090 w/ active backplate
Memory 2x32GB DDR4 3600 Corsair Vengeance RGB @3866 C18-22-22-22-42 TRFC704 (1.4V Hynix MJR - SoC 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
Case Fractal Design R6
Audio Device(s) Logitech G560 | Corsair Void pro RGB |Blue Yeti mic
Power Supply Fractal Ion+ 2 860W (Platinum) (This thing is God-tier. Silent and TINY)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE ( Sexy white keycaps)
VR HMD Oculus Rift S + Quest 2
Software Windows 11 pro x64 (Yes, it's genuinely a good OS) OpenRGB - ditch the branded bloatware!
Benchmark Scores Nyooom.
Anyone has ideal settings for a 3070? What are thoughts about 3070 undervolting compared to a 4070 with as purpose ocasionally game at high settings (not ultra) in 1440p.
As I find the price gap between 3070 and 4070 significant, is dls 3 that much better than dls 2, better said is a 3070 not future proof enough?
It's different for every single physically existing GPU, let alone every single GPU variant or design. No two are identical.
Run a game/test that lets the GPU run at 100%, and see the MHz value and voltage it settles at once it's heated up and hitting power and/or thermal limits.

Then lock the curve flat to that value, and keep re-trying that same MHz value with lower and lower voltages. Spend time on each adjustment, and test multiple programs. What is stable in a DX11 test may not be stable in a DX12 test, and vice versa.

I'll see if i can find whats stable on my 3070, it's stable but i haven't bothered testing it out for more efficiency or higher clocks since it's running quite well on air already.

Edit:
1620MHz, 0.737v
Screenshot 2023-07-29 135546.png


I cant remember the wattage values, but i do remember it had it running less wattage than the stock wattage of the GTX 1080 it replaced
 
Last edited:
Joined
Dec 12, 2012
Messages
721 (0.17/day)
Location
Poland
System Name THU
Processor Intel Core i5-13600KF
Motherboard ASUS PRIME Z790-P D4
Cooling SilentiumPC Fortis 3 v2 + Arctic Cooling MX-2
Memory Crucial Ballistix 2x16 GB DDR4-3600 CL16 (dual rank)
Video Card(s) MSI GeForce RTX 4070 Ventus 3X OC 12 GB GDDR6X (2610/21000 @ 0.91 V)
Storage Lexar NM790 2 TB + Corsair MP510 960 GB + PNY XLR8 CS3030 500 GB + Toshiba E300 3 TB
Display(s) LG OLED C8 55" + ASUS VP229Q
Case Fractal Design Define R6
Audio Device(s) Yamaha RX-V381 + Monitor Audio Bronze 6 + Bronze FX | FiiO E10K-TC + Sony MDR-7506
Power Supply Corsair RM650
Mouse Logitech M705 Marathon
Keyboard Corsair K55 RGB PRO
Software Windows 10 Home
Benchmark Scores Benchmarks in 2024?
1620 MHz on a 3070 seems excessive. That card only has a 220 W TDP. And stock boost clock is higher than a 3080, so you'll losing as much as 15-20% performance in games with a lighter load.

@Chippendale Unless your card is really loud, I'd try going for 1905 MHz at around 0.9 V, or 1800 MHz at around 0.8 V.
For heavier games the first option would actually offer more performance at similar power to stock. The second option should drop power usage across the board, but you'd lose a bit of performance in lighter games. It's worth having two profiles and switching them depending on the game.

What I've personally used:
1080 - 1905 MHz @ 0.9 V
2070 S - 1905 MHz @ 0.95 V
3080 - 1800 MHz @ 0.8 V, second profile 1905 MHz @ 0.9 V (pretty much never used this one as the card was ridiculously power hungry)
4070 - 2610 MHz @ 0.91 V, second profile 2805 MHz @ 1.0 V (didn't need to use this one yet)
 

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,593 (2.35/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6000CL30┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Display(s) 43" QN90B / 32" M32Q / 27" S2721DGF
Case Caselabs S3┃Lazer3D HT5
1620 MHz on a 3070 seems excessive. That card only has a 220 W TDP. And stock boost clock is higher than a 3080, so you'll losing as much as 15-20% performance in games with a lighter load.

@Chippendale Unless your card is really loud, I'd try going for 1905 MHz at around 0.9 V, or 1800 MHz at around 0.8 V.
For heavier games the first option would actually offer more performance at similar power to stock. The second option should drop power usage across the board, but you'd lose a bit of performance in lighter games. It's worth having two profiles and switching them depending on the game.

What I've personally used:
1080 - 1905 MHz @ 0.9 V
2070 S - 1905 MHz @ 0.95 V
3080 - 1800 MHz @ 0.8 V, second profile 1905 MHz @ 0.9 V (pretty much never used this one as the card was ridiculously power hungry)
4070 - 2610 MHz @ 0.91 V, second profile 2805 MHz @ 1.0 V (didn't need to use this one yet)

Same here, I was 1920 @ 0.98V on the 2060 Super and 1980 @ 0.93V on the 3070 Ti. Before that was like 1900 @ 0.93V on the 1070 I think. The GA104 on that profile was essentially a 180-220W card. No need to dip so low to 1600 for a regular non-Ti.

I am really behind, I still gotta put together a undervolt curve for my 4070 Ti. Just haven't felt the need for Afterburner yet as it by default doesn't run as aggressive as the 3070 Ti. Ada also seems to throw a bit of a wrench into the works with its higher Vcore.
 

