• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA GeForce RTX 4080 Super Founders Edition

It would be supposition to assume that sold out = high demand or high sales. Unless you have the number of units Nvidia actually shipped or sales numbers it's impossible to say how many units were shifted. Heck being sold out isn't even a guarantee that any units were sold from a singular shop period, you are making another assumption that said store was able to acquire stock at this point in time in the first place. That's very well possible given there have been GPU launches in the recent and further past where certain retailers were assigned 0-5 units.

It's beneficial for companies to ship only what they know will sell out at launch because people tend to make the same logical mistake you did here. Most people immediately think sold out = hot / popular without actually taking two seconds to consider the fact that sold out in only analogous to that store being out of stock and nothing more. It's not an indicator sales or popularity without additional supporting information.

dont fall for the illusion of high demand at launch when deliberate limited quantities are at work. Its not uncommon for these types of marketing strategies to take shape. Limiting availability invites a sense of increasing need to pull the trigger or overall generally draws the illusion for a more high-demand desirable product. Can you imagine the headlines "RTX 4080 SUPER SOLD OUT WORLDWIDE" and its effects. Now add e.g. "only 1000 units were supplied at launch for the ~8 billion population" -all of that excitement just evaporates.



$950 for a XTX, hardly an underdog. Deserves all the criticism it gets. At this sort of price you'd expect nothing short of perfection!

my local microcenter had more than 120 rtx 4080 "super" units in stock on release day. no idea what the exact number is, they won't say. but it was more than 120. they were all gone by 1 pm. they had 20 units of the MSI card that looks like a founders edition rip off. that one went really fast. restocks have mostly sold out. the only 4080S restock not selling out are the PNY cards, which isn't really surprising. more cards come in every 2-3 days and the "premium" brands go right away
 
my local microcenter had more than 120 rtx 4080 "super" units in stock on release day. no idea what the exact number is, they won't say. but it was more than 120. they were all gone by 1 pm. they had 20 units of the MSI card that looks like a founders edition rip off. that one went really fast. restocks have mostly sold out. the only 4080S restock not selling out are the PNY cards, which isn't really surprising. more cards come in every 2-3 days and the "premium" brands go right away

The problem I have with this is that it took you 8 days to produce that number from the original comment. Right off the bat that eliminates you being a potential Microcenter employee and it eliminates the possibility of you having said number as a basis for your original comment in the first place. How you are making the observation that cards are restocking every 2-3 days other than potentially looking at the online numbers (which aren't accurate to real store numbers or restock times) is beyond me. I very much doubt you are driving down to the store everyday. If someone makes the argument that sold out = high sales and only returns with the stock numbers 8 days later you have a clear incentive to fabricate / fudge stock numbers. You are essentially saying that you went back to Microcenter specifically for some random internet argument, used the inaccurate online numbers, or somehow obtained the stock number through other 2nd hand accounts after the fact several days later. The 2nd option doesn't make much sense either given Microcenter only publishes current stock online and not past. In addition, most restocks don't come every 2-3 days. Anyone that's every had to fight scaplers for a video card knows that Microcenter gets a restock weekly or biweekly for a particular product a majority of the time.

Regardless of whether your number is accurate or not, the facts presented render your provided figure unadmissable as anything that can be taken as fact unless further evidence is provided.

Stock can vary a lot from location to location as well. In essence your agrument at this point is that because store x had 120 (unverified) units, all stores must have at least that much stock. It's one disengous argument based on a series of assumptions piled on top of another, which just leads me to believe that your orignal comment was in fact made without any knowledge of stock numbers.
 
my local microcenter had more than 120 rtx 4080 "super" units in stock on release day. no idea what the exact number is, they won't say. but it was more than 120. they were all gone by 1 pm

...and you were at the MC counter until 1pm bagging each sold unit?

I'm just curious was one of the customers a man with a black leather jacket who kept returning to the counter with the same card?
 
...and you were at the MC counter until 1pm bagging each sold unit?

I'm just curious was one of the customers a man with a black leather jacket who kept returning to the counter with the same card?
My local BestBuy had a bunch too. They were sold out before lunch.
 
