Search results

For the afternoon, this thread would have been absolutely dead if it wasn't for erek's posts...

The 3d cache really helps on open world games. As for core counts, if you already happen to have a 5800X3D, you can run a game on six of the cores and see how it does. If you don't notice a difference, the 5600X3D should run it as well as the 5800X3D as the clock speeds are nearly identical...

Pretty much agree only I'm not sure if two plugs are enough without load balancing.

Or it means Nvidia cheaped out and left out load balancers. Or it means someone tried driving 600W through 300W of connectors -- the PCIe 8 pin is rated for 150W. You need four of them to drive a 5090, two is insufficient. Anybody discussing this issue without addressing the other design...

Can you imagine the clown show that would happen if Nvidia tried to get the 5090 UL rated?

The failure that is the 12VHPWR connector compared to previous in one picture (that was posted earlier): Note the condition of the four old-school PCIe 8 pin connectors. They were subjected to the exact same load that destroyed that 12VHPWR connector and yet they look unscathed. That picture...

Good to see it has soft power. If the WireView is good enough, I would use a tight tolerance on Current Balance as a leading indicator. The precise current balance will vary greatly from system to system. However, on any individual system, it should not, as the wire lengths are not going to...

The 4090 also only has two load balancers, previous cards had 3. The 5090 is really bad, as it only has one and the highest load. What else besides too much power can possibly melt the connector? Think about what causes this issue in the first place. Take into account that most 5090...

Because tribute must be paid...think of the poor federal and local congresskritters. Won't anyone think of them!

^^ This. Ebay is not what it once was, but these clicks filter out a lot of riff-raff. You can still score on old electronics. I picked up several Titan Xps for $131 each, but I had to troll for 3 months to get it. Could have bought them at $180 all day and all night. Haven't sold on Ebay...

Does the WireView provide a software method to automatically shutdown the board?

erek, I'm not sure what you are trying to get here. I have criticized the need for these "do-dads" but that is leveled at Nvidia. There is no reason to think this device will not work to solve the melting issue. I don't have one of these, nor one of the affected cards, but from what I've...

That's actually true for most power systems. High end 4000 and 5000 series video cards? Not so much.

I think so depending on exactly what you mean. Another way of stating my point of view is I want three independent current shunts feeding the VRMS that cannot crossfeed between each other. Each shunt provides a portion of the power required to operate the card. The approach worked swimmingly...

That's how I read it too. Frankly, I'm surprised Nvidia left the loophole in there. IIRC, Nvidia locked down the BIOS because of the BIOS edits that were done on Maxwell to crank up the power. Then Nvidia locked down the bios on Pascal and shunt modding became the norm.

Current balancing, load balancing - tomato, tomahto. They are the same thing, and no, my opinion has not changed.

Looks like AMD has brought back the 5600X3D as well as the 5800X3D. Hard to judge the hotness, but to someone who likes AM4 and the 5800X3D, $180 for one of these sounds like a deal for those with AM4 motherboards and DDR4 laying around. MC 5600 X3D $179.99

Just checked stock at the local Microcenter and a surprise, they have 5600X3Ds in stock too. I didn't realize they brought that one back too. $180 for one of those does not seem like a bad deal as it's nearly 1/2 off the 5800X3D and very close in performance.

Oh wow, that isn't very smart. In a lot of cases I'm thinking that's like giving a toddler a loaded gun. Unlike shunt modding, which can likely be detected, this doesn't alter the card at all. So some user cranks their board up to 800W, the connector burns up, and they RMA it with "I was just...

I thought the BIOS was locked down by Nvidia. They have done this since Pascal. Did someone break the encryption?

And poof! With one little mouse click, life is better.

I doubt a 600W transient spike would do it as this sort of damage requires time to accumulate. Also the card's power supply would limit it before it got there. The TDP rating is a continuous rating, so transient spiking above the average (330W for the 5080) is already baked into the cake...

I still would just keep an eye on it. The power feed to the card is dynamic, it will change with load. Also, watch the balance - did the other pins go up as well? Watch all six, not just one. If a pin really goes south, it will be obvious.

I agree about Windows 7. On crying shame is Win 10 LTSB 1507 felt damned good. Lightweight and you could actually turn of the telemetry with user controls. It was sandbagged by lack of Pascal support. Also programmed by real programmers with no vibe coding crap and much less UI...

IMHO, don't fix it. I don't think that looks bad at all and if those are contact temps, they are actually similar to what I measured on the 6 power contacts of the old school PCIe 8 and PCIe 6 connectors on my Titan Xp loaded to ~230W, the max one was about 43C. The 12VHPWR does have a higher...

I'm not sure what this means because I've never measured wire/contact temperatures on even the old connectors. However, I've been running a test overnight on my pair of Titan XPs. They are out on the bench but have been operating at nearly max power (230W) for 20 hours. I stuck a thermocouple...

Agree. I think some if it could be "single rail" power supplies that can drive all their output down a single wire. If that wire somehow gets shorted to ground, the wire will burn up before the power supply overcurrent protection can save it. A minor quibble, but the terms current balancing...

I don't know for sure if is the beefier PCIe 8 connector and don't believe I made that claim. In fact, when commenting on that quad PCIe 8 adapter, I mentioned that I wasn't sure it was enough. Me too. As you can tell I like discussing this stuff and it has been a nice detraction from my...

More to toss in the mystery. It looks like the 5080 has two shunts and is more like the 4090 which Buildzoid reported also has two, I initially assumed it also had one like the 5090. This suggest that two shunts are not enough as 5080s and 4090s have burned these connectors. Makes the quest...

How do you know those connectors, by themselves are good enough? The littlegreenman posted several pictures of burned up PCIe 8 connectors. All the 1080TIs had current balancing and I suspect the practice has gone back a long way. I think all the 3090 TIs had current balancing. The 3090 TI...

How about yet another tack? Think of each power wire individually. They are all rated at 8.33A. If you don't want them to burn up, you would fuse them at 8.33A. Since blowing a fuse is inconvenient, we'll let the power supply monitor the current in each of the six legs and if any of them...

How do you propose to solve the hot wire problem? You may not like the current balancing solution, but it has worked in the past on Nvidia video cards. I have thought about what I'm claiming is the problem and I've backed up every claim I've made. Several claims by others have been backed up...

Then just leave the load carried by each in amps. But realize the power dissipation in the wire will be those amps times the resistance of the wire. Power dissipation in the wire will raise its temperature. Wire itself does tend to be uniform and there is much more variation at the contact...

It sounded like the guy(s) doing that quad PCIe 8 adapter were talking about putting some sort of shunting in there, that should help but may be tricky as there is still only a single 12V feed (shunt) into the card. The reason this works better on the PCB is you have a multi-channel step down...

Connector resistance isn't the only thing and it appears you have tunnel vision about it. Didn't mean to offend, don't have a dog in this hunt, and the results posted by der8auer and Buildzoid are 100% conclusive. I have no motivation for further research. So much for the hope...moving on.

Why not? What specification does it violate? All I'm looking at is that nice 9A on each connector pin. Keep in mind, for this current balancing scheme to work the range of resistances and spread of resistances all the way back to the ATX 12V power supply have to be reasonable, this is real...

Go beyond the connector and look at the resistance from the GPU input all the way back to the ATX power supply output. Again, it's not just the connector. If one of my 12V lines is twice as long as another, it will have 2x the resistance. Not likely, but same concept. How well matched are...

You don't care about trying to shove 24A though a contact rated for 8.33A? Your own spreadsheet confirms what I've been saying, I circled the relevant parts in the red and green boxes.

I think we are seeing a bit of that today :) .

Not 100% sure how you are doing the calcs, but they look close. I don't know why the 8A line is only 6.94A in your non-balanced case. You realize your confirming the argument I've made all along? I said I don't care about the wasted power, I said I wanted the connector to be within spec...