Live data from Hacker News

Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

igorslab.de

171–180 of 182 posts

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#171
post #10

Assuming you can install two GPUs using 600W, that's basically running a hair dryer in your desktop since most of the energy is converted to heat.

The RTX 3080 already uses 320W (peak?). By next GPU generation who knows. Good lord, just 2 generation back GPUs like the 1070Ti consumed 180W. Even if you add a CPU (of that era) like the i7-7700 with it 65W draw, that just 245W. Now a video card alone can draw more power.

I have a 2xRTX 3090 rig under desk. At close to full GPU load the temperature on the desk is about 3F higher than in the room.

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#172

Earlier quoted context omitted.

It's not, because if you plugged a device needing 600W using an MPX connector, it wouldn't work at all, since MPX can only do 500W. Here's the thing. When you have a lot of amps, they create noise in wires next to them. They'll also need thick wires and will get hot. When you have a lot of volts, they'll go through shielding and start a fire. Do you really want to place something with a lot of amps and volts next to…

Given the RF noise and the heavy-duty electrical engineering, wouldn't it be better to isolate the GPU in its own external enclosure?

the RF noise form computer components is enough to be taken care of by regular metallic shielding, because the amps are low for actual compute operations. an inch of space on its own is an effective barrier. this is very different than having a wire running enough amps to power a space heater, located next to your data transmission lines. even having the GPU in its own external enclosure, you would not want the data cable to be bundled with the power cable for anything that pulls over half a kilowatt.

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#173

Earlier quoted context omitted.

>You really won't run your rig at 500W all the time either. While this is true, I don't agree with your statement about cooling. It's very possible to have sustained 500W+ draw while playing a modern game on a 3080/3090. Sometimes you might play games for a few hours with a fairly consistent 500/600W power draw - honestly I think it's cheap and easy to cool this kind of load. A few 140mm fans and a mid-level CPU cool…

> A few 140mm fans Yeah... Nearly nobody has a few 140mm fans on their computers. Most people won't have any fan lager than 120mm. Powerful computers usually have a pair of them. Maybe 3 if you count the PSU.

People who are using 3080/3090's are much more likely to have systems with larger numbers of fans. And they usually have far more than two.

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#174
post #99
post #89

Earlier quoted context omitted.

It's funny because here in europe, power plugs with 16A (some are 20A) are actually the norm: at 230V they can deliver 3600W/4600W !

I was gonna say... I didn't even know overloading a power socket was a problem people had. Sometimes we'll run 5 gaming PCs + 5 monitors from one socket and it's always worked fine.

I've melted extension cords at lan parties before... There is always going to be a point where the # of electrical contrivances trips a certain threshold and starts causing problems.

If you've ever been to dreamhack or one of those gigantic lan gaming conventions, you would probably have noticed the massive power distribution transformers staged every 300 feet or so on the floors.

I honestly don't know how they would manage power delivery if every single person was running a ~1kW gaming load.

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#175
post #89

Earlier quoted context omitted.

It's funny because here in europe, power plugs with 16A (some are 20A) are actually the norm: at 230V they can deliver 3600W/4600W !

I wonder how hard it is to do a bottom-up conversion of the USA to 230. If the "all-electric" branch of the environmental movement are right, we'll eventually need to adopt the superior European kettle technology. Getting everyone on the same plug would be great.

Imagine how many transformers you are talking about replacing. You'd also have to redo everyone's electrical panel.

It is substantially easier to run additional customer connections for those edge cases where an existing 200A residential service will not suffice.

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#176

Earlier quoted context omitted.

excessive conspicuous consumption Energy is a proxy for accomplishment and development. Humanity's #1 goal should be to maximize energy use per capita while minimizing energy use per activity, such that every human is doing as many desirable activities as possible. Gaming at 1kW for an hour costs 30 cents or less in electricity. As others have mentioned indirectly, going for a Sunday drive to see the changing leaves…

>minimizing energy use per activity Maybe this is what feels off? A kilowatt feels like too much for a videogame. This isn't a bitcoin mining thing, is it?

my rtx 3070 uses like 70-80% on average when I play new world. my cpu (5600x) is very low (~25%), but I probably blow like ~250 watt or more per hour and I do not have a expensive setup (no rgb, etc.)

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#178

Earlier quoted context omitted.

Definitely agree that a kilowatt seems very high for gaming and starts to have an environmental impact. Still, driving a car at 100kph takes around 20 kW of power, so driving somewhere is still a bigger burden on the environment than gaming.

That's peak power. You don't run at 100kph straight for hours but you will run your gaming rig at say 500W for three hours in a day (particularly for people willing to shell out money for such a gaming rig). The total energy per day might not be that different, you'd have to account for how much is involved in commuting in traffic.

> That's peak power. You don't run at 100kph straight for hours

Have you been to the US? I regularly drive 4+ hours at 125kmh+ anytime I'm trying to get to the Sierra, or Utah or Arizona, which is often.

In fact, almost every weekend I drive at least 3 hours round trip, half freeway half mountain roads, if I go to the closest good climbing. 5+ if I go to my preferred locale. The distances people regularly cover are enormous.

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#179

Earlier quoted context omitted.

There are literally millions of kilometers of highway all over the world with a speed limit of 100km/h. I think the total number of man-hours spent at that speed or higher globally would be pretty shocking. My commute to and from the office alone would be about 90 minutes at that speed.

Do you live 150km away from your office? A super commuter commutes >50 miles one way, that's a little over half of 90 miles (which is 150km). Again, the point is what matters isn't peak power or rate of any quantity but the integral, the total. Most of the time commuting is spent moving much slower than 60mph.

No, I'm about 60km one direction - I live rural and work in a nearby city. Each direction is still only 40 minutes or so, so it's not as onerous as it might sound.

Re: Up to 600 watts of power for graphics cards with the new PCIe 5.0 powerconnector

#180
post #147

This is insane. Give me back the 75W GPUs instead. No extra power connector, just the slot.

It's been a few years since the release of the GTX 1650 but it's still a pretty serviceable card. Cheap, too.

That's the point. Where is the 3050?
Post reply on HN