new 7700x/6700xt build sucking too many watts

C

cardboardbox

Weaksauce
Joined
Apr 4, 2014
Messages
105
I dont know if I did anything wrong but my new 7700x/6700xt build is sucking 2x or more power than my previous gaming rig. Here's what I have now:

7700x
XFX 6700xt
32gb ram
1 m2 ssd
1 sata ssd
1 2.5" hard drive (will be removed later this year when I add another m2 ssd)
550w psu

I know the PSU is weak and I will upgrade it soon, thought I could get away with it.

Here is my old rig for reference, try not to laugh, lol:

i7-4970
Sapphire 5700xt
16gb ddr3
2 sata ssd
600w psu

I'm testing with some older games, right now its The Outer Worlds Spacer's Choice Edition. The 7700x build is at 700w at the main menu while the old 4970 rig is at 345w. I know my newer parts use more power but not 350w more. I'm using the same settings in the game on both systems. The reason I started checking out power consumption is the new build loses video output regularly so I got suspicious that my power supply wasnt supplying enough power.

Any idea what I'm doing wrong or is the power consumption about what you'd expect?

edit: just realized that I'm using a 600w in my old pc, I'm an idiot. Will upgrade the PSU in the new pc today if I have time

edit2: This website says I need a 500w PSU https://outervision.com/power-supply-calculator

edit3: I changed maximum cpu state to 99%, that saved 30 watts, did the auto undervolt in Adrenalin, that saved another 25, but then I capped my fps at 60 in the game and that saved 200 watts. Its been a loooong time since I had a pc that would give me 100 fps at any resolution so I didnt realize to do this. 60 fps is fine for me. I'm still curious how I'm idling at 130-160watts though. Adrenalin shows my cpu using 0.7w and gpu 7w.
 
Last edited:
Pushing over 80% of a psus capacity will reduce its efficiency and you will get more power use at the wall then the componants are using.

Also 700w use from a 550w psu doesn't sound very smart.

Just set the power limit targets for the CPU and GPU lower and let then throttle. As evident by the 200w difference capping frame rate, you are well into the proformance slope of the power/proformance curve
 
To reduce power consumption, I would suggest looking at the core and soc voltages in the BIOS and see if they're out of spec. There are quite a few AM5 boards that push way too much voltage, which resulted in the widespread AM5 CPU death coverage in the news, where CPUs were quite literally exploding. Some motherboards have BIOS updates available that are supposed to fix, or at least make better the power issues. You may want to check if there's a firmware update for your board.

Seems to be a problem on AM4 as well, I just got a 5700G for my network server and noticed it was idling in the mid 50s, which was ridiculous. Ruled out the heatsink, and noticed in the BIOS that it was pushing a 1.43v core voltage, which was far too high. Looked at the settings in the voltage menu, and it was set to 1.35v. Typical Asus and their garbage voltage regulation. I was able to use voltage offset of like -0.8v to get it down to the proper 1.35v core and temperatures went from the mid 50s to the mid 30s. Power consumption at idle also went way down.
 
cardboardbox said:
I dont know if I did anything wrong but my new 7700x/6700xt build is sucking 2x or more power than my previous gaming rig. Here's what I have now:

7700x
XFX 6700xt
32gb ram
1 m2 ssd
1 sata ssd
1 2.5" hard drive (will be removed later this year when I add another m2 ssd)
550w psu

I know the PSU is weak and I will upgrade it soon, thought I could get away with it.

Here is my old rig for reference, try not to laugh, lol:

i7-4970
Sapphire 5700xt
16gb ddr3
2 sata ssd
600w psu

I'm testing with some older games, right now its The Outer Worlds Spacer's Choice Edition. The 7700x build is at 700w at the main menu while the old 4970 rig is at 345w. I know my newer parts use more power but not 350w more. I'm using the same settings in the game on both systems. The reason I started checking out power consumption is the new build loses video output regularly so I got suspicious that my power supply wasnt supplying enough power.

