Nvidia RTX 4090's Power Draw May Be Too Much For Its Power Connector To Handle

Slashdot is powered by your submissions, so send in your scoop

The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.

Gamers Nexus covered this. [youtu.be]

So did JayzTwoCents. [youtu.be] Good information on how it might be the bend radius of the new 600Watt cables that are causing insufficient pin contact for the amount of current each pin is supplying.

Why on earth we insist and putting the connector in a position where such a bend is required is beyond me. GPUs are big enough that in many cases there's no option other than breaching this bend radius. It would make more sense to have the connector come with right/angle connectors or be in a different location on the card (such as the back, which used to be a problem when our PC cases were full of HDD and GPUs were insanely long, but they've gotten shorter and fatter).

Because on the back it would run up against another big thing in computer cases. The CPU cooler which can be pretty big. Memory and M.2 SSDs are also in the general area with their own cooling issues.

"Back" was ambiguous. I was talking about the side opposite to the display connectors, the end that points to the front of the PC. Given the lack of 3.5" bays on modern PCs and the fact that GPUs stopped being insanely long after about the 10x0 series back half a decade ago so we should have ample space on that side.

Several of the cards terminate before reaching the edge of the heatsink/fan assembly on that side of the card. Nvidia still put the connector there for many of their rtx2000 cards, but this complicated disassembly with extension cables and glue to ensure they didn't come unplugged.

The standards for graphic card mounting are so old and outdated that there are problems no matter how things are manipulated. I'm not sure I buy the critiques about bends near the connector either, as it appears that half of the

I have a ninety degree board so the "back" points directly at the case fans. The NVIDIA solution, if I had one would point at the side-panel.

The new connector wasn't about GPU power requirements, it was about PSU signaling what kind of power it can make available and what kind of cable is connected so that the GPU knows what power budget it has for boosting.

Your welding plug (I know you used is as a joke but I feel the need to point out a flaw not related to the joke about size) doesn't have sense pins or other communication so it doesn't fit the bill.

"That the way it's been done for billions of years." (Slides a $9 billion brand new Gerald R. Ford-class aircraft carrier into a rowboat berth.)

These things already draw 600W peak, TDPs these days are for a âregularâ(TM) workout, crunching numbers will put a significant strain above the designed spec. Just a light overclock and they are drawing 150W above TDP already, combine that with a very thin connector that is often poorly installed or modified for âprettyâ(TM) cable management and water cooling, these things will start smoking a lot more often.

We have some high power cards with passive cooling in datacenters and workstatio

Change it to Anderson SB50. That should be good enough.

Or redesign the card to take higher voltage, 12V isn't really suitable for those powers. 400V 3-phase directly to the card should be good for the forseeable future.

These are regular connectors in a 16 pin housing. Each pin can take 8A without problem. That means the connector can take 768W if everything is perfectly balanced. Call it around 600W real-world. This card only pulls 450W regular. Of course if you plug this connector or the PSU-side in wrong, contaminate it or the pins or do stupid things to your card that causes power-spikes, this connector may well melt. But that is on the user, not the design.

Indeed. Unless you do stupid crap like plugging in only one of the connectors on the other side. Or plugging them in halfway.

There is a difference between a current rating and continuous current rating within an enclosure. The pins rated for 8A will be fine at 8A so long as they are not adjacent other pins and have access to airflow. Group all 16 pins together and place them within a plastic shroud and bad things can happen.

Derating the current capacity of a conductor is often required when wiring in order to conform to electrical codes. Are you placing the cables within a conduit? Are the cables bundled with other cables?

I should also point out that the main form of cooling for these connectors will be the PCB and connected power wires. Heat from the connector will go up the power lines and dissipate to the environment - basically the wire acts as a heatsink. So if you tidy the power lines and perhaps group them together within a cable sheath, you are going to get less cooling.

At the end of the day there will have been many factors at play. The problem is that the design does not allow for these factors.

Just looked at the datasheet for Nanofit connectors - they are a similar Molex connector. Anyway, they are also rated for 8A per pin. But there is a catch. Once you load all 16 pins the rating drops to 5.5A for 20 AWG wire. And this is based on a temperature of 30 degC - you have to drop the current even more with higher temperatures. So I am not making stuff up, you really do have to reduce the current rating of connectors. This is true for all connectors that have multiple pins grouped together.

