7800X3D, 7900X3D, 7950X3D..

[EightyDuce]

3 hours ago, some1 said:

That's a lot @EightyDuce, 300W more than my Pc with similar specs. Is that only the computer, or also including Monitor, and other things you have on the desk?

 

Just the tower and anything that's plugged into it. 

[some1]

11 minutes ago, Mac D said:

Just out of curiousity : how "correct" are these kind of PSU with measuring the wattage in comparision to digital power monitor ??

I think the PSU should be more accurate, and also give you the actual output power readout, while the power meters at the wall socket measure power draw including PSU efficiency losses, so they skew the numbers high. 

Regardless, even basic power meter that is off by a few % will get you in the right ballpark.

11 minutes ago, EightyDuce said:

Just the tower and anything that's plugged into it. 

Interesting I wonder what else in your PC can take so much power above what I got here. 

[EightyDuce]

6 minutes ago, some1 said:

I think the PSU should be more accurate, and also give you the actual output power readout, while the power meters at the wall socket measure power draw including PSU efficiency losses, so they skew the numbers high. 

Regardless, even basic power meter that is off by a few % will get you in the right ballpark.

Interesting I wonder what else in your PC can take so much power above what I got here. 

To be clear, the 880-917w (Cyberpunk 2077 RT Ultra) was from the wall while benching (built in benchmark). Same bench in RT Overdrive was pulling 940-947w. DCS Marianas F18 free-flight in VR was pulling 610-642w. Diablo 4 in-game, pulling 520-560w-ish.

 

If all you're doing is DCS, you have headroom. But some things may be more demanding. I'm actually surprised CyberPunk 2077 RT overdrive pulled that much power....makes me think I should have gone with a 1200w+ PSU instead of a 1000W lol.

 

@some1 Do you by chance have Cyberpunk 2077 and could run the built in benchmark to see if your power-draw spikes?

Edited June 27, 2023 by EightyDuce

[some1]

56 minutes ago, EightyDuce said:

 

@some1 Do you by chance have Cyberpunk 2077 and could run the built in benchmark to see if your power-draw spikes?

Sure, here's a benchmark run of Cyberpunk at the highest setting. 640W max, and that includes a 42" TV that's connected to the same UPS and everything else I have on my desk, not necessarily connected to the PC.

I mean you have a GPU that at stock setting won't take more than 450W, a CPU that takes less than 100W (during CP2077 benchmark it's more like 40-50W), so either you have something else inside your PC that eats additional 300W, a very inefficient PSU, or a power meter that is somewhat pessimistic.

1 hour ago, EightyDuce said:

makes me think I should have gone with a 1200w+ PSU instead of a 1000W lol.

Don't worry too much, even if the readings are true, you have to multiply them by PSU efficiency, so that's around 800W of actual PSU load.

[EightyDuce]

15 minutes ago, some1 said:

Sure, here's a benchmark run of Cyberpunk at the highest setting. 640W max, and that includes a 42" TV that's connected to the same UPS and everything else I have on my desk, not necessarily connected to the PC.

I mean you have a GPU that at stock setting won't take more than 450W, a CPU that takes less than 100W (during CP2077 benchmark it's more like 40-50W), so either you have something else inside your PC that eats additional 300W, a very inefficient PSU, or a power meter that is somewhat pessimistic.

Don't worry too much, even if the readings are true, you have to multiply them by PSU efficiency, so that's around 800W of actual PSU load.

PSU is an MSI MPG A1000G, so I don't think it's the issue. I'd be more inclined to think that the 15 year old kilowatt P3 may be the issue. Unfortunately I don't have another way to measure other then in-windows reporting and napkin math. 

 

Edit: so I just thought of something and grabbed one of the smart plugs that records power use/draw I had on my 3D printer. Cyberpunk 2077 RT overdrive benchmark indicates 463-480w draw at the wall. No idea how accurate these are (sengled smart plug), but it's a heck of a spread from the killawat P3. 

Edited June 27, 2023 by EightyDuce
Smart plug tested for what's it worth.

[kksnowbear]

You guys are measuring different things (at least somewhat) and it's affecting your results.  Apples and oranges, if you will.

