One of the biggest consumer graphics chips ever produced by AMD? Youtuber claims Big Navis Die Size is 536mm2.

[AlTech]

4 minutes ago, LAwLz said:

And MLID was wrong on pretty much everything else he said about the 30 series so I wouldn't bet on him getting the price right either.

Not really. He got a couple things wrong as I've said before and he fixed that. This just comes across to me as hating on MLID for no valid reason.

[leadeater]

5 hours ago, AluminiumTech said:

For the 3080 specifically the cooler required is between $150-200. MLID says the total bill of materials for the 3080 cards is around $500-600.

Bill of materials is pretty well impossible to know or guess because Nvidia does what ever they want with the GPU package prices, and what they charge AIBs is not what it costs either.

[Albal_156]

6 hours ago, AluminiumTech said:

For the 3080 specifically the cooler required is between $150-200. MLID says the total bill of materials for the 3080 cards is around $500-600.

Would make sense if the AIBs can't make 3090s for MSRP. Though I'd say that bill of metrials couldn't be accurately predicted.

 

46 minutes ago, leadeater said:

Bill of materials is pretty well impossible to know or guess because Nvidia does what ever they want with the GPU package prices, and what they charge AIBs is not what it costs either.

You could say its pretty expensive but an actual price would be difficult as you said.

 

Think AMD will have the upper hand with that as well with the cooler looking like a standard three fan or dual fan affair and likely being much cheaper to produce. Fs in the chat for the AIBs and for us as well because I don't think AIB partner cards will be coming for RDNA 2 at launch which makes sense; the AIBs are leaky af and NVIDIA would known exactly what AMD had well in advance and could have prepared for it. Hopefully the reference cooler is decent and not terrible like AMDs previous blower cards but I think I'll be waiting until I can consider all the options.

[DeScruff]

On 10/5/2020 at 10:10 PM, TetraSky said:

I'm surprised it took AMD this long to made a beefy GPU die. I wonder why they limited themselves to that size though instead of going ham on it and equaling or surpassing Nvidia's die size for more transistors ?

For a while >10 years ago ATi/AMD and Nvidia dies were getting bigger and bigger, but that trend ended.
The big two disadvantages of a big die from a manufacturer's viewpoint is that:

A) Your making a lot less chips per wafer. - More wafers = Longer to manufacture, and costs more.

B) The chance for a chip to fail is higher. And with less chips in general your failure rate can be extremely high.

Kinda an extreme example.
The one on the right has dies that are 2x the size as on the left.(for ease of making this image). The Green dots are imperfections on the wafer and are in the same exact spots on both wafers. Red are failed, Blue are perfect.

Left has 44/52 chips that are functional. Right has 5/12 chips that are functional.
Even if all the bigger ones were functional, that only making 27% of the functional smaller chips. Hooray exponentials!

[Loote]

I still think it would be interesting to prove they can, but maybe a halo card is not as important as some people claim. I also saw multiple mentions that AMD's architectures weren't going to scale to bigger die sizes, that was also said to change in RDNA - finally being able to make big dies which actually improve performance.

 

Can't wait for it to come out.

[porina]

12 hours ago, AluminiumTech said:

For the 3080 specifically the cooler required is between $150-200. MLID says the total bill of materials for the 3080 cards is around $500-600.

There's no way BOM cost is that high unless either nvidia or their partners are losing money on every card sold. BOM cost is literally the cost of the components used to make a product, and doesn't include anything else. No licencing fees, shipping, development, non-product related costs... 

 

One way to get such a high number is if you incorrectly use retail values for example, but that is not how manufacturers count BOM cost. Or it could be wrong in many other ways such as grossly over-estimating the cost of key components. Take the cooler, if you look at it, there's not really anything special about it in terms of mechanical construction. I can't see it being $150 retail value, let alone $150 BOM.

[LAwLz]

1 hour ago, porina said:

There's no way BOM cost is that high unless either nvidia or their partners are losing money on every card sold. BOM cost is literally the cost of the components used to make a product, and doesn't include anything else. No licencing fees, shipping, development, non-product related costs... 

 

One way to get such a high number is if you incorrectly use retail values for example, but that is not how manufacturers count BOM cost. Or it could be wrong in many other ways such as grossly over-estimating the cost of key components. Take the cooler, if you look at it, there's not really anything special about it in terms of mechanical construction. I can't see it being $150 retail value, let alone $150 BOM.

Agree.

For comparison, the BOM for the iPhone 4 was 188 dollars.

 

So MLID (the clown that he is) is claiming that the cooler for the 3080 costs as much to make as an iPhone. There is simply no way that is true.

And like I said earlier, he was wrong on pretty much everything else regarding the 3080 so I don't get why we should believe him when it comes to the BOM for the cooler.

[Bombastinator]

5 hours ago, LAwLz said:

Agree.

For comparison, the BOM for the iPhone 4 was 188 dollars.

 

So MLID (the clown that he is) is claiming that the cooler for the 3080 costs as much to make as an iPhone. There is simply no way that is true.

And like I said earlier, he was wrong on pretty much everything else regarding the 3080 so I don't get why we should believe him when it comes to the BOM for the cooler.

I don’t agree that MLID is a clown.  He is a gossip columnist and weirdly gossip column work requires immense journalistic rigor to obtain even middling results, so I suspect he has reasoning and diligence behind his estimates.  That said anything done there can only be an estimate and might not be accurate so “believing” as in taking as gospel simply should not be done with anything on that site.  And of course there is always the possibility that something or even simply some part of the equation was simply pulled from his ass. So a fuzzy picture, possibly with immense errors.  Salt must be used.  Cant take it for more than it is. “Best possible” doesn’t always even mean “good” and is very rarely even close to perfect. Even a fuzzy picture can be better than none though.

Edited October 7, 2020 by Bombastinator