Ok, just did a little more searching. I assume these new connectors are based on MicroFit3 connectors. At least it looks like a MicroFit3 housing would fit. So the current rating depends on the guage wire you are using. The maximum wire size is 18 AWG which allows for 8.5A nominal. But you have to derate and when using 16 circuits it derates to 5.25A. But now you have to adjust this value for ambient temperature as 5.25A is only valid for 30 deg C.

Seriously, there is a section of the product specif

Well, I applaud you actually trying to find things out. These are Molex Mini-Fit Jr. though. And it seems Molex does not have the current derating info online.

The point is however that derating should already have been taken into account and a lot more than just these connectors would blow up if it was not. At least Intel thinks these are good for 9.2A per contact in the stated configuration: https://www.cybenetics.com/att... [cybenetics.com] (page 52ff) with a max 30C heat up.

So the real problem here is not any derating fro

Uh, that nonsense speculation is bog-standard electrical shit any half-assed lineman would know.

The idea is not in dispute. The claim that this may be the cause for the overload is dispute. Do you really think the PC industry has been using these for 20 years and does not know how much current they can take in a double-row configuration in standard conditions inside a PC?

The _actual_ problem is loading these connectors up to the limit. They will still be fine in many situations, but even slight issues, like mechanical stress may then break them. And since there is more than one of these cards out ther

These pins are rated for 8A when installed into these connectors double-row with others adjacent and no airflow, just a maximum ambient temperature. Seriously. Using them singularly, they probably would be rated higher but that is not the way this type of pins get specified.

Too many connectors are just holding those currents momentarily, not continuously. Assuming that you can load any connector wirh the max rating continuously is the most common mistake in the electrical world. Heat also weakens the contact pressure beteeen pin and socket, which accelerates the issue by increasing the heat and weakening the spring tension between pin and socket permanently. Multi-pin connectors also have issues with more heat developing on the center pins. So a connector rated for 50A shall not

These are rated for 8A continuous. These are the same inserts that are used for CPU power, PCI-E power and main power connectors.

Just looked it up, actually it is 9A continuous. And these are the ratings when installed in the housings, which means when in two rows with neighbors. Seriously.

I have seen way too many cases where that rating still is done under ideal circumstances and not real world circumstances.

It's the same rating as the lifetime of LED lamps that are like 3 years, but only if you don't use them. So I'm serious that the figures are inflated and that it's not a good design.

Ask yourself who did the rating? A certified rater like UL or some other independent agency or the manufacturer?

Oh, sure. You should put in ample reserves. Especially when things get installed by non-experts. But if these are handled correctly and are undamaged, 9A is not a problem in rows of two. They do not even get warm. As soon as any problem is in the mix, not so much. But this has not really any connection to whether it is a single contact or multiple ones in a socket. A single contact under full load, when damaged or subjected to too much side-force (as seems to have happened here) will still heat up massively

I have seen way too many cases where that rating still is done under ideal circumstances and not real world circumstances.

Ratings don't tell you anything about the circumstances. The ratings define a temperature rise. In combination with the thermal limits of the connector and cables it's up to engineers to choose if they are fit for purpose.

In the case of Amphenol CEM-5 connectors like this one, 9.5A gives a 30degree temperature rise for a part with a maximum thermal limit of 105C. A 600W derating is applied for safety. While PVC insulation melts lower than this the pictures clearly show the connector overheating.

Ambient / real world thermal or power conditions aren't what killed this connector. Poor connection or poor assembly did.

Well, yes. But that was made possible by poor safety margins. In a user-installed connector, you should always have ample safety margins or some will experience the observed effects. The bad press that causes is not offset by the fault actually being on user side.

According to some people that been in the business like that jay two cents, the problem is most likely how easy it is to bend the pins when trying to install the card. The connector requires a clearance of 35mm from the first bend, and that is very easy to get wrong. And as bad contacts heat up, you get fire

According to some people that been in the business like that jay two cents, the problem is most likely how easy it is to bend the pins when trying to install the card.

That would do it. These connectors have two contact points in the spring side. If you bend a pin, you get only one good contact point or maybe even none. Do that to several contacts and it heats up too much.

The connector requires a clearance of 35mm from the first bend, and that is very easy to get wrong.