For example - and this is just one example, mind you - it appears as if some1, being in Poland (unless I'm mistaken) is using 220v AC line (see HWinfo "UPS Input Voltage").  Meanwhile, EightyDeuce <sp>, I assume, is in the USA and (probably) therefore using 120v AC line.  PC switched-mode PSUs are usually much more efficient at 220v than 120.

That's one thing.

Also, one of you is reading UPS load via software that we probably cannot fully account for its calculations (in any event this is not "wall power", and although it might be a realistic representation of what the PC PSU draws, it's still software-driven).  The other appears to be using an inline meter device, which is (possibly...probably) measuring differently.

Another factor is how these devices (your measuring devices) actually determine load; i.e. is it just VxA (which is actually watts, but not necessarily true AC power), or is it something more akin to RMS calculation, with power factor considered?

Finally, although you both might have *similar* systems, you're not necessarily running the exact same tests on the exact same hardware.  This could matter a fair amount as well.  You both may well have (relatively) efficient power supplies...but the odds that they're exactly equal in efficiency, at the same moments in the same tests...well, the odds are not very high.

Different measuring devices, different types of load, different conditions, different efficiency...could easily explain the delta in your measurements**. FWIW.

Still a good idea to measure, don't get me wrong...just a lot of factors to consider even when you're using instruments.

**(EDIT: I think we can all agree that there obviously must be some explanation for the two grossly different numbers EightyDeuce got when using different devices - which should itself be telling.)

Edited June 27, 2023 by kksnowbear

[some1]

Sure, but we're not in a business where it has to be measured super accurately. 50W this or that way won't change the overall picture. EightyDeuce measurments are... something else

I also checked my UPS with an inline meter and it also showed similar numbers. Reading them from UPS software is just more convenient.

[kksnowbear]

29 minutes ago, some1 said:

Sure, but we're not in a business where it has to be measured super accurately. 50W this or that way won't change the overall picture. EightyDeuce measurments are... something else

I also checked my UPS with an inline meter and it also showed similar numbers. Reading them from UPS software is just more convenient.

Yes, and my comments were taking EightyDeuce's grossly different numbers into account.  Had they just been taken at face initially, it would be misleading.

Software is still software, a meter is still...not (necessarily) software...

My point was that doing this sort of thing, especially when making recommendations to those less knowledgeable, means it's important to account for differences.  Otherwise it leads to misunderstandings (which can get out of control in a hurry).

(Incidentally, to the question of whether it has to be 'super accurate'...well, all I can say is that if it's oversized by 1000 watts, it'll run just fine even if it's absurdly expensive, where if it's undersized - even by 50 watts -  it *will* change the overall picture.  It'll cause problems, and that's a fact.)

Circling back a moment to the question of what power supply is appropriate to a 7800x/4090 system...

Well...it depends.  You'd want to know which 4090, as even different cards from the same manufacturer can be rated differently.

For example, MSI recommends a 1000W power supply for their Suprim Liquid model, stating the minimum required is 850W, while the recommended PSU for a Gaming Trio 4090 is 850W.  So that's 150W difference in the recommended PSU from the manufacturer, both cards being 4090s.

Another (fairly important) thing to consider is that all switched-mode PSUs like those in our PCs operate most efficiently in the range of 50% load.  So, in other words, you could *double* the calculated PSU output and it's not "too much" (except maybe the cost, that is).

I am aware that people often throw money at (grossly) over-rated power supplies - it's actually one of my pet peeves, TBH.  But the absolute fact is that sizing a PSU's output capacity on even twice the calculated load is not really absurd (just pricey!)

Now, there are a lot of things to consider where trying to determine load - not the least of which being that the machine will seldom run at a maximum load (even when actually gaming), which affects determining what 50% is, of course...but it still applies that (especially if you're concerned with efficiency) you have to factor in how the equipment actually works.

Edited June 27, 2023 by kksnowbear

[EightyDuce]

I have an APS battery backup UPS coming Thursday for my server room but I'll hook it up to my PC and see what it reads. 

I'm not sure how accurate the sengled plug is, but the Kill-a-watt was, and from a quick Google, still is a popular measuring tool and folks report as accurate but there's no telling if it went off at some point of its 15 years of sitting in a drawer. 

Unfortunately until the UPS gets here I have no other way to measure, but the readings have been surprising if nothing else... Especially the disparity between the killawatt and the smart plug. 

[some1]

1 hour ago, kksnowbear said:

Software is still software, a meter is still...not (necessarily) software...

I can read the same numbers from a display directly on my UPS. It shows voltage, frequency, load in VA, load in Watts, and other things, same as an inline meter. It's just more convenient to display them on my PC screen sent through an USB connection. So is that a meter or a software?

Also you're diving into minutiae that are becoming less relevant to the discussion here. The efficiency curve at various loads, or 110V vs 230V difference is something that may save you a few cents in the long run, but it makes little difference when it comes to the choice of PSU output power. Here are example efficiency curves from Corsair manuals. You have something like 1% drop in efficiency between 50% and 75% load, and 2% drop between 110V and 230V. Plus the efficiency curve of the more powerful PSU is shifted left, so the difference between them at the same wattage is even smaller.

1 hour ago, kksnowbear said:

Well...it depends.  You'd want to know which 4090, as even different cards from the same manufacturer can be rated differently.

For example, MSI recommends a 1000W power supply for their Suprim Liquid model, stating the minimum required is 850W, while the recommended PSU for a Gaming Trio 4090 is 850W.  So that's 150W difference in the recommended PSU from the manufacturer, both cards being 4090s.

First of all you need to know which CPU will be used with the card. These recommendations come from the simple fact that these GPUs can be paired with any possible CPU. An 80W AMD, or a 250+W Intel. So in this context it makes sense that a card that's slightly more power hungry will require at least 850W PSU. But we're discussing 7800X3D, not Intels, and it this context, any 4090 model will fit in the recommended 850W. 

[kksnowbear]

What model/size backup did you buy, if I can ask? (I assume you mean "APC" as opposed to "APS")

Will there be loads besides the one PC?

[kksnowbear]

I said nothing that excludes considering other factors besides which GPU. In fact I said quite the opposite.

And unless you claim to know more than MSI does about their own product, your statement about "any 4090" is inaccurate. I just cited an example where they specifically recommend a 1000w PSU.

Not sure what you mean by "the recommended 850w", but that is not what some might recommend.  Maybe you're OK with recommending a PSU that barely meets (or fails to meet at all) the manufacturers recommendations...

My comment about 220 v 120 is solid, factual, and on point. It was also concerning why you and EightyDeuce were getting different measurements, not necessarily about choosing a PSU. Among other things, some people don't have 220v line to feed their PC. Yet it is still among the factors that matter.

It is one factor of many, just as i said.  Every statement I've made is 100% relevent to any discussion about PSUs. Can't help if you don't understand these things as well others might.

Starts to look like you're just touchy that you're wrong about power supplies. Unfortunate.

[EightyDuce]

50 minutes ago, kksnowbear said:

What model/size backup did you buy, if I can ask? (I assume you mean "APC" as opposed to "APS")

Will there be loads besides the one PC?

I thought it was an APC-branded UPS, however, it's actually a "CyberPower 1500VA / 900Watts True Sine Wave Uninterruptible Power Supply (UPS)". It's just going to have the Synology NAS, Unifi Dream Machine SE and two switches on it. We don't typically lose power for long but recently with the weather, have had frequent quick outages and the NAS hasn't been happy about it. 

[some1]

1 hour ago, kksnowbear said:

My comment about 220 v 120 is solid, factual, and on point. It was also concerning why you and EightyDeuce were getting different measurements, not necessarily about choosing a PSU

EightyDeuce is seeing 400W difference depending on which device he uses to measure his own PC. Same PC, same cables, same voltage. How's your talk about 220V vs 120V relevant to that? It's neither on point nor on topic. 

1 hour ago, kksnowbear said:

And unless you claim to know more than MSI does about their own product, your statement about "any 4090" is inaccurate. I just cited an example where they specifically recommend a 1000w PSU.