They are "cheap" therefore they are not good. This logic not only make some people prefer Nvidia over AMD, but it also valid within Nvidia cards too.

Not the case.

They're just a less known brand, their actual cards are fine and perform well.

Check our reviews to see how they objectively compare against the competition.

People have herd mentality.

Hence the abundance of 7800X3D and ASUS GPU purchases. These aren't necessarily bad parts, in fact they're pretty good, but they're the best known, and so sell the most.
 

Not the case.

They're just a less known brand, their actual cards are fine and perform well.

Check our reviews to see how they objectively compare against the competition.

People have herd mentality.

Hence the abundance of 7800X3D and ASUS GPU purchases. These aren't necessarily bad parts, in fact they're pretty good, but they're the best known, and so sell the most.

Their CS the few times I've contacted them has been good at least here in the US. I still tend to go with other brands becuase at least with the high end stuff they seem to stick with a slightly lesser designs when it comes to vrms etc probably 0 to worry about and at msrp they are more than fine.
 
I read so many comments on the net about 4090s being better value, but 4090 FE is 55% more price than 4080 super, and is less than 55% more performant, I am so flush with cash now so looking at solving VRAM woes, but annoyed its triple slot, they have so over spec'd the cooler on this and seems to be some attitude now that a GPU can consume a case by itself.

Also buying one would still feel stupid as the MSRP is crazy compared to last gen.
 
I read so many comments on the net about 4090s being better value, but 4090 FE is 55% more price than 4080 super, and is less than 55% more performant, I am so flush with cash now so looking at solving VRAM woes, but annoyed its triple slot, they have so over spec'd the cooler on this and seems to be some attitude now that a GPU can consume a case by itself.
It can be pushed to 600 W, the cooler is fine.

Put a single slot waterblock on if necessary.

In heavy path tracing loads, the 4090 is more than 40% faster than the 4080, so it really depends how you're loading the card.
 
YES! This. cant beleive I didnt notice until I saw the article on videocardz. Why is nobody talking about this yet? Bigger deal than the extra few cuda cores IMO.
Guess it doesnt apply to those who dont buy founders though.
Reminds me of the cost cutting they did on the GTX285 compared to the GTX280 back in the day!
GTX280 one of my fav cards of all time!

Links to the TPU circuit board analysis pages.



Yep, I read about this on reddit, put in 4080 FE super review on google to get to TPU review, and was surprised to see although Wizzard mentioned the phase count he decided to not mention it was a reduction.

Even though it might not really be relevant as the original card was probably overbuilt, I think it should have been mentioned.

It can be pushed to 600 W, the cooler is fine.

Put a single slot waterblock on if necessary.

I know the cooler is fine, its actually over fine, this should have been releasable as a 2 slot card.
 
I know the cooler is fine, its actually over fine, this should have been releasable as a 2 slot card.
I should have phrased it "the cooler is appropriate". Sure at 450 W stock loads it may seem a bit excessive, but not all cases have great airflow, and as mentioned, the 4090 can pull up to 600 W if you tell it to.

Being a little overbuilt is better than a little underbuilt.
 
Well this is the 4080 super, they could have locked the power bar lower, it has the same TDP as the 3080?, and a bigger card can hinder case airflow. In my system a triple slot card will be right up to the back of a PCB. Also would cover up a NVME slot thats exposed with a double slot card.

The card seems designed for the card itself and no concern for overall system integrity.

Just release a 2 slot variant, cheaper with less materials used and restrict the power boost availability, job done.
 
I read so many comments on the net about 4090s being better value, but 4090 FE is 55% more price than 4080 super, and is less than 55% more performant, I am so flush with cash now so looking at solving VRAM woes, but annoyed its triple slot, they have so over spec'd the cooler on this and seems to be some attitude now that a GPU can consume a case by itself.
Well this is the 4080 super, they could have locked the power bar lower, it has the same TDP as the 3080?, and a bigger card can hinder case airflow. In my system a triple slot card will be right up to the back of a PCB. Also would cover up a NVME slot thats exposed with a double slot card.

The card seems designed for the card itself and no concern for overall system integrity.

Just release a 2 slot variant, cheaper with less materials used and restrict the power boost availability, job done.
I thought you were referring to the 4090.
 
I thought you were referring to the 4090.
Nope 4080 Super sorry. I think I mentioned the 4090 as a value comparison, sorry for the confusion. Of course for sure the 4090 I can understand being triple slot.

But anyway it wont be a showstopper, if I get one I will just grit my teeth with the slots.

Something I am thinking over, whether to wait for next gen (ar at least end of gen price drop), or buy now.
 
I own the RTX 4080 Super FE and they did a good job with its cooler. It is far more effective and quieter than the 2-slot cooler on my RTX 2070 Super FE (which is like 215W TDP -- about a hundred watts less than the 4080).

I also own an Asus TUF Gaming RTX 3080 Ti which has an even bulkier cooler but it's pretty quiet too.
 
My 320W 3080 has never gone above mid 50s with its 2 slot and is usually no higher than low 40s. Also pretty quiet.

I think the turing FE cards were an older design?
 
My 320W 3080 has never gone above mid 50s with its 2 slot and is usually no higher than low 40s. Also pretty quiet.

I think the turing FE cards were an older design?
Yeah the fans were enlarged amongst some other changes with the Ada FE cards.

Very similar design language, but refined between RTX 30xx and RTX 40xx.

Unfortunately the 3080 Ti which is a significantly larger die/more power hungry than the 3080, uses the same two slot cooler. It's very pretty, but now I use it as a display piece above my wardrobe, and the 400 W card is under a waterblock.
 
Yeah the fans were enlarged amongst some other changes with the Ada FE cards.

Very similar design language, but refined between RTX 30xx and RTX 40xx.

Unfortunately the 3080 Ti which is a significantly larger die/more power hungry than the 3080, uses the same two slot cooler. It's very pretty, but now I use it as a display piece above my wardrobe, and the 400 W card is under a waterblock.
The die is not larger between 3080 / 3080Ti. They are both GA102. The 3080Ti does use more power though.

They were originally both supposed to be GA103 based cards, hence the 2 slot cooler, but GA102 yields were so bad that apparently samsung gave nvidia the cast off dies for free, so they ended up being used as 3080s instead.

I also like to use the reference coolers as display pieces while the card is water cooled:
Resized_20240422_081159_1713787934694.jpeg
 
The die is not larger between 3080 / 3080Ti. They are both GA102. The 3080Ti does use more power though.

They were originally both supposed to be GA103 based cards, hence the 2 slot cooler, but GA102 yields were so bad that apparently samsung gave nvidia the cast off dies for free, so they ended up being used as 3080s instead.

I also like to use the reference coolers as display pieces while the card is water cooled:
View attachment 344730
I get what you are saying, but the die is not technically the exact same. It's 102-225 vs 102-200. They're all cut from the same silicon sure, but it's not like it's just an upclocked 3080 or something. Much more of the chip is physically active including a wider memory bus.

Blame Samsung for the 10nm 8nm process. The GDDR6X being very power hungry didn't help either. In fact much of the difference in performance from 3090 to 3090 Ti was from the better density memory chips of the Ti offering a lower overall power consumption (helping the core have more of the power budget) compared to the double amount of chips in the 3090 consuming so much power.

1713788581508.png


3080Ti is much closer to 3090 than to 3080. They technically have the same power limits but we know that the 3080 Ti/3090 run hotter and use more power. All of Ampere was power limited essentially all the time, so you can still have a 3080 push 400 W, but it's to minimal gain. 3080 Ti FE especially ran hot because it reused the small FE 3080 2 slot cooler, while only the 3090 got the bigger cooler.
 
The part # is different of course, but the number after the "GA102-" does not indicate die size, just stepping. They are still both GA102.
Yes. 3080Ti is more like 3090 than 3080.
 
The part # is different of course, but the number after the "GA102-" does not indicate die size, just stepping. They are still both GA102.
Yes. 3080Ti is more like 3090 than 3080.
We know this, but you could argue the 2060 KO which used same die as 2070/80 is technically the same die too, and while it did have slight performance improvements over the regular 2060, because of the massive differences in activated parts of the die, you still can't really call the 2060 KO and the 2080 the same chip. This is why I tend to go with the activated silicon reasoning argument rather than what may be physically there but inactive.