This may be a requirement the adapter manufacturer has put in there, but the connectors only specify some maximum mechanical load as far as I recall (did read several datasheets for them a few months back because I was doing a custom connector for). But yes, too much mechanical load on the contacts can also get you bad contact and heating up and these 35mm may be what the wires used give you for the maximum force. For example, if somebody bends the whole wire bundle close to the connector by 180 degrees, that may well do it.

I think the bottom line is that loading a connector of this type up to close to its maximum load is something you can do if only experts and careful people handle it. Throw in a few fat-finger types and you get the observed effects.

It doesn't help that PSU manufacturers don't supply right angle connectors to reduce the strain.

I imagine the manufacturers are the ones most concerned about this. If the connector melts they are the ones who will be getting the PSU RMAed, and possibly the graphics card too if that gets damaged. Apparently even at the RRP they can't make much money on a 4090, and some may be losing money on what is considered a halo product.

Indeed. I guess manufacturers should stop to make these "bleeding edge" stuff. It just does not make much sense. Or maybe the PC buyers are too irrational and will buy something smaller only of the manufacturer also makes the larger thing.

These are regular connectors in a 16 pin housing.

It's a 16pin connector. However, 4 of them are for sideband. That leaves 12 pins. The spec indicates 6 pins at 9.2A for a total of 55.2A and 662.4W. As to what is happening, so far testing indicates it's a matter of bending the wires too close to the connector. It's causing stress, and imperfect contact, within the connector.

Oops, you are right. So they already are overloading the individual pins by a bit. These should be the same pins and springs as the regular Molex Mini-Fit Jr. and these days they are rated for 9A. (Used to be 8A? May also be my memory playing tricks on me.) If you put too much force (not a lot) from the side on them, contact will be imperfect. This is not a high-quality connector, just a regular one, with a pin and female connector costing just something like 10 cent if you buy then individually. A Molex Me

Unlikely. These would melt a lot. No, even the original Molex part is at its limits here.

These are regular connectors in a 16 pin housing.

No. 12 power pins, 4 side-band (not current carrying).

Each pin can take 8A without problem.

That means the connector can take 768W if everything is perfectly balanced.

It should be noted that the IEC only provides a safety rating of 5A. This would be a worst case scenario - so all circuits loaded and in a warm environment. Looks like we have a worst case scenario here.

With a 5A rating we only have 30A total (there are only 12 pins that carry power), or 360W assuming the power supply maintains 12V output. Below the 450W power draw of the card. Good luck getting IEC certification....

I speculate that we will eventually have a recall on these boards due to being a fire hazard. Over time, the resistance of the connector will increase due to corrosion. Looks like the card and power supplies use tinned pins - (should have used gold). As the resistance increases the heat will build up and eventually things will melt / combust. Even if this only happens on a small number of boards, it still warrants a recall. And considering that it has already happened and that it is going to get worse with time - expect either a recall or a new firmware that castrates the card.

And fyi, power-spikes are a non-issue. The issue is with heat which is based on average current level. And you can not plug it in wrong as it is keyed. But contaminating it or the pins is a good example of what could happen.

And you can not plug it in wrong as it is keyed.

I fear you may be underestimating people's incredible ability to get things wrong ;-) (I guess one possibility is that the plug is somehow not fully engaged - so that the pins are only making partial contact, leading to higher resistance?)

But that is on the user, not the design.

Frequently not. The design should take into account reality. If you know that the cable will tend to be heading for a 90 degree turn right behind the connector, but your design is susceptible to poor contact or over stress the connector when that happens, that's a design problem. If you release something with a specified cycle of 10 reconnects or so, then you almost certainly have a design issue. If typical household environments have a high likelihood of too much contamination, then your design would be

Remember that most of these do not melt. Something wrong has to be done to them. But yes if you allow user-installation on something like this, you have to put in generous safety margins. Loading a 600W connector to 600W in that scenario is obviously asking for trouble.

True, this is going off of anecdotes with poor discipline with respect to overall data collection.