And I just explained to you why the manufacturers put 850W as minimum for these cards. The power draw of a Suprim Liquid card is no secret. This is a factory overclocked card set at 480W out of the box (+30W vs stock) and can be pushed up to 530W (hard limit in BIOS). It's well within capability of 850W PSU when paired with ~80W 7800X3D CPU, but becomes a tight squeeze when you install it in a system with something like the latest 250W Intel CPUs. Hence such manufacturer's recommendations. Nothing more, nothing less, these are just generic guidelines without considering the CPU type, so they play it safe.

But here we're discussing 7800X3D, which is a CPU with exceptionally low power draw for a modern high end CPU. It did not even exist back when those examples you cite were written and when recommendations were set.

Anyway, I'm done. It was a simple question for which I provided a simple answer based on my experience with my own PC, that somehow turned into two page argument with people who don't even have the hardware, yet like to theorize a lot.

Edited June 27, 2023 by some1

[kksnowbear]

The only point I made about 230 vs 120 is that it's more efficient, thus being among the reasons you two will see different power consumption.

Note well: I said "among the reasons".  I said it is becuase...it is.

Fact: 220 line is more efficient than 120.  And it most assuredly *is* on point once you turned the discussion to "why is your load different from mine if they're similar systems" (even though you did this while posting readings from a different device, measuring by a different means on a system that is...different.)

I cited a number of reasons that apply to your question. Simple: They're different.

And I went on to say it's important to identify the reasons for the differences...

...while you keep arguing with a manufacturer's specs for their own hardware, simply because it doesn't agree with your assertion that 850w is enough. Not according to MSI it's not.

But what does MSI know? They only design, manufacture, and sell *thousands* (millions??) of these things.

 

Edited June 29, 2023 by kksnowbear

[some1]

 

15 minutes ago, kksnowbear said:

...while you keep arguing with a manufacturer's specs for their own hardware, simply because it doesn't agree with your assertion that 850w is enough. Not according to MSI it's not.

According to MSI it is enough. It meets the minimum requirement.

https://www.msi.com/Graphics-Card/GeForce-RTX-4090-SUPRIM-LIQUID-X-24G/Specification

Now I wouldn't recommend using it with 250W Intel CPU in such configuration and I would recommend at least 1000W PSU for that, just like MSI does.... but we're not in a thread about Intel CPUs.

 

Edited June 27, 2023 by some1

[kksnowbear]

I said earlier (still there, go look) that it was recommended 1000w, required 850.

In spite of your obvious internet prowess lol, you're still off base...because using the minimum required PSU is always a good idea ("not") - especially in a system where someone drops a load of cash to put in a 4090 to begin with.

Two other AIB vendors with models of the 4090 that recommend 1000w PSU:
- Asus Strix OC edition
- Gigabyte 4090 Gaming OC 24G

Since eVGA is no longer in the GPU game, that pretty much covers the 'big three', here in the States anyhow.

You have no legitimate, confirmed information from an official source that proves which system those manufacturers were considering when they made their recommendations...

... do you?

So, there's no way for anyone here (including you) to say what they took into account.  And they don't specify "Well, if you have a 7800x, then it's really only 850"

And again, the recommendation is 1000w for (now three) GPU manufacturers' 4090 cards.

You're not arguing with me, bud - looks like you're arguing with them.

Edited June 27, 2023 by kksnowbear

[some1]

1 hour ago, kksnowbear said:

You have no legitimate, confirmed information from an official source that proves which system those manufacturers were considering when they made their recommendations...

I know for a fact that these recommendations are not based on 7800X3D, because these CPUs did not exist on the shelves back when 4090 cards went on sale. They are just three months old, while the cards were released nearly a year ago. And I know for a fact, that every single "high-end" CPU has higher peak power consumption than 7800X3D. Often two or three times higher.

1 hour ago, kksnowbear said:

So, there's no way for anyone here (including you) to say what they took into account.  And they don't specify "Well, if you have a 7800x, then it's really only 850"

And again, the recommendation is 1000w for (now three) GPU manufacturers' 4090 cards.

It's simple really, if 1kW PSU is recommended to run 4090 with something like Intel 12900k or 13700k (250W CPUs), then 850W PSU is perfectly fine with 80W 7800X3D. And if you think that one needs at least 1 kW PSU to run even with 7800X3D, then automatically you imply that all those recommendations from MSI, ASUS, Gigabyte are worthless, because they should say at least 1200W for Intel and high-end Ryzens (non-X3D).