I did mistype when I said "significantly larger die", that is technically incorrect. Just going off memory of the CUDA core count and memory bus comparison and the "active silicon" thing I mentioned.

Do Zen x600 and x800 chips use the same die? Maybe technically, but I'd argue having 25% fused off/disabled for 2 fewer cores kinda changes it. I guess it's a matter of interpretation.
 
Ok here we go, I am sure this will cause a few smiles.

I grabbed the 4080 Super out the box, and had a look to see how it would fit, well wow.

The triple slot is actually not a problem, as there is 3 empty spaces to the next usable pci-e slot, so its just extra bulk, but the length of the card I will probably remove 5 of my drive bays, its 50/50 if it will go in with the bays left in place, but the gap is going to be 1-2 mm if it fits. I wonder what made Nvidia think they had to lengthen this card from the standard length by a few cm and how much that was needed.

If the bays go, I need to ditch a HDD as there is 4 in the rig and the bottom HDD casing is only 3 bays, the SATA SSD can be dumped anywhere I guess even if I just chuck it in the 5inch bays. The obvious HDD to ditch in terms of file organisation is annoyingly the newest one.

If I ditch the HDD, might replace with a 860 EVO 4TB to avoid having to squeeze files on other drives.

Need to wait 2 weeks for 12V cable to arrive, so time to think it over.
 
Ok GPU swap is done.

In the end the storage wasnt the headache, I had to remove the drive bays which is better in the long run anyway so is breathing space around the card, I shoved the SATA SSD in the 5 inch bay above my long disabled DVD optical drive, its not particularly loose so will stay there, I kept the HDD in its caddy and blutak'd the caddy to top of the 3 bays still in the case, obviously I need to be wary if I ever tilt the case or move it around but its good enough for now with it upright, no vibrations etc. I will after moving/deleting enough data remove a spindle from the rig, so that contraption is temporary.

At first the card was dead, and had to shut down PC via remote desktop, looks like it wasnt satisfied with me only using 2 8 pin cables, so a 3rd went albeit via pigtail which I know Nvidia doesnt like. I do have a native cable en route for the PSU but shipping is super slow.

The nightmare was when I removed the 3080 I had to move the wifi bracket as it was in the 3rd slot which the triple slot 4080 super needs, I had removed it fine to move it but then decided to tighten the screw attachments as one became loose, the twisting yanked out one of those super fragile connectors from the wifi card, then when removing the wifi card, the other one came out urrgh. 3 hours later both back in, but then whilst attaching the bracket back to case one came out again, with the slightest of movements (who designed these connectors wtf), I then figured out a way I could reattach it without taking it all out the case again, luckily only took about 20 minutes instead of another 3 hours.

After kind of getting the wifi cables in a way I think the 4080 super would cover them without disconnecting, I installed the GPU, the thing is so big I could not see the PCIe slot at all, I did it blind more or less just keeping it aligned with the slots on the case, and then I felt it slide in so was good, the card without me supporting it was tilting away from the case a little, perhaps even bending the board a little, in addition because the card is raised a lot from the bracket, getting the first screw was a real challenge, something a work bench installation wouldnt have a problem with, I had a tiny gap to get those first few turns on the screw whilst lifting the card the other end so it was aligned, after that was easy to screw in with screwdriver, and once the first was in, the other 2 were somewhat easier.

So whilst its in now, I think the size is just ridiculous, side by side to the 3080 (which I think is a big card in itself) it looks enormous, probably at least twice the heatsink material, as its not just triple slot, its also taller and longer, this is with both cards the same TDP, it just isnt needed Nvidia. For same reason a 3rd PCIe cable really shouldnt be needed either. Previous Nvidia products with similar or same TDP were 2 cables and much smaller heatsink.

I think when heatsink's get this big the vendors need to be moving onto AIO.

Final note I suspect one or both of the wifi cables is loose again with 4080 maybe knocking it too much during installation, as wifi signal is mega low, but I use ethernet day to day, so if its broken it is what it is now.
 
Last edited:
Back
Top