However it jives with what I have heard said from electrical engineers I know when they first saw the specifications. That style connector with that sort of power delivery sounds like it is just asking for trouble, particularly when oriented to head straight for where a case cover usually sits a few centimeters away. Since I heard people call it months ago at work and now precisely their predictions are coming

You should give these EEs credit. I did not see it back then, but it is not much of a surprise. Sure, this is not asking for every single one of these to melt, but it was pretty clear it would be more than a few and given the price-tag on the cards, that is just completely unacceptable. Connectors melting should be _rare_, like 1 in 1'000'000 or so when new. That is why you either restrict the load to something like 50% of rated load or make them extra sturdy. Just like in all other engineering, reliability

It sounds like the wiring needs to be upgraded to handle the current. That would go along with a better grade of plastic maybe PPS (Polyethelene Sulfide Plastic) which is rated up to 212C but that would require some data other than "it melted." It would be great to use an IR thermometer to see how high the temperatures go and validate the wire grade and temperature rating of that as well..

No, it is the contact pins loaded up to the limit in a relatively cheap design (Molex Mini-Fit Jr.). Any problem with the installation, like mechanical side-load on the connector, and the connector does not have reserves and heats up. Smart electrical designers always put in reserves. Guess the ones doing this design were not smart and thought you can go up to the maximum rated load with no problems.

It sounds like the wiring needs to be upgraded to handle the current.

The wires aren't melting. The connector is.

Given the relatively small surface area of the connector and the fact that the relevant surface area had no exposure to ambient air, it's not clear why you would blame ambient temperature for the connector melting. They're pushing too much power through too few pins of too small size.

It's the connector that's the issue, not the wire. Connectors are weak points.

12V, 600W means 50A, so any connector weaker than an Anderson SB50 would be insufficient.

It's the connector that's the issue, not the wire. Connectors are weak points. 12V, 600W means 50A, so any connector weaker than an Anderson SB50 would be insufficient.

It's the connector that's the issue, not the wire. Connectors are weak points.

12V, 600W means 50A, so any connector weaker than an Anderson SB50 would be insufficient.

This thing has 6 pins for ground and 6 pins for 12V, of which each is rated for 9A. Loading it to 50A is dicey though and essentially any problem can kill it. As was observed.

With up to 200A? No surprise.

Nothing to do with it. Power = Voltage X current. A switchmode powersupply (like basically all are these days) can convert voltages with ~90% efficiency.

If you have a 1000W powersupply at 120V @ 8.3A from your wall socket, a switchmode powersupply can provide 900W at whatever voltage, e.g. 12V @ 75A.

The capacitors exist to sort out tiny transients above 75A.

To elaborate on the above. You can get approximately 1500 watts out of ~120V ~15A outlet. You have 800W or 1000W power supply that takes most of it. This in turn split into multiple rails, because power supply can't deliver max wattage on a single rail. Don't remember exactly why, but it has something to do with AC to DC conversion. Basically, no household power supply does 50A for anything. You are talking Tesla Charger or Arc Welder territory.

Yes, you can get 15 amperes from a standard outlet. But the power supply transforms the 120 volts down to 12 volts, 5 volts, and possibly 3.3 volts as well. But assuming for the moment that only the 12-volt rail is providing current, you have a factor-of-ten reduction in voltage. That corresponds to a factor-of-ten increase in available current at 12 volts; the power remains constant, and power is the product of voltage and current, therefore a ten-fold decrease in one dictates a ten-fold increase in the other.

It's not quite as simple as that. First, there are the other output voltages that I mentioned above. Second, the overall power delivery efficiency is probably somewhere in the 80 to 90 percent range. Still, as a ballpark figure the current available on the output of the power supply is in the multiple-tens-of-amperes range.

How do you think small arc welders and plasma cutters can run from a 120V 15A outlet yet deliver 80 or more amperes of current? Low voltage and high current! It's all in the voltage transformation. You might want to look into it, given that it's the foundation for the grid that supplies all of our electrical power.

You can get approximately 1500 watts out of ~120V ~15A outlet.

V x A = W. 120V x 15A = 1800W.

This in turn split into multiple rails, because power supply can't deliver max wattage on a single rail.

What makes you think that? Pretty sure there's no reason you can't have 100A on a 12V rail if you want to. If you can't and somebody knows why I would be keen to know.

Basically, no household power supply does 50A for anything. You are talking Tesla Charger or Arc Welder territory.

No because again it's not just about amperage. A Tesla charger does have a reasonably high amperage but not drastically so if you compared to the ~150A that you can do at 12V out of a 120V/15A household socket (120V * 15A = 1800W. 1800W / 12V = 150A). A 150kW Tesla charger on the 480V platform is about 180A given

The DC on a CCS2 or Tesla charger are only one phase. Still, I like the comparison because it is ridiculously many amp through that tiny connector on the graphics card.

V x A = W. 120V x 15A = 1800W. A Tesla charger does have a reasonably high amperage but not drastically so if you compared to the ~150A that you can do at 12V out of a 120V/15A household socket (120V * 15A = 1800W. 1800W / 12V = 150A).

V x A = W. 120V x 15A = 1800W.

A Tesla charger does have a reasonably high amperage but not drastically so if you compared to the ~150A that you can do at 12V out of a 120V/15A household socket (120V * 15A = 1800W. 1800W / 12V = 150A).

You can only pull 80% of that continuously without the breakers tripping.

This in turn split into multiple rails, because power supply can't deliver max wattage on a single rail. Don't remember exactly why, but it has something to do with AC to DC conversion.

Power supplies have multiple "rails" (which might just be traces, today) which have different voltages. They used to also commonly have multiple 12 volt rails because of the current limits of the voltage regulators that were used back then, but I haven't seen a power supply like that in ages. These days they "all" just have one output circuit per voltage. Power supplies used to put out more current at 5V than anything else because the logic ran on 5V, but now the logic runs on all kinds of weird voltages, a

Basically, no household power supply does 50A for anything. You are talking Tesla Charger or Arc Welder territory.

If an RTX 4090 specification says that it can pull up to 600W (it does, and can) that means the power supply is providing 12VDC @ 50A, because W = V * A. However, that same 600W into the power supply is only ~5A at 120VAC line voltage depending on power supply efficiency.

Seriously, this is like 6th grade math.

This problem basically tells me that the connectors have been overrated and should be re-rated.

It's way too common to rate connectors at a current under ideal conditions, not under real world conditions. That's one of the problems that causes electrical fires in households. Oxidation and weakening of the spring tension ensuring contact are factors that ages the connector and decreases the rating over time. When the connector heats up then the spring tension weakens even more and the problem accelerates.

This problem basically tells me that the connectors have been overrated and should be re-rated. It's way too common to rate connectors at a current under ideal conditions, not under real world conditions. That's one of the problems that causes electrical fires in households. Oxidation and weakening of the spring tension ensuring contact are factors that ages the connector and decreases the rating over time. When the connector heats up then the spring tension weakens even more and the problem accelerates.

This problem basically tells me that the connectors have been overrated and should be re-rated.

Well, yes. Alternatively the card designers screwed up by loading these up to the limit. If you load a 600W connector with 600W and push out a few 1000 of them, some are going to have faults and some people will screw up the installation. The latter seems to have happened here by bending the wires in a way that puts too much mechanical load on the pins. Hence any sane design would not put more than, say 450W, on these connectors so there are some reserves for smaller problems.

Hence any sane design would not put more than, say 450W, on these connectors

The sane answer is to use a rated connector at the rated power in a service in which it is suitable for. It shouldn't be up to the end user (here being NVIDIA / AMD) to apply a derating due to mechanical deficiencies which aren't listed in the connector datasheets.

Hence any sane design would not put more than, say 450W, on these connectors so there are some reserves for smaller problems.

What is kind of hilarious is that Nvidia already did this, kind of. The 4090 has four power connectors on it, but only 3 are required. If you use the 4th one, it will lower the overall load on each connector to give you some headroom for sub-optimal connection resistance.

I think the people that are running into problems are using only three connectors, and probably two of them are on the same cable with a bare minimum (or less) gauge wire for hauling the amperage being asked of it. And then combine that

This problem basically tells me that the connectors have been overrated and should be re-rated.

Well yes and no. It tells me that the connectors are rated just fine for power, but not rated for the required service (plug/unplug cycles, bend radius etc). The issue isn't the power rating of the connector, it's that it is mechanically fragile to the point of making a poor connection under stress.

It can't possibly be 50A, your household power outlet is only rated for 15A.

Watts = Volts x Amps I suggest you look into what a power supply does. Read up on transformers, for example.

You might want to google ohms law.

Because we all know that GPUs are running at line voltage and on alternating current, right?

There's a reason there is a power supply in there. All amperes are not equal.

Aehm, my passive cooled 460W spare one puts out 38A at 12V. The 850W one I use in my PC is rated at 70A at 12V. So be shocked.

I would be shocked if you find one rate for more than 10A.

