NVIDIA discussion [2024]

homerdog said:
What is the estimated BOM for the 4080? MSRP of 4080 is 50%* higher than 3080. Does it really cost 50% more to produce? I mean a whole GPU, not just the die.

*inflation for adjusted. $700 in Sept. 2020 is $800 in Sept. 2022. :oops:
Click to expand...
Haven't you been listening since the 40 series release? Die size, BOM and historical reference aren't relevant anymore, it's only relevant what Nvidia decide the price is.
 
In general I don't feel people should emphasis BOM as it relates to pricing. BOM does affect pricing in the sense it changes the margins but these aren't cost+ pricing models.

Just take the RTX 4080 Super, it uses a more enabled GA103 die and faster memory over the RTX 4080. Maybe TSMC 4N has dropped so much in cost over the 1 year or so?

But we have more examples in the past of price drops over a time period of just months. RTX 770/780 post release. GTX 280. Fury Nano (somehow cheaper to produce then Fury XT suddenly after a few months?) 5700/xt (not even released...) just as some from both sides.
 
www.guru3d.com

Scammers Selling Used RTX 4090 Graphics Cards without Memory or GPU in China

In recent developments originating from China, a concerning trend of counterfeit RTX 4090 graphics card sales has emerged. These counterfeit cards are notably lacking critical components, including the GPU and memory modules.
www.guru3d.com www.guru3d.com
One notable case reported in Hong Kong involved an individual who purchased an RTX 4090 for approximately $1,660, only to discover that the card was non-functional. Despite their attempts to seek assistance from local authorities, their efforts proved fruitless, leading them to seek media coverage through specialized Chinese press blogs. Online forums, particularly Chiphell, have seen a surge in similar complaints. Prospective buyers, particularly those interested in used GeForce RTX 4090 cards, often fall victim to enticing pricing without realizing the potential for fraudulent transactions.
Click to expand...
 
trinibwoy said:
Eh when has die size and BOM ever dictated price? Price has always been a function of competition and what people are willing to pay. Basic economics.
Click to expand...

It's that and the fact that the floor price will also be impacted by all of the other R&D costs. When you buy an Nvidia gpu you're paying for all of the graphics researchers, the DLSS training, the engineering hours on the SDKs, the driver team etc. If they just just put a low margin on the BOM they'd go out of business.
 
Scott_Arm said:
It's that and the fact that the floor price will also be impacted by all of the other R&D costs. When you buy an Nvidia gpu you're paying for all of the graphics researchers, the DLSS training, the engineering hours on the SDKs, the driver team etc. If they just just put a low margin on the BOM they'd go out of business.
Click to expand...
That's a good point but they won't be going out of business if they lowered gaming GPU prices lol. Their market cap is higher than the GDP of Turkey.
 
homerdog said:
That's a good point but they won't be going out of business if they lowered gaming GPU prices lol. Their market cap is higher than the GDP of Turkey.
Click to expand...

Yah, at this point I guess the gaming business is small to them, and it's not the majority of their operating expenses. Still, there's more to pay for in a product than a bill of materials. It's BOM plus labour, and in tech products labour costs can be enormous once you factor in R&D.
 
homerdog said:
That's a good point but they won't be going out of business if they lowered gaming GPU prices lol. Their market cap is higher than the GDP of Turkey.
Click to expand...
But that's mostly thanks of datacenter revenue, not gaming. If they made less money on gaming, they would naturally invest less and less of their R&D into gaming GPUs, and then everyone would complain even more about the lack of innovation. Alternatively, NVIDIA actually pricing their GPUs aggressively and crushing AMD/Intel to the point where the latter deprioritise discrete graphics in favour of AI is not really in anyone's interests.

Of course, you could argue they are already putting less R&D into gaming: Turing was a huge step forward with an insane amount of graphics-centric innovation, but (arguably) in 4 years they delivered fewer significant architectural changes with Ampere/Ada than some other GPU vendors (including PowerVR at its peak) achieved with just 1 year projects. On the other hand, their tensor cores have been heavily rearchitected every single generation and there's just a huge amount of AI/HPC-specific stuff in Hopper in general.

I guess we'll see how much of a graphics-specific improvement (features and PPA) there is in Blackwell... I'm cautiously optimistic, although I guess if they believe AI is at the core of future 3D graphics, then it might end up a bit disappointing for everything else, we'll see...
 
The fact that Nvidia is making more and more money on DC products puts more pressure on the gaming side of the business to bring in more profits, not less, to keep up with the DC side and be just as profitable. If it would be the other way around the gaming business would just wither and die eventually - which I doubt that anyone here want to happen. The idea that Nvidia would for some reason subsidize the gaming products from their DC profits is just insane, that's not how any commercial company does business.
 