Any idea what I'm doing wrong or is the power consumption about what you'd expect?

edit: just realized that I'm using a 600w in my old pc, I'm an idiot. Will upgrade the PSU in the new pc today if I have time

edit2: This website says I need a 500w PSU https://outervision.com/power-supply-calculator

edit3: I changed maximum cpu state to 99%, that saved 30 watts, did the auto undervolt in Adrenalin, that saved another 25, but then I capped my fps at 60 in the game and that saved 200 watts. Its been a loooong time since I had a pc that would give me 100 fps at any resolution so I didnt realize to do this. 60 fps is fine for me. I'm still curious how I'm idling at 130-160watts though. Adrenalin shows my cpu using 0.7w and gpu 7w.
Click to expand...
My Intel 12700K idles at 182 W hits 600 when GPU runs full tilt.

https://www.amd.com/en/products/cpu/amd-ryzen-7-7700x

16
Max. Boost Clock
Up to 5.4GHz
Base Clock
4.5GHz
L1 Cache
512KB
L2 Cache
8MB
L3 Cache
32MB
Default TDP
105W
Processor Technology for CPU Cores
TSMC 5nm FinFET
Unlocked for Overclocking
Yes
CPU Socket
AM5
Thermal Solution (PIB)
Not included
Recommended Cooler
Premium air cooler recommended for optimal performance
Max. Operating Temperature (Tjmax)
95°C
 
Last edited:
cdabc123 said:
Pushing over 80% of a psus capacity will reduce its efficiency and you will get more power use at the wall then the componants are using.

Also 700w use from a 550w psu doesn't sound very smart.

Just set the power limit targets for the CPU and GPU lower and let then throttle. As evident by the 200w difference capping frame rate, you are well into the proformance slope of the power/proformance curve
Click to expand...
I'll have to figure out how to set power limit targets. I'm already running the cpu on eco mode in ryzen master and I've done the auto undervolt in Adrenalin.

I know 700w from a 550w isnt smart but in my defense, the psu calculators said 550w would be enough. I'll switch to a 650w gold I have as a backup for my work pc and pray my work pc psu doesnt die before I find a nice backup on sale.

Or if I knew how to set power limits I guess I could just do that and keep using the 550w.
 
GiGaBiTe said:
To reduce power consumption, I would suggest looking at the core and soc voltages in the BIOS and see if they're out of spec. There are quite a few AM5 boards that push way too much voltage, which resulted in the widespread AM5 CPU death coverage in the news, where CPUs were quite literally exploding. Some motherboards have BIOS updates available that are supposed to fix, or at least make better the power issues. You may want to check if there's a firmware update for your board.

Seems to be a problem on AM4 as well, I just got a 5700G for my network server and noticed it was idling in the mid 50s, which was ridiculous. Ruled out the heatsink, and noticed in the BIOS that it was pushing a 1.43v core voltage, which was far too high. Looked at the settings in the voltage menu, and it was set to 1.35v. Typical Asus and their garbage voltage regulation. I was able to use voltage offset of like -0.8v to get it down to the proper 1.35v core and temperatures went from the mid 50s to the mid 30s. Power consumption at idle also went way down.
Click to expand...
I'll look in the bios tonight but I'm not sure what numbers would be in spec. I just did the latest BIOS update and updated the mobo drivers. Idle power consumption is now about 105w. Still going past 700w in some games though. The Last of Us main menu is using 740w. :(
 
drutman said:
My Intel 12700K idles at 182 W hits 600 when GPU runs full tilt.

https://www.amd.com/en/products/cpu/amd-ryzen-7-7700x

16
Max. Boost Clock
Up to 5.4GHz
Base Clock
4.5GHz
L1 Cache
512KB
L2 Cache
8MB
L3 Cache
32MB
Default TDP
105W
Processor Technology for CPU Cores
TSMC 5nm FinFET
Unlocked for Overclocking
Yes
CPU Socket
AM5
Thermal Solution (PIB)
Not included
Recommended Cooler
Premium air cooler recommended for optimal performance
Max. Operating Temperature (Tjmax)
95°C
Click to expand...
I have one of those as a work pc in my office. I'll check the power consumption at some point but it doesnt work that hard. No gaming so far.
 
cdabc123 said:
Pushing over 80% of a psus capacity will reduce its efficiency and you will get more power use at the wall then the componants are using.

Also 700w use from a 550w psu doesn't sound very smart.

Just set the power limit targets for the CPU and GPU lower and let then throttle. As evident by the 200w difference capping frame rate, you are well into the proformance slope of the power/proformance curve
Click to expand...
A good power supply will remain better than 90% efficient as it approaches its upper power limit.
 
The biggest problem there is AMD"s implementation of the chiplet design. The memory-I/O controller that's part of the CPU package sucks up more than 20 to 30 watts on its own even when absolutely nothing at all whatsoever is going on. Whereas your old Intel system's CPU package eats up less than 10 watts at idle (and that's for the entire CPU package). This is the reason why I switched my main PC from AMD back to Intel with the arrival of the 12th-Gen (aka Alder Lake) CPUs.

And to correct the OP, there has never been an i7-4970 for desktop use. It's actually an i7-4790 (with or without the "K"), which has only 4 cores and 8 threads. That user's current Ryzen 7 7700X has twice as many CPU cores and twice as many threads, which itself accounts for a large percentage of the power consumption increase.
 
I switched to a 650w gold PSU but I'm still losing video output when starting a game maybe 25-50% of the time. It doesnt seem to be just games that are using a lot of power. Going to swap out the 6700 XT for a 5700 XT next and if that doesnt solve the problem I give up.
 
pendragon1 said:
did you not say that its pulling over 700w? just get a proper sized psu ffs...
Click to expand...
why should I go higher when every psu calculator says 500w or 550w is enough? I'm clearly missing a setting of some sort to bring down the power consumption to a normal range.
 
cardboardbox said:
why should I go higher when every psu calculator says 500w or 550w is enough? I'm clearly missing a setting of some sort to bring down the power consumption to a normal range.
Click to expand...

Power calculators are generally made by people that don't know that much about hardware, and take the data the manufacturers give them.

Problem is that manufacturers regularly lie about their power consumption, or use flawed metrics to come up with some "TDP" number that is not at all representative of real world use cases. To add to that, AIBs and motherboard manufacturers regularly ignore design specifications of the OEMs to get themselves higher in the performance charts, which further pushes up power consumption.

If you want a safe power supply margin, take the power figures from all of the components you're using and double them. You'll get closer to real world power usage. Sometimes you'll have to triple it, an example would be my i9-10850k. Intel gives it a 125W TDP, but it will regularly pull over 300W under peak loads.
 
cardboardbox said:
why should I go higher when every psu calculator says 500w or 550w is enough? I'm clearly missing a setting of some sort to bring down the power consumption to a normal range.
Click to expand...

Nobody in the history of building computers has ever trusted those PSU calculators. They do nothing but drive clicks to their site so they can put ads for power supplies in your face. Your rig needs 700-800W PSU to be in the efficiency band. You can get one from any good brand for $89-99 with no issues.
 
I will be on the lookout for a 750-800w PSU. In the meantime I solved my problem by uninstalling Ryzen Master. I was only using it to put my cpu in eco mode. As soon as it was gone, no more issues. I'm measuring my power consumption, will only be playing games that are consuming under 600w until I can upgrade the PSU.
 
Depends what you do a 7700x while gaming can stay under 85w, a 6700xt under 250w, say 350w for both, 100w for everything else...

Should be true that 550w would be more than enough, where does the 700w value come from ? Is it off the wall ?
 
Seem to come from Adrenaline ? If so, having a breakdown of what consume what if possible would be nice, cpu, gpu, others ?
 
LukeTbk said:
Seem to come from Adrenaline ? If so, having a breakdown of what consume what if possible would be nice, cpu, gpu, others ?
Click to expand...
Adrenaline usually only specifies consumption for GPU only.

Because I don't have PCAT / any tools for direct measurement to component / to the wall, I myself usually run both GPU-Z and HWInfo to log the lowest and highest consumption of CPU and GPU.
 
riev90 said:
Because I don't have PCAT / any tools for direct measurement to component / to the wall, I myself usually run both GPU-Z and HWInfo to log the lowest and highest consumption of CPU and GPU.
Click to expand...
I would suspect the methodology could be a bit flawed, achieving 700 watt from only the CPU and GPU and from the chip on a 550 PSU would be impressive to start with (would that mean getting close to 780-800 watt off the wall ?), even if the value are perfect maybe they did not happen at the same time.

How high the cpu and how high the GPU seem to go, are you playing a game during that evaluation or using your computer for something that push CPU more ?
 
LukeTbk said:
Depends what you do a 7700x while gaming can stay under 85w, a 6700xt under 250w, say 350w for both, 100w for everything else...

Should be true that 550w would be more than enough, where does the 700w value come from ? Is it off the wall ?
Click to expand...
Sorry for the delayed response, been on vacation.

I use a killawatt to measure the watts so that number is at the wall.
 
LukeTbk said:
I would suspect the methodology could be a bit flawed, achieving 700 watt from only the CPU and GPU and from the chip on a 550 PSU would be impressive to start with (would that mean getting close to 780-800 watt off the wall ?), even if the value are perfect maybe they did not happen at the same time.

