• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

So, who's getting a 5XXX series card?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Nvidia never did crossfire.
This is what happens when people talk AMD GPUs in an Nvidia thread! Reaper in this case seems to be a 9070 XT version from PowerColor.
 
This is what happens when people talk AMD GPUs in an Nvidia thread! Reaper in this case seems to be a 9070 XT version from PowerColor.
Yeah I was trying to get the agent to get that AMD crap out of this thread. 😅
 
These cards look pretty basic and boring. They are nowhere close to the last-gen Red Devil or even Hellhound (even though these were mediocre quality, they at least look better).
Some other brands were showing AMD cards, too, like ASRock or Gigabyte. I just see them as worse quality in general than Nvidia cards. I don't know why as they could make it equal. I mean, check ASUS or Gigabyte. Nvidia coolers are always of better quality, have better finished shrouds, have more heat pipes, or have better fans.
 
Last edited:
Nvidia coolers are always of better quality, have better finished shrouds, have more heat pipes, or have better fans.
I've not looked at this, but I would speculate one reason. If the coolers can't be made universal, and have to be dedicated separately between red and green, there may be costs involved where higher volume NV sales allow them to do better for a given cost.
 
I've not looked at this, but I would speculate one reason. If the coolers can't be made universal, and have to be dedicated separately between red and green, there may be costs involved where higher volume NV sales allow them to do better for a given cost.

It's the same with AMD vs Intel motherboards. Intel has more enthusiast/overclocker oriented models and is generally better designed. AMD is still considered a lower/budget or gamer series for most manufacturers. I don't feel it's right, but this is what I see testing and comparing various products.

I highly doubt that the cooler's cost is so different and affects the total price of $500+ cards so much. It's also not only the cooler itself but fan curves, RGB illumination (like it or not), additional software ... everything is somehow better prepared on Nvidia cards.
 
I was just thinking since my board could run dual gpu I thought hey crossfire, but it looks as dead as my great uncle.

So that was my brief interest in team red ha.

No offence meant of course..
 
For some reason, Crossfire is still mentioned on some motherboards, but officially, it's as dead as SLI. I assume you can make any AMD cards run in Crossfire, as long as drivers support it and you have 2x PCIe x16 on your mobo (at least 8+4). I don't know if drivers for the latest cards enable Crossfire. For sure, SLI was disabled some time ago.
I guess all those CES PCs with 2x RTX5090 were presented more as AI builds. Nowadays, everything is AI, even if it is based on older technology and has nothing to do with AI.
 
Yeah any multi gpu system these days is typically created for high performance compute: LLM/NN, rendering, decode/encode, the odd people that do F@H, etc.
 
With maximum wattage at 675 and the base card already using 575w. Is there a point to nicer 5090s? Doesn't seem like a lot of overhead for clock boost or overclocking. The cheapest might be the best option. Hello power color and zotac, lol.
Post magically merged:

Yeah any multi gpu system these days is typically created for high performance compute: LLM/NN, rendering, decode/encode, the odd people that do F@H, etc.

We are not odd. lol
 
Exactly why I'm shocked they didn't do double HP12V connectors. It seems more like a safety issue to me. I'm hoping that the OEMs release double HP12V cards for the 5090 late this year or next year. But who knows, Nvidia has been restrictive with their power designs for awhile now. I think the 5080 and lower will do fine with power boosting but the 5090 may have more restrictions. Not sure if Nvidia is going to try and prevent the 5090 competing with their other dedicated AI products by binding the max core freq or some such. I'm still very tempted in tearing my first one apart and trying to power mod it to as close to 1KW as possible. I just know I'll have to work in another HP12V connector somehow or try and source a higher powered variant.
 
I mean... you can get 675W of in-spec power. That's 100W of headroom...or around 17%. NV has been limiting card power for ages now. Typically, they (card partners) only let you increase the power limit several % and not often into double digits. The only reason you'd need more than what a single connector offers is if you're modifying them to go past that limit. Otherwise, they are safe and everything in-spec up to 675W.

You may get lucky and a partner comes out with one.... but I wouldn't hold my breath....and know MSI isn't considering it.
 
I wouldn't be surprised if Nvidia puts a bigger hold on the power stage in this generation.

First off we do need to know what the avg pull is during gaming and see if it comes close to the limits. Maybe the limit is only for certain types of workloads. It'll be very interesting with these cards. My bet is that we can assume that the GB202-300 ASICs are a tuned down version of their data center grade ASICs. There is some information out there that shows the data center versions use 700W to 1200W per GPU. So its just a matter of what is allowed, what can be hacked, and what is a realistic power usage of the GB202 for gaming.

If anyone were to come out with a dual HP12V variant it would have been EVGA. My bet is on the Asus Matrix having this capability if Nvidia allows it.
 
Well if you look carefully you will see that the industry has made some changes because of the 12vhpwr issues.
If you remember, the ASUS 4090 Matrix was the first (and only) card to monitor the current on the individual pins of the 12vhpwr, and they made this a standard feature for the 50 series.
W1i6Eoo.png

In addition ZOTAC added a 12vhpwr "safety light" for the 50 series, which is meant to fullfil a similar purpose.

Finally CableMod announced a modification to their cables based on the new NVIDIA spec, probably not something major, but its the first revised cable (thus far, only the connector on the card was adjusted).

I know me and EarthDog are in disagreement over this but the way I see it every 4090 is a ticking time bomb, especially those with the H+ connector.

Direct 12vhpwr PSU<>GPU cables scare the hell out of me, they really do. I prefer Corsair's approach where it's PCI-E on the PSU side and 12vhpwr on the GPU side.
 
I wouldn't be surprised if Nvidia puts a bigger hold on the power stage in this generation.

First off we do need to know what the avg pull is during gaming and see if it comes close to the limits. Maybe the limit is only for certain types of workloads. It'll be very interesting with these cards. My bet is that we can assume that the GB202-300 ASICs are a tuned down version of their data center grade ASICs. There is some information out there that shows the data center versions use 700W to 1200W per GPU. So its just a matter of what is allowed, what can be hacked, and what is a realistic power usage of the GB202 for gaming.

If anyone were to come out with a dual HP12V variant it would have been EVGA. My bet is on the Asus Matrix having this capability if Nvidia allows it.

EVGA is dead. Matrix is ridiculously expensive, and the last generation was highly limited. No other brand exceeds Nvidia's specifications anymore, just because it's against their warranty terms. At least no brand unlocks voltages and very high power limits. If you want that, you have to do the hard mods. RTX4000 was not "hacked" in any way. People had problems even doing anything with the firmware. Some models had some mods, but each of them required a different way to even flash firmware.
In short, double power connectors would be useless for a card with a power limit significantly below the maximum power design.

With maximum wattage at 675 and the base card already using 575w. Is there a point to nicer 5090s? Doesn't seem like a lot of overhead for clock boost or overclocking. The cheapest might be the best option. Hello power color and zotac, lol.

PowerColor doesn't sell Nvidia cards. TUL (the corporation that owns the PowerColor brand) has an agreement with AMD, and all its brands sell AMD cards.
Anyway, I wouldn't spend money on RTX5090 from most brands, because of various quality issues. The list is quite long like Palit, Gainward, PNY, Zotac, Inno3D, Biostar (I doubt they release top GPU models), and all those weird Chinese brands. I was positively surprised about quality of Colorful and I use their RTX4080 for 2 years. It has an excellent power design and cooler. RGB/OC software could be better.
I assume there will be basic and not overpriced models from ASUS, Gigabyte and MSI, and I would aim for that, probably even FE models as the last two generations were surprisingly good.

Finally CableMod announced a modification to their cables based on the new NVIDIA spec, probably not something major, but its the first revised cable (thus far, only the connector on the card was adjusted).

I know me and EarthDog are in disagreement over this but the way I see it every 4090 is a ticking time bomb, especially those with the H+ connector.

Direct 12vhpwr PSU<>GPU cables scare the hell out of me, they really do. I prefer Corsair's approach where it's PCI-E on the PSU side and 12vhpwr on the GPU side.

I wouldn't trust CableMod after they failed twice with connectors. They had the worst connectors initially, then corrected them, and they were still melting. Even cheap Chinese brands made it better. Literally, some brands like AsiaHorse (you can buy on Amazon) are much cheaper, look great, and have strengthened connectors, even with heatsinks.

I don't know what to think about direct PSU-GPU 12VHPWR cables. I had no problems with them, but you are right; they give one more spot that may fail. I have FSP and Deepcool 1000/1300/1350W PSUs that were working fine with RTX4080, but I had no RTX4090.
I have mixed feelings about Corsair 12VHPWR cables in ITX PSUs. I guess the same is in ATX as they share the "Type" among various models. I mean, they're hard to bend and feel a bit weird, but this is also good as it's not easy to bend them near the connector, which is the main reason for failure.
Every adapter adds another potential point of failure, so as long as all these 90° adapters look good, I don't trust them with high-wattage cards.
 
Last edited:
Another way to look at the power connector thing: Steam has a reported MAU of 132M as of some point in 2024. As of December the 4090 is 1.18% share - more than any AMD dGPU! Combine the two implies there's over 1.5M monthly active 4090's on systems with Steam. How many power connector problems are there? We might never know, but if it were a significant %, you'd expect far more reports. You're probably more likely to encounter any other warranty covered problem on a 4090 than the power connector.

Edit: December seems to be another glitch month for Steam Hardware Survey. It totals around 120%. They did correct a similar glitch previously present for October, where it had totalled around 130% but is now back to 100%. Manually allowing for this, let's call 4090 share near enough 1.0% which is more consistent with previous months.
 
Last edited:
EVGA is dead. Matrix is ridiculously expensive, and the last generation was highly limited. No other brand exceeds Nvidia's specifications anymore, just because it's against their warranty terms. At least no brand unlocks voltages and very high power limits. If you want that, you have to do the hard mods. RTX4000 was not "hacked" in any way. People had problems even doing anything with the firmware. Some models had some mods, but each of them required a different way to even flash firmware.
In short, double power connectors would be useless for a card with a power limit significantly below the maximum power design.



PowerColor doesn't sell Nvidia cards. TUL (the corporation that owns the PowerColor brand) has an agreement with AMD, and all its brands sell AMD cards.
Anyway, I wouldn't spend money on RTX5090 from most brands, because of various quality issues. The list is quite long like Palit, Gainward, PNY, Zotac, Inno3D, Biostar (I doubt they release top GPU models), and all those weird Chinese brands. I was positively surprised about quality of Colorful and I use their RTX4080 for 2 years. It has an excellent power design and cooler. RGB/OC software could be better.
I assume there will be basic and not overpriced models from ASUS, Gigabyte and MSI, and I would aim for that, probably even FE models as the last two generations were surprisingly good.



I wouldn't trust CableMod after they failed twice with connectors. They had the worst connectors initially, then corrected them, and they were still melting. Even cheap Chinese brands made it better. Literally, some brands like AsiaHorse (you can buy on Amazon) are much cheaper, look great, and have strengthened connectors, even with heatsinks.

I don't know what to think about direct PSU-GPU 12VHPWR cables. I had no problems with them, but you are right; they give one more spot that may fail. I have FSP and Deepcool 1000/1300/1350W PSUs that were working fine with RTX4080, but I had no RTX4090.
I have mixed feelings about Corsair 12VHPWR cables in ITX PSUs. I guess the same is in ATX as they share the "Type" among various models. I mean, they're hard to bend and feel a bit weird, but this is also good as it's not easy to bend them near the connector, which is the main reason for failure.
Every adapter adds another potential point of failure, so as long as all these 90° adapters look good, I don't trust them with high-wattage cards.
Yep this is exactly my point about 12vhpwr<>12vhpwr. When you have a single connector that carries 450-600W, and one side is bent or dislocated, you will run into issues. At least with the GPU you can visually inspect the cable, make sure it doesn't line up against the side panel, etc. When it's single 12vhpwr also on the PSU side, you never know what happens under that PSU shroud, or how the cable squashes when you close the (rear) side panel.
0FeYckU.png
Regarding the Corsair cable, I tried 3 different cables. The first one was the original non-braided Type-4 12vhpwr, and it was VERY difficult to bend and twist, almost impossible to route this cable. The second one was a braided Type-4 12vhpwr, much easier to manage but eventually I saw big voltage drops with it (~0.4V). The third one came with my HX1000i, and it was easy to manage and route even though visually it looked like the first cable.
I also tried a few CableMod cables. Neither of them melted. But the first one got hot to the point I saw glue oozing out of the (back) of the connector, on the outer layer of the cable. The second one was very difficult to route (it was 4x PCI-E to 12vhpwr, very thick). The third one after a few weeks developed a large voltage drop (0.45V). The fourth one was new in box and had a dislocated pin... Yeah, I'm not recommending those.
 
Back