DegustatoR said:
The fact that Nvidia is making more and more money on DC products puts more pressure on the gaming side of the business to bring in more profits, not less, to keep up with the DC side and be just as profitable. If it would be the other way around the gaming business would just wither and die eventually - which I doubt that anyone here want to happen. The idea that Nvidia would for some reason subsidize the gaming products from their DC profits is just insane, that's not how any commercial company does business.
Click to expand...

When there's limited CoWoS capacity they can't just expand DC stuff and gaming still brings healthy profit, not to mention synergies in R&D, why would you shoot a milking cow just because another cow milks more and you can keep both?
 
Kaotik said:
When there's limited CoWoS capacity they can't just expand DC stuff and gaming still brings healthy profit, not to mention synergies in R&D, why would you shoot a milking cow just because another cow milks more and you can keep both?
Click to expand...
There aren't much synergies in R&D there nowadays and CoWoS capacity being limited doesn't mean much when you can sell gaming grade GPUs on AI markets. It's just a running misconception that a company should (why?) sell its products cheaper because it can compensate the lack of profits on that market by profits from the other market which does good right now. Nvidia is not a charity foundation, it doesn't make products so it can give them away for free to those who can't buy them.
 
del42sa said:
https://videocardz.com/newz/modder-shrinks-geforce-rtx-4070-ti-to-itx-form-factor

like this small form factor (y)
Click to expand...
Cards should be smaller across the board. My baseline MSI RTX4070 Ventus 3x is gigantic. It barely fits in my case. There's no reason for this. It is hard locked to 200W. Yea I could have gotten the Ventus 2x but I didn't really think about it and it didn't occur to me that they would make something so large for a 200W card.

And then there's the 4090 which is so heavy it needs a kickstand or it might break off the PCIe slot or twist until the core pops off. Ridiculous.
 
homerdog said:
There's no reason for this. It is hard locked to 200W. Yea I could have gotten the Ventus 2x but I didn't really think about it and it didn't occur to me that they would make something so large for a 200W card.
Click to expand...

Noise tolerance preferences can vary greatly. 200w is actually not simple to cool with just 2 fans, I'm guessing the Ventus 2x will likely need closer 2k fan speeds (if not higher) which at least for me is not preferable.

Also it might be counter intuitive but those triple fan long style coolers on the base models are simpler/cheaper than perhaps a smaller 2 fan design to achieve the same results. You'll notice the base 3 fan models often don't outperform some smaller 2 fan designs for example. But those 2 fan designs often have to employ more "complex" heatpipe layouts and fin designs among other things.
 
arandomguy said:
Noise tolerance preferences can vary greatly. 200w is actually not simple to cool with just 2 fans, I'm guessing the Ventus 2x will likely need closer 2k fan speeds (if not higher) which at least for me is not preferable.

Also it might be counter intuitive but those triple fan long style coolers on the base models are simpler/cheaper than perhaps a smaller 2 fan design to achieve the same results. You'll notice the base 3 fan models often don't outperform some smaller 2 fan designs for example. But those 2 fan designs often have to employ more "complex" heatpipe layouts and fin designs among other things.
Click to expand...

Yah, fans and plastic shrouds are a lot cheaper than copper and vapor chambers.
 
arandomguy said:
Noise tolerance preferences can vary greatly. 200w is actually not simple to cool with just 2 fans, I'm guessing the Ventus 2x will likely need closer 2k fan speeds (if not higher) which at least for me is not preferable.

Also it might be counter intuitive but those triple fan long style coolers on the base models are simpler/cheaper than perhaps a smaller 2 fan design to achieve the same results. You'll notice the base 3 fan models often don't outperform some smaller 2 fan designs for example. But those 2 fan designs often have to employ more "complex" heatpipe layouts and fin designs among other things.
Click to expand...
I can't really argue this point, but my GTX970 consumed around 200W after OC (probably more like 175W) and it wasn't nearly as long and wasn't very loud. It was also way cheaper but I guess that ship has sailed.
 
You must log in or register to reply here.

Similar threads

P
GTC 2024
2 3
Replies
56
Views
4K
trinibwoy
T
W
New Qualcomm silicon for new Surface May 2024
2
Replies
22
Views
2K
Cyan
Cyan
P
AI Desktop Tools - Training and Inferencing
2
Replies
29
Views
3K
pharma
P
D
RTX Video HDR modded to work in games, RTX Game HDR?
Replies
7
Views
1K
DegustatoR
D
Shifty Geezer
Digital Foundry Article Technical Discussion [2024]
62 63 64
Replies
1K
Views
99K
Remij
R
Back
Top