How high the cpu and how high the GPU seem to go, are you playing a game during that evaluation or using your computer for something that push CPU more ?
Click to expand...
Yes I was playing a game. Its a bronze psu, so 700 at the wall is 560? Maybe there's a little extra wiggle room before it melts? lol
 
cardboardbox said:
Yes I was playing a game. Its a bronze psu, so 700 at the wall is 560? Maybe there's a little extra wiggle room before it melts? lol
Click to expand...

Can't know what the real power draw is without using equipment to measure it.

The 80PLUS standard doesn't apply to the real world, the testing conditions they use are entirely unrealistic and can't be replicated outside the testing parameters they use. Unless your PSU is in open air, with no obstructions and has 68 degree intake air at all times, the numbers they give are meaningless.
 
GiGaBiTe said:
Can't know what the real power draw is without using equipment to measure it.

The 80PLUS standard doesn't apply to the real world, the testing conditions they use are entirely unrealistic and can't be replicated outside the testing parameters they use. Unless your PSU is in open air, with no obstructions and has 68 degree intake air at all times, the numbers they give are meaningless.
Click to expand...
Not meaningless, just specific to comparisons between units in a controlled environment, much like all benchmarking,
 
vick1000 said:
Not meaningless, just specific to comparisons between units in a controlled environment, much like all benchmarking,
Click to expand...

That's not the problem. The problem is that Plug Load Solutions created the 80PLUS standard and turned around and said their testing methodology applies to every day use case scenarios, which is not true at all.

Which is what I was pointing out in the above post, the guy was saying he had a "bronze rated" PSU and was trying to apply flawed testing methodology to gauge what the power draw of his system was. It won't work.
 
GiGaBiTe said:
That's not the problem. The problem is that Plug Load Solutions created the 80PLUS standard and turned around and said their testing methodology applies to every day use case scenarios, which is not true at all.

Which is what I was pointing out in the above post, the guy was saying he had a "bronze rated" PSU and was trying to apply flawed testing methodology to gauge what the power draw of his system was. It won't work.
Click to expand...
But the testing is not meaningless. In the same scenario, if you replace a bronze rated unit with a gold, the gold will have higher efficiency. Same as if you replace a 12900K with a 13900K. You bench numbers may not equal a reviewer's numbers, but you will see a similar increase percentile over the 12900K.
 
vick1000 said:
But the testing is not meaningless. In the same scenario, if you replace a bronze rated unit with a gold, the gold will have higher efficiency.
Click to expand...

You have a fundamental lack of knowledge on how power supplies function if you think this is true, which it is not. Power supply efficiency does not scale linearly with ambient temperature, you cannot make the claim that a gold unit will be better than a bronze unit outside strict testing procedures. Your weird conclusion that power supplies somehow equate to different CPU models is confounding and plain wrong.
 
GiGaBiTe said:
You have a fundamental lack of knowledge on how power supplies function if you think this is true, which it is not. Power supply efficiency does not scale linearly with ambient temperature, you cannot make the claim that a gold unit will be better than a bronze unit outside strict testing procedures. Your weird conclusion that power supplies somehow equate to different CPU models is confounding and plain wrong.
Click to expand...
The 80+ ratings exist for a reason. They are supposed to be tested and confirmed by the manufacturer before certification. They most certainly CAN be used to compare units, provided there is no deception by the tester. The 80+ certification is standardized within a set of parameters, if you test outside of those parameters, of course you my get anomalous results. Poor input voltage or heat will effect efficiency, or course. But it doesn't make a bronze rated unit better than a gold rated, as long as they were properly evaluated to get those certifications. If you can find a comparison which shows a bronze unit out performing a gold (same wattage rating) in any high load scenario, high ambient or not, let's see it. I call BS on your claim. I have been building PCs since 1996, and have modified, repaired or restored many audio/guitar tube amplifiers. I understand electronics completely.

The CPU comparison is about benchmarking, meaning performance gains should be present, even among different systems, though not identical. So you see a comparison between units performed correctly online, you can expect similar percentage of results even in a different system. That's what benchmarks are for, and how we make buying decisions.
 
vick1000 said:
The 80+ ratings exist for a reason. They are supposed to be tested and confirmed by the manufacturer before certification.
Click to expand...

Yes, 80PLUS exists to make Plug Load Solutions money, easy money at that. They make the money machine go Brrrr. The last I checked, they got something like $15 grand per unit tested. 80PLUS has no benefit to the consumer, other than slightly raising the cost of the unit. It doesn't tell them the quality of the unit, how safe it is to use, or how it will operate in real world conditions. PERFECT EXAMPLE is Gigabyte's GP-P750GM and 850W units. They had 80PLUS Gold ratings, yet they have a nasty habit of exploding randomly because of design faults and killing thousands of dollars in powered equipment.