Can't have one-size-fits-all recommendation when there's 200W difference between CPUs. 

 

Besides, I have already shown you how much power my 7800X3D system is using. Even if I put a hypothetical 600W GPU in it, which is the theoretical max for some of the RTX4090 cards and the limit of 12VHPWR connector, then it would still sit well below 850W. That's just the way it is with 7800X3D.

 

Edited June 27, 2023 by some1

[BitMaster]

I don't know why this is so complicated.

If one spents 1500€/US$ on a GPU plus a ton of money for CPU, Board, RAM, screen, VR, Chair, Hotas, Rudders, 25 Modules + 5 Maps.....why do you argue about 150watts ??? 

My guess is that those of you who favor the lesser wattage and likely lower ATX rating ( Bronze, Silver, Gold... ) PSU have not done PC repair for a living and seen what

people like me have encountered over the years.  Cutting corners on a PSU is a bad idea. Here, you rather follow the "50% Load, buy double that and make it Platinum too" advice

Your devices will be better protected by things like OCP, OVP, SCP, etc... plus in most cases much better components inside the PSU for a cleaner, healthier DC flow. 

 

I have such an example next to me, 1000w Titanium, 100€.  a joke of a PSU... 1080ti it killed and a nice MSI X570 board ....   cutting corners often ruins another day. 

I warned that guy more than once, tried to force him in buying a better PSU...he didnt follow my advice.   about 9 month later, see result 1 line above.

 

If you buy a 1500€ GPU and couple it with a 100€ PSU and then experience strange things....don't call my number, that's how I put it these days.

[some1]

20 minutes ago, BitMaster said:

I have such an example next to me, 1000w Titanium, 100€.  a joke of a PSU... 1080ti it killed and a nice MSI X570 board ....   cutting corners often ruins another day. 

I warned that guy more than once, tried to force him in buying a better PSU...he didnt follow my advice.   about 9 month later, see result 1 line above.

That's more of a counter example that a cheapo 1000W won't save you even if its well above "recommended" specs. The guy would have been better off buying something less powerful from a reputable brand for the similar price, rather than heeding your "50% Load, buy double that" advice. 1 kW to run X570 with 1080ti, really?

Edited June 27, 2023 by some1

[Thinder]

On 6/27/2023 at 9:37 PM, BitMaster said:

I don't know why this is so complicated.

If one spents 1500€/US$ on a GPU plus a ton of money for CPU, Board, RAM, screen, VR, Chair, Hotas, Rudders, 25 Modules + 5 Maps.....why do you argue about 150watts ??? 

My guess is that those of you who favor the lesser wattage and likely lower ATX rating ( Bronze, Silver, Gold... ) PSU have not done PC repair for a living and seen what

people like me have encountered over the years.  Cutting corners on a PSU is a bad idea. Here, you rather follow the "50% Load, buy double that and make it Platinum too" advice

Your devices will be better protected by things like OCP, OVP, SCP, etc... plus in most cases much better components inside the PSU for a cleaner, healthier DC flow. 

 

I have such an example next to me, 1000w Titanium, 100€.  a joke of a PSU... 1080ti it killed and a nice MSI X570 board ....   cutting corners often ruins another day. 

I warned that guy more than once, tried to force him in buying a better PSU...he didnt follow my advice.   about 9 month later, see result 1 line above.

 

If you buy a 1500€ GPU and couple it with a 100€ PSU and then experience strange things....don't call my number, that's how I put it these days.

Those were the original offers, Seller agreed to reduce to £90, I reduced mine to £85, the RM1000X happen to be brand new, no reason not to swap considering that the 850 wouldn't guaranty any headroom for future upgrade, like the 750 before it which I had to swap for the 850.

 

 

Just to give yet another idea of how the need for stronger power supply increases with new generations of technologies, my system is now DDR4, my GPU like all GPU actually in the market is PCI_E Gen 4, we have no idea of what a PCI_E Gen 5 system will need apart for some clues like this:

i913900k

An AMD Ryzen™ 9 7950X3D has a  Default TDP of 120W.