My Seasonic Prime has a single 12V rail. Max current on the rail is 83A. Limiting factor is current per connector and current per cable. My other PSU, a Bequiet Straight has 4 12V rails, each rail supplies more than 20A per rail, the rails for the GPU line (2 in parallel for the GPU output) supply 26A each so again more than 50A going to the GPU connector.

I think this is limitation due to size of step-down transformers that simply would not fit into PSU enclosure.

SMPS like PC power supplies do not have step down transformers. No consumer PC ever has. The limit is to do with switching frequency and performance of th

I would be amazed if you found a modern power supply with a rated wattage sufficient for this GPU that couldn't do at least 60A on the 12V rail. PCI-E power connectors in any reasonable supply are rated at 288W each, which is exactly why this 600W card has four connectors on it, and requires at least three to be used.

For example, this 850W EVGA power supply [media-amazon.com], which isn't even at the top of the range for their 850W supplies, can do 70.8A on the 12V rail according to it's spec sheet. And higher wattage power

They are pumping up to 50 amps at 12 volts through the connectors. There has been multiple investigative videos on the subject already by multiple outlets that speculate on what is happening. JayzTwoCents recently did a video on it ( https://www.youtube.com/watch?... [youtube.com] ) where he speculated that bending the cable caused a couple wires to pull back and make less surface area contact in the connection. The problem is that they are pushing 600 watts through this cable and less contact means more current throu

Actually, these connectors are fine up to 600W. But just barely. And only if handled correctly. Add any problem like side-wards mechanical strain and the connector pins do not reach the maximum rated current anymore and start to heat up, possibly massively.

So _technically_ this connector is still just barely sufficient. In the real world you want reserves and never load a pretty cheap piece of equipment (and the Molex Mini-Fit Jr is that) to its limits and even less so when intended to be installed by non-e

If you are using 20 AWG wire, I have no problem imagining it at all. This card can pull up to 600W across 3 cables, with an optional 4th. So you are talking about up to 200W per connector. And some power supplies ship with GPU power cables that have two connectors chained - some of them properly use 16 AWG wire to the first connector which would properly handle the 12VDC@33A load without overheating. But not all of them do.

Cheap daisy chain cables with 18AWG wire (or thinner) to the first connector woul

They're pushing too much power through too few pins of too small size.

Actually the pins are sized correctly and sufficient in number. The core issue here is that they are too sensitive to external stresses causing them to make poor contact.

It's still a not fit-for-purpose connector, but it isn't actually undersized.

This isn't a cooling issue. This is a insufficient-contact for the current inside the connector. It results in like 30A of current going through the contacts that were intended for 9.2A.

Not necessarily. I have no problem believing that someone would spend $1700 on a GPU, but go cheap on the power supply that provides shitty cables with insufficient thickness for the kind of amperage that $1700 GPU requires. Less thickness = more resistance = more heat. Until something melts.

There may be more comments in this discussion. Without JavaScript enabled, you might want to turn on Classic Discussion System in your preferences instead.

FAA Warns of Aviation Safety Risks Without US Mandate On 5G Limits

Rishi Sunak Is the First Crypto Enthusiast To Serve In UK's Top Office

Nothing motivates a man more than to see his boss put in an honest day's work.

High Qualty & Foctory Pitee

100% On-Time Shipment Protection

18 Years Manufacturering Experience

Nvidia RTX 4090's Power Draw May Be Too Much For Its Power Connector To Handle - Slashdot

Featured Products

News & Blog

Rapporti di crimini dalla carta assorbente della polizia suburbana del sud, sud-ovest

Kwajalein Lisesi Yaşlıları Yıllık Sokak Boyama Etkinliğinde İz Bırakıyor |Makale |Birleşik Devletler Ordusu

Lucid Releases Pricey At-Home Wall Charger

Caesarstone anuncia la fecha de los resultados del tercer trimestre de 2022

lars ljungkull turns plastic bags into sustainable skis with a marbled effect

Product Safety Report: 10pcs Wooden Magnet Cube Building Blocks Set sold via AliExpress (2206-0287) - GOV.UK

The Most Popular Elux Disposable Vapes in - GISuser

Braskem and SML Collaborate on PCR Stretch Film | plasticstoday.com

SEX ON THURSDAY | Lions and Tigers and Sex Toys, Oh My! | The Cornell Daily Sun

She-Hulk Finale Ending Explained: Hulk's Return and THAT Cameo | Den of Geek