PLS doesn't do anything of substance to police their own standards for abuse or fraud. They don't go out on marketplaces and get counterfeit 80PLUS rated units pulled from sale, or copycats trying to pretend to be them.

This makes the 80PLUS standard worthless.

vick1000 said:
They most certainly CAN be used to compare units, provided there is no deception by the tester. The 80+ certification is standardized within a set of parameters, if you test outside of those parameters, of course you my get anomalous results.
Click to expand...

Why are you even arguing when you agree with my point that I originally made?

vick1000 said:
Poor input voltage or heat will effect efficiency, or course. But it doesn't make a bronze rated unit better than a gold rated, as long as they were properly evaluated to get those certifications. If you can find a comparison which shows a bronze unit out performing a gold (same wattage rating) in any high load scenario, high ambient or not, let's see it. I call BS on your claim. I have been building PCs since 1996, and have modified, repaired or restored many audio/guitar tube amplifiers. I understand electronics completely.
Click to expand...

The efficiency difference between a bronze and a gold is AT MOST 5% under IDEAL conditions. This can easily be eaten up by thermal performance loss.

80plus.png


Audio amplifers are not switch mode power supplies. Mosfets are not linear devices when it comes to output vs case and ambient temperature. You don't need fancy charts to see that a gold rated unit can be derated to bronze efficiency by heat, you can just look at the specifications of the mosfets used in the supply.

1692567669630-png.591980


Here's a random mosfet I pulled the datasheet from one of my suppliers. At room temperature (what PLS does), you can pull 200 amps. At 50C, which is typical inside many PSUs inside of hotbox computer cases, you lose around 10 amps of capacity, and the drain-source junction resistance increases. In reality, power supplies can operate well over 50C, so you have further derating and higher junction resistances, which means less efficiency. PLS doesn't test this at all, and what I was talking about in my original post that you still don't seem to understand.
 

Attachments

  • 1692567669630.png
    1692567669630.png
    30.7 KB · Views: 0
GiGaBiTe said:
Yes, 80PLUS exists to make Plug Load Solutions money, easy money at that. They make the money machine go Brrrr. The last I checked, they got something like $15 grand per unit tested. 80PLUS has no benefit to the consumer, other than slightly raising the cost of the unit. It doesn't tell them the quality of the unit, how safe it is to use, or how it will operate in real world conditions. PERFECT EXAMPLE is Gigabyte's GP-P750GM and 850W units. They had 80PLUS Gold ratings, yet they have a nasty habit of exploding randomly because of design faults and killing thousands of dollars in powered equipment.

PLS doesn't do anything of substance to police their own standards for abuse or fraud. They don't go out on marketplaces and get counterfeit 80PLUS rated units pulled from sale, or copycats trying to pretend to be them.

This makes the 80PLUS standard worthless.



Why are you even arguing when you agree with my point that I originally made?



The efficiency difference between a bronze and a gold is AT MOST 5% under IDEAL conditions. This can easily be eaten up by thermal performance loss.

View attachment 591983

Audio amplifers are not switch mode power supplies. Mosfets are not linear devices when it comes to output vs case and ambient temperature. You don't need fancy charts to see that a gold rated unit can be derated to bronze efficiency by heat, you can just look at the specifications of the mosfets used in the supply.

1692567669630-png.591980


Here's a random mosfet I pulled the datasheet from one of my suppliers. At room temperature (what PLS does), you can pull 200 amps. At 50C, which is typical inside many PSUs inside of hotbox computer cases, you lose around 10 amps of capacity, and the drain-source junction resistance increases. In reality, power supplies can operate well over 50C, so you have further derating and higher junction resistances, which means less efficiency. PLS doesn't test this at all, and what I was talking about in my original post that you still don't seem to understand.
Click to expand...
In most of the testing of power supplies I've seen the efficiency is often better than the 80 PLUS definition at 100% load. I consider it the worst-case scenario. Still, I wouldn't be budgeting my power supply to match the maximum output required by my PC. You always want some headroom. If OP needs 560W, then I'd be looking at a 750W PSU to account for situations needing more power and to handle transients. 10% loss of output due to high temps still leave you with 675W to work with.

Anandtech's hotbox testing is run at 45C ambient.

1692979100098.png

1692979243421.png
 
You must log in or register to reply here.
Back
Top