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.18/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
Cooling Alphacool Apex UV - Alphacool Eisblock XPX Aurora + EK Quantum ARGB 3090 w/ active backplate
Memory 2x32GB DDR4 3600 Corsair Vengeance RGB @3866 C18-22-22-22-42 TRFC704 (1.4V Hynix MJR - SoC 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
Case Fractal Design R6
Audio Device(s) Logitech G560 | Corsair Void pro RGB |Blue Yeti mic
Power Supply Fractal Ion+ 2 860W (Platinum) (This thing is God-tier. Silent and TINY)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE ( Sexy white keycaps)
VR HMD Oculus Rift S + Quest 2
Software Windows 11 pro x64 (Yes, it's genuinely a good OS) OpenRGB - ditch the branded bloatware!
Benchmark Scores Nyooom.
1620 MHz on a 3070 seems excessive.
excessive meaning... what?
I'm happy with how it runs, doing 1440p 165Hz and my rift S like this
1690620283613.png
 
Joined
Feb 24, 2023
Messages
2,237 (5.21/day)
Location
Russian Wild West
System Name DLSS / YOLO-PC
Processor i5-12400F / 10600KF
Motherboard Gigabyte B760M DS3H / Z490 Vision D
Cooling Laminar RM1 / Gammaxx 400
Memory 32 GB DDR4-3200 / 16 GB DDR4-3333
Video Card(s) RX 6700 XT / RX 480 8 GB
Storage A couple SSDs, m.2 NVMe included / 240 GB CX1 + 1 TB WD HDD
Display(s) Compit HA2704 / Viewsonic VX3276-MHD-2
Case Matrexx 55 / Junkyard special
Audio Device(s) Want loud, use headphones. Want quiet, use satellites.
Power Supply Thermaltake 1000 W / FSP Epsilon 700 W / Corsair CX650M [backup]
Mouse Don't disturb, cheese eating in progress...
Keyboard Makes some noise. Probably onto something.
VR HMD I live in real reality and don't need a virtual one.
Software Windows 10 and 11
My 3070 can run at ~120W, the performance of a 2080Ti at the wattage of a 1660 Super
Same is possible on RX 6800 (if we don't count RT perfromance which doesn't exist on AMD cards below 7900 series)
 
Joined
Jul 13, 2023
Messages
106 (0.37/day)
System Name Game Rig 2
Processor 13700f
Motherboard Alienware R16 Stock
Cooling Alienware R16 Liquid Stock
Memory 16 GB R16 Stock
Video Card(s) 4060 TI Stock
Storage 2 TB Kioxa XG8 Stock
Display(s) ASUS PB287Q
Case R16 Stock
Power Supply R16 Stock 500 Watt
Mouse Razer Abyssus
Keyboard Cherry KC 6000 Slim
Software Micro 11 Home
It's different for every single physically existing GPU, let alone every single GPU variant or design. No two are identical.
Run a game/test that lets the GPU run at 100%, and see the MHz value and voltage it settles at once it's heated up and hitting power and/or thermal limits.

Then lock the curve flat to that value, and keep re-trying that same MHz value with lower and lower voltages. Spend time on each adjustment, and test multiple programs. What is stable in a DX11 test may not be stable in a DX12 test, and vice versa.

I'll see if i can find whats stable on my 3070, it's stable but i haven't bothered testing it out for more efficiency or higher clocks since it's running quite well on air already.

Edit:
1620MHz, 0.737v
View attachment 306712

I cant remember the wattage values, but i do remember it had it running less wattage than the stock wattage of the GTX 1080 it replaced
interesting ty
 
Joined
Apr 13, 2017
Messages
116 (0.05/day)
System Name AMD System
Processor Ryzen 7900 at 180Watts 5650 MHz, vdroop from 1.37V to 1.24V
Motherboard MSI MAG x670 Tomahawk Wifi
Cooling AIO240 for CPU, Wraith Prism's Fan for RAM but suspended above it without touching anything in case.
Memory 32GB dual channel Gskill DDR6000CL30 tuned for CL28, at 1.42Volts
Video Card(s) Msi Ventus 2x Rtx 4070 and Gigabyte Gaming Oc Rtx 4060 ti
Storage Samsung Evo 970
Display(s) Old 1080p 60FPS Samsung
Case Normal atx
Audio Device(s) Dunno
Power Supply 1200Watts
Mouse wireless & quiet
Keyboard wireless & quiet
VR HMD No
Software Windows 11
Benchmark Scores 1750 points in cinebench 2024 42k 43k gpu cpu points in timespy 50+ teraflops total compute power.
Ventus 2x oc edition won't let me do lower than 0.91V according to gpu-z. Below 910mV, it only reduces frequency. At 920mV, I'm using 2580 MHz, making it about 150Watts and cool the hot-spot below 90C (guessing 90C+ is bad for other components on mobo).
 
D

Deleted member 182555

Guest
asus dual 4060 biosmod... 0,675v-1,1v (0,88v-1,065v) voltage unlock 170w (+32w) tdp

biosmod also defeats extreme oc mod.... 12665 vs wip 12680 with my much worse CPU score around 1/4

curve.jpg
score.jpg
 
Last edited by a moderator:
Joined
Jun 6, 2022
Messages
622 (0.90/day)
3070 Ti
The custom profile is 26 months old. Rock stability.
It offers the same performance (results with +/- 1%, within the margin of error) with a lower consumption of ~50W only on the GPU. The advantages are obvious, from consumption at the outlet (~ -70W) to noise and temperatures.
I can use memory overclocking to reduce consumption even more, but... no. This video card must be used until at least 2025 because it killed my wallet in 2021.

Default: 284 W peak
Custom: 234 W peak
Note:
The nVidia recording program is not perfect but it is free. It affects the performance by about 2-5%, depending on the game.

 
Top