I'm not gonna upgrade my system for the stake of spending money, I will be looking for increased performances for both GPU and CPU, meaning very likely to be faster than either i9-13900KS, 7950-X3D and RX 7900 XTX, the only thing I might keep for a while would be the Cl14 RAM, here is why, this is the very reason why I chose to stick with a DDR4 system together with DDR5 RAM immaturity.

AMD Ryzen 9 7950X3D Initial Performance: AM5 Enters the X3D Era

So either I'm stocked with AMD (I'm not a groupie, I have nothing against Intel as long as it is good, I loved my EVGA GTX 1080Ti) or if I consider going for a mixed system, my next CPU might need 110.83% more power plus cooling...

So anyone telling me for certain that they know that I won't need more than 850 W for my future upgrade (PCI_E Gen5, stronger GPU than my RX 7900 XTX, next generation of RAM new die with lower latency etc) is just lying.

Next generation of GPUs and PCUs are gonna need more power, not less, that's the way manufacturers are increasing performances and it comes with increased heat radiation so more cooling is also needed, right now I'm just able to keep my system at <> 10°C below thermal limits (90/90°C) for both GPU and CPU in average with picks at - 5°C and I got 6 case fans among which two high airflow Noctuas.

In short, better more than too little, I experienced that first hand when I purchased my 750, having to swap it for the 850 (new) after only two years, that's something only experience can tell you, you just cannot tell what your system need for power will be tomorrow once all the technologies you're waiting for to be marketed are mature.

 

  

On 6/27/2023 at 3:18 PM, kksnowbear said:

Well...it depends.  You'd want to know which 4090, as even different cards from the same manufacturer can be rated differently.

For example, MSI recommends a 1000W power supply for their Suprim Liquid model, stating the minimum required is 850W, while the recommended PSU for a Gaming Trio 4090 is 850W.  So that's 150W difference in the recommended PSU from the manufacturer, both cards being 4090s.

Recommended PSU 1000W (Min. 850W)

That's what I mean. If people want to go by the minimum and ignore the recommended wattage, they risk to find themselves in the position where I was when I upgraded to the 5600X/1080Ti.

 

Edited June 29, 2023 by Thinder

[EightyDuce]

Just got the UPS hooked up and running the same Cyberpunk 2077 Overdrive benchmark it was pretty much locked at 486w output (similar to what the smart plug reported) with one momentary peak to 513W. It appears Kill-o-Watt is the outlier.

 

[kksnowbear]

2 hours ago, EightyDuce said:

Just got the UPS hooked up and running the same Cyberpunk 2077 Overdrive benchmark it was pretty much locked at 486w output (similar to what the smart plug reported) with one momentary peak to 513W. It appears Kill-o-Watt is the outlier.

 

Out of curiosity, what model is the Kill-A-Watt unit you used originally?  (And brand, if it's not actually a Kill-A-Watt brand, there are tons of knock-offs.)

Also, could you tell me the length of time that the CyberPunk benchmark takes to run (is it always the same amount of time)?

Edited June 30, 2023 by kksnowbear

[EightyDuce]

1 hour ago, kksnowbear said:

Out of curiosity, what model is the Kill-A-Watt unit you used originally?  (And brand, if it's not actually a Kill-A-Watt brand, there are tons of knock-offs.)

Also, could you tell me the length of time that the CyberPunk benchmark takes to run (is it always the same amount of time)?

 

Its a P3 Kill-A-Watt. Can't remember where I bought it, probably Newegg or Amazon, but it's been close to 15 years. 

 

The benchmark is a Cyberpunk built in benchmark that runs same amount of time for each run, couldn't tell you the length right now as I'm at work. Off top of my head its like 1-2 min long. 

Edited June 30, 2023 by EightyDuce

[usmave]

is it worth to upgrade from a 5800x3d, 32gb 3600 cl 16 (2x16gb), 4090 to

 

a) 7800x3d, 64 gb cl 6000

b) just upgrade to 64gb cl 18 (2x32gb)

 

playing in vr only, pimax crystal is coming soon.