General Discussions

[affe-mit-waffe]

Hello Lady's and Gentleman's.

I am Asking maybe a wired question but i cannot understand why you choose the produce such small ship.

When you can attack or take advance of the high end counter part like 2080 and 2080Ti,

Why you guy's from AMD do not have taken the change to crush the concurrence for once, you deliberately choose to make that ship small (251 mm²) when you must know that the perfect or sweet spot for chip is between 330 mm² or even 380 mm². 

I really don't understand why you choose to stay behind, just give us high end rather that passing for second in this race once again.

Thank for your time and have a nice day folks!

[majaczos22]

You know that Navi 10 is made in 7nm process, right? It makes a chip with the same number of compute units 40 or even 50% smaller. But it's also 2x more expensive. In short - Navi is as fast as 450mm² 14nm chip and costs as much (or even a touch more) to make.

[mstfbsrn980]

No one from AMD will answer this question because it's a sales strategy. If AMD builds a 350mm² GPU, special solutions are needed to cool this GPU. Or GPUs come with low TDP relative to mm² as nVidia does. They can't make money. Inference to mm² is not correