*Official Benchmark Thread* - Post it here or it didn't happen :D

[Papusan]

 

Let'm burn. One by one. What brands doesn't matter. Will never come a proper fix before most of them melt to crispy crunch...

 

 

A Zotac RTX 5090 Solid completely burns out its connector!

 

 

 

 

 

 

[Mr. Fox]

4 hours ago, Papusan said:

 

Let'm burn. One by one. What brands doesn't matter.

 

A Zotac RTX 5090 Solid completely burns out its connector!

 

 

 

 

 

 

This may not mean much, but it seems like all of the melted connections I have seen have been on air-cooled 4090 and 5090 GPUs. I wonder if that is somehow contributing to the melting plastics. It may be mostly due to liquid cooling being less common. My liquid cooled cards have a thermal pad to help wick away heat from the connector to the backplate, and it definitely works. The backplate is always warmer around the  connector.

 

It also looks as if both of those examples above are using the octopus adapters that ship with the GPUs and not aftermarket cables/adapters.

[Papusan]

35 minutes ago, Mr. Fox said:

This may not mean much, but it seems like all of the melted connections I have seen have been on air-cooled 4090 and 5090 GPUs. I wonder if that is somehow contributing to the melting plastics. I may be mostly due to liquid cooling being less common. My liquid cooled cards have a thermal pad to help wick away heat from the connector to the backplate, and it definitely works. The backplate is always warmer around the  connector.

 

It also looks as if both of those examples above are using the octopus adapters that ship with the GPUs and not aftermarket cables/adapters.

 

8 out of 10 cards sold in the market is probably air cooled cards. And the average gamer Joe #if buying xx90's" prefer air cooled cards. 

 

And not sure even with help some won't connect the tiny trash correctly to the card. Even pros can fail the same task. And if you connect it properly that doesn't necessarily mean the current will be balanced between each of the 12V pins. And you may see if you succeded after a few months/a year. Use of clamp meter is needed if you want to be 100% sure everything is correctly connected. But I expect the average pc user will never test it this way. It's plug and play.

 

 

[electrosoft]

1 hour ago, Papusan said:

 

Let'm burn. One by one. What brands doesn't matter. Will never come a proper fix before most of them melt to crispy crunch...

 

 

A Zotac RTX 5090 Solid completely burns out its connector!

 

 

 

 

 

 

 

Myself personally, I am not a fan of any downward pressure/force on these cables especially the official adapter. Since Ampere, I have always run my cables up over the back suspended and supported by either another cable and/or a cable tie from the top of the case.

 

Second, the way they jut out really leaves no choice for many to put some bent stress on it depending on case depth. Even in my Phanteks Enthroo which is pretty massive with the cable connector run completely straight as possible before the suspended hold, it is protruding past the glass cover (or where it would be).

 

If you're running a down run on the cable, I would actually give it some slack and use some cable ties to remove any and all type of pressure off the cable from the bottom and give it support. That expansion and contraction of the connector along with that constant pull on it can't be good.

 

lastly, I would just sacrifice aesthetics and if that means having to run it with the panel off do it that way.

 

You can always run a right angle adapter that you trust, but even then I would still give the cable some slack and then use a zip tie or retention mechanism on the bottom of the case to remove all pull from the cable.

 

With the Wireview 2 pro warranty in place, I do plan on transitioning to a single run and potentially swapping in either my pre-existing MSI AI300p (AI = barf) or an Asrock 1600w PSU and then running the 800w vBIOS. 

 

1 hour ago, Mr. Fox said:

This may not mean much, but it seems like all of the melted connections I have seen have been on air-cooled 4090 and 5090 GPUs. I wonder if that is somehow contributing to the melting plastics. I may be mostly due to liquid cooling being less common. My liquid cooled cards have a thermal pad to help wick away heat from the connector to the backplate, and it definitely works. The backplate is always warmer around the  connector.

 

It also looks as if both of those examples above are using the octopus adapters that ship with the GPUs and not aftermarket cables/adapters.

 

1 hour ago, Papusan said:

 

8 out of 10 cards sold in the market is probably air cooled cards. And the average gamer Joe #if buying xx90's" prefer air cooled cards. 

 

It's probably a bit of three  things:

 

#1. Air cooled cards vastly outsell waterblocked cards (aka Joe Gamer) who plug n play and just go even twisting and forcing their connections if need to make them fit which is problematic in and unto itself.

 

#2. Those who block their cards are also cognizant of the connector issues and usually take a higher level of care and attention to their cards and cable(s).

 

#3. The nature of the block, coverage and cooling can provide better cooling to the connector assuming the imbalance of amps and subsequent heat is the issue and as a result the extra heat is being dissipated at a higher level than normal connections. It isn't as if the imbalanced amps are killing the connectors. It's the absolute meltdowns as a byproduct that are doing it.

 

So you have not only a much smaller audience who block but within that audience a higher degree of care and maintenance along with added cooling to the connector.

 

 

 

 

[Papusan]

On 11/23/2025 at 10:56 PM, Papusan said:

And what you think will happen once China again can get whatever they want from Nvidia and AMD? Component shortages will increase further and push the prices on pc hardware to new height we have never seen. China they wiill be delighted... 

 

China's huge big fleet of tech engineers will make wonders with the better western tech to boost their military technologies. Greed, flattery and stupidity come at a huge price. Some is so stupid that they can't see it themself. Sad they don't know who'm they themself are. Maybe look into the "golden"mirror above the golden sink. I'm sure you will see stupidity. Fooled by Jensen and Xi Jinping.

The United States wants to give NVIDIA back the ability to sell its GPUs to China.

 

RTX6000 aready tested with Geekbench. Come with 84GB GDDR7. Better use the scarce vram chips for consumer products than for AI workstations in China. But max profits beats common sense.

 

Nvidia with good help of the jokes in the US adm now targeting AI hardware demand in China. 

 

NVIDIA RTX 6000D made-for-China GPU: 84GB GDDR7, lower clocks, slower than RTX PRO 6000

 

 

And Samsung is finally ready to help Nvidia to feed the Chinese market with GPUs for AI workflow.
Samsung begins mass production of 28 Gbps GDDR7 3GB memory, faster variants now sampling

 

[Mr. Fox]

A few of you might know I was put on unpaid furlough on Friday, along with about 8 or 10 other people where I worked, including my daughter and the company owner's nephew that has worked there for 15 years.

 

I had my second job interview for an opportunity that looks really promising and I expect to be an even better job than what I had (which was great in every respect until Friday). I have never been unemployed before, but I am hoping to get an offer on this next week. So, please keep me in your thoughts and prayers.

 

I hope all of you in the US celebrate a great Thanksgiving with your loved ones. We all have so very much to be thankful for, even when there are days that it doesn't seem like it.

[Mr. Fox]

You know things are circling the drain when Steve is kicking Winduhz to the curb.

 

 

[Papusan]

3 hours ago, Mr. Fox said:

You know things are circling the drain when Steve is kicking Winduhz to the curb.

 

 

 

Railway to Hell....... Microsoft just can't stop making wrong choices. 

 

[electrosoft]

Sugi put up a nice post over on the OCN forums discussing the MSI connector and the differences in their PSU cable versus their adapter.

 

 

Basically the 4 headed dongle doesn't have complete coverage around all sides solid and is more of a crimp top connector suspect to pressure/bending and opening up over time.

 

This supports my argument about bending/pressure/pulling on the adapter making it susceptible to opening up a bit or changing over time with constant amps and heat/contraction AND force being applied. 

 

I'm a firm proponent of making sure your connection is devoid of any external, unneeded force or pull while it is violently trying to handle 600w even with a heavy duty made adapter, but the MSI yellow terror may be subject to a higher chance of suffering from structural fatigue due to its design.

 

Here is how I run mine:

 

 

There is zero pull or drag on the connector and any force is actually upward if any. 

 

I can't imagine *4* PCIe cables attached to the adapter then all that just being pulled down by gravity over time let alone bent and pinched to go with it would have long term viability.

 

We obviously know many other cables and connectors are suspect to meltdowns as we've seen them in the wild, but this is a first design choice that MAY lend the MSI adapter to have a higher failure rate especially in conditions where it isn't optimally paired and supported.

 

 

 

 

[Papusan]

 

Where is the QC ? Gigabyte also failed QC when they started to use thermal putty on their graphics cards.

 

Brand new 5070 TI, noticed this immediately, anyone have seen something like this before? I have not plugged it in, and do not intend to do so.

 

 

 

 

 

[Papusan]

On 11/25/2025 at 10:40 PM, Papusan said:

Nvidia with good help of the jokes in the US adm now targeting AI hardware demand in China. 

 

NVIDIA RTX 6000D made-for-China GPU: 84GB GDDR7, lower clocks, slower than RTX PRO 6000

 

 

And Samsung is finally ready to help Nvidia to feed the Chinese market with GPUs for AI workflow.

Samsung begins mass production of 28 Gbps GDDR7 3GB memory, faster variants now sampling

 

Yup. For AI GPUs to China. 

 

Nvidia has reportedly asked Samsung to double its production of GDDR7, possibly for AI GPUs to China—although we're hoping for some more VRAM-stacked RTX 50-series Super cards

 

Oh great, the latest rumours suggest Nvidia might stop bundling VRAM with its GPUs, because who cares about cheap graphics cards
 

[Raiderman]

Happy Thanksgiving everyone!

[Mr. Fox]

31 minutes ago, Raiderman said:

Happy Thanksgiving everyone!

Happy Thanksgiving, brother. And, to all the same. 

 

Life has its ups and downs, sometimes significant challenges, but there are always more reasons to give thanks than we can count. Blessings to all of you and your loved ones.

[chew]

Out of space for files.

 

Link instead 

 

[Mr. Fox]

5 hours ago, chew said:

Out of space for files.

 

Link instead 

 

Wow, nice. I am glad to see that board released. I am going to have to check out the owner's manual to see if they effed up the bifurcation on the PCIe slots like A$$zeus did with the X870E Apex abortion. Do you know if you can install anything X1 or X4 in the bottom PCIe slot without castrating the GPU slot to X8?

 

How do you connect the two 8-pin CPU power cables. Does that rear I/O have a decorative cover  that unsnaps? In the photos it looks like they are covered by the rear I/O.

[Mr. Fox]

https://hwbot.org/benchmarks/cinebench_-_r23_multi_core_with_benchmate/submissions/5931059

[Raiderman]

On 11/28/2025 at 4:16 AM, chew said:

Out of space for files.

 

Link instead 

 

Delete all of your old screenshots to free up space for the new. 

Nice results!

[Mr. Fox]

45 minutes ago, Raiderman said:

Delete all of your old screenshots to free up space for the new. 

Nice results!

Yes, very nice results. The downside to that motherboard is the unfortunate fact that it is single chipset (X870) and lacks the extra PCIe lanes of the dual chipset (X870E). It is sad that Gigabyte chose to not make the Tachyon an X870E part. It may as well been made as a B850 motherboard. The single chipset configuration basically ruins the capability to use any add-in cards without sacrificing GPU bandwidth.

 

It never ceases to amaze me how idiotic the motherboard engineers are at AMD, Intel and all of the AIB partners. They frequently implement the dumbest things imaginable. They are all guilty of it and none of them consistently demonstrate intelligent design. It almost feels like it is a deliberate nasty finger salute to the people that buy their products, but I suspect it is just ignorance more than anything else... maybe combined with a touch of apathy.

 

For X870E, the USB4 is a flagrant waste of resources that could have been put to better use. It is unfortunate that AMD was so shortsighted and mandated the waste of PCIe lanes for something hardly anyone I know that buys an enthusiast board give a rat's butt about. The second chipset allows potential for the motherboard manufacturers to allocate resources that can be put to good use. More often than not, however, they do not. They waste it on something stupid.

 

Deleting old screenshots is extra work. When I ran out of space for attachments I just started using embedded links from photos posted at overclock.net and hwbot.org and imgur. Much faster and easier for me.

[Papusan]

 

What a coincidence with the leak of the Astral 5090 XOC. Within a week after the Matrix bios got released for everyone, and that almost none of the owners of Asus cards could use. Hmm, I don't believe in Santa bro @Mr. Fox. I'm too old for that. What else could it be? A$us try avoid further bad publicity for the brands overpriced GPU's=? Hmmm. I'm sure I know the answer. Same with you😎

 

https://www.overclock.net/posts/29533017/

 

Edit. Jup, Asus had to do something....

Matrix BIOS can even be flashed to "HP OEM" lmao. Binning hundreds of Astrals to find one that clocks the highest in Valley, then ASUS ruins it.

https://www.overclock.net/posts/29533316/

 

[Mr. Fox]

13 hours ago, Mr. Fox said:

https://hwbot.org/benchmarks/cinebench_-_r23_multi_core_with_benchmate/submissions/5931059

Now that the CPU is sorted, I got the RAM dialed in stable as well.

Since I have more spare time on my hands temporarily, I decided to dismantle the Thermal Grizzly Mycro direct die block and clean the jet plate fins. It wasn't too terrible, just normal accumulation from the loop during roughly a year of use. I use two filters on each PC so that helps. I've had to clean the filters a couple of times.

I ordered a set of Thermal Grizzly TG GPU Guard gaskets and will put liquid metal on the 4090 and 5090 when I can muster enough motivation to do it. 

 

[Talon]

2 hours ago, Papusan said:

 

What a coincidence with the leak of the Astral 5090 XOC. Within a week after the Matrix bios got released for everyone, and that almost none of the owners of Asus cards could use. Hmm, I don't believe in Santa bro @Mr. Fox. I'm too old for that. What else could it be? A$us try avoid further bad publicity for the brands overpriced GPU's=? Hmmm. I'm sure I know the answer😎

 

https://www.overclock.net/posts/29533017/

 

Edit. Jup, Asus had to do something....

Matrix BIOS can even be flashed to "HP OEM" lmao. Binning hundreds of Astrals to find one that clocks the highest in Valley, then ASUS ruins it.

https://www.overclock.net/posts/29533316/

 

 

Very nice for your Astral LC card! Hilarious that magically Asus can release the XOC vBIOS after the Matrix made their entire 5090 line (except white Astral) complete garbage overnight lol. Good to see them realize most will never flash this vBIOS and it's only the very enthusiast community so risk is tiny. 

[Mr. Fox]

OK. New problem. My CyberPower 1650W UPS does NOT like this XOC vBIOS, LOL. SCREAMING BLOODY MURDER. I will be using the Matrix vBIOS one my "quiet" vBIOS position for normal daily driver usage and gaming. I launched Cyberpunk 2077 and all the sirens began blaring on my shunted 5090. I do not even need the EVC2 now because the XOC vBIOS gives me the voltage I was setting manually using the EVC2 LOL. I will test with benchmarks to confirm, but I may move the EVC2 over to the 4090 Suprim if I no longer need it on the 5090. Check the voltage and power now. Nice.

 

[Mr. Fox]

Process Lasso is definitely the way to go if you have a dual CCD processor with 3D V-Cache. No need for Micro$lop XBOX feces bloatware and the extra chipset driver crap. Just select the EXE and choose the affinity and BOOM, every time you launch that game it uses the CCD with the cache. When you close the game it is back to using all cores and threads. Easy, peasy... no hokey pokey nonsense with flakey rubbish from the Redmond Reprobates.

 

[electrosoft]

-----------------------------------------------------------------

 

6 hours ago, Papusan said:

 

What a coincidence with the leak of the Astral 5090 XOC. Within a week after the Matrix bios got released for everyone, and that almost none of the owners of Asus cards could use. Hmm, I don't believe in Santa bro @Mr. Fox. I'm too old for that. What else could it be? A$us try avoid further bad publicity for the brands overpriced GPU's=? Hmmm. I'm sure I know the answer😎

 

https://www.overclock.net/posts/29533017/

 

Edit. Jup, Asus had to do something....

Matrix BIOS can even be flashed to "HP OEM" lmao. Binning hundreds of Astrals to find one that clocks the highest in Valley, then ASUS ruins it.

https://www.overclock.net/posts/29533316/

 

 

For most of us, it is going to be the 800w Matrix vBIOS we run anyhow. It is superior with more granular control. I don't want my 5090 locked at an unnatural 1.15v at all times or trying to suck down the sun and go nova especially lacking the proper cooling for support which realistically is LN2 or at least a chiller.

 

800w Matrix vBIOS is going to end up as the perfect one.

 

Matrix vbios = Gigachad+ 800w

 

 

3 hours ago, Talon said:

 

Very nice for your Astral LC card! Hilarious that magically Asus can release the XOC vBIOS after the Matrix made their entire 5090 line (except white Astral) complete garbage overnight lol. Good to see them realize most will never flash this vBIOS and it's only the very enthusiast community so risk is tiny. 

 

This.

 

Do you plan on attempting to use it @Papusan since you can't flash the Matrix vbios?

 

2 hours ago, Mr. Fox said:

OK. New problem. My CyberPower 1650W UPS does NOT like this XOC vBIOS, LOL. SCREAMING BLOODY MURDER. I will be using the Matrix vBIOS one my "quiet" vBIOS position for normal daily driver usage and gaming. I launched Cyberpunk 2077 and all the sirens began blaring on my shunted 5090. I do not even need the EVC2 now because the XOC vBIOS gives me the voltage I was setting manually using the EVC2 LOL. I will test with benchmarks to confirm, but I may move the EVC2 over to the 4090 Suprim if I no longer need it on the 5090. Check the voltage and power now. Nice.

 

 

Yowza! Looking forward to some chilled results @Mr. Fox!

 

Crazy the Asus XOC bios just comes out full tilt like that and pretty much goes crazy. +4000 memory and basically all the W and V you probably need. You could always push past 1.15 w/ the EVC2 though....

 

--------

 

@Talon now that you have a Suprim 5090 with the Matrix goodness, how is it stacking up against your HOF 5090D in just basic air testing so far?

 

With the Asus XOC and Matrix 800w out in the wild, what are the plans now?

 

 

[Mr. Fox]

1 hour ago, electrosoft said:

Yowza! Looking forward to some chilled results @Mr. Fox!

 

Crazy the Asus XOC bios just comes out full tilt like that and pretty much goes crazy. +4000 memory and basically all the W and V you probably need. You could always push past 1.15 w/ the EVC2 though....

I am going to have to fire up the chiller tomorrow and play with the power slider to see if I can reign in some of that because pulling 1700W from the wall is not sustainable. It will trip circuit breakers (and I have this on a 20A circuit) and exhausts the capacity of my 1500W UPS. If that does not work will I will just go back to the Matrix vBIOS and use the EVC2. I'm not gonna lie though... I just LOVED seeing that 1.150V default. 

 

For now I just flipped my GPU over to using the stock Zotac 600W vBIOS. I think letting it run wild and free on the XOC vBIOS probably isn't a smart thing to do. It might be safe, but I don't want to just assume that and hope it turns out OK. Temping fate to such an extreme degree (1700W from the wall, probably 1300W from the GPU) with the petite arson PCIe power cable should put the fear of God in all of us that possess an ordinary amount of common sense.

 

Here is something interesting. I use a GPU overclocking program in Linux called LACT. It shows both core and hotspot temperature readings on the 5090. So, maybe the register is actually there and the NVIDIOTS gave Windoze software developers marching orders to not expose it. -?- I think they rank right up there with bad actors like crApple when it comes to being among the kings of command and control dictatorships.

 

I received my Thermal Grizzly GPU die gaskets for liquid metal, so I really need to make time for that. The package comes with 3 for 4090 and 3 for 5090, so I can do both GPUs. I am expecting to get an offer on the job I want most on Monday or Tuesday and I am sure that I am going to initially not have much discretionary time to do things until I get past my orientation, which usually take a bit more time in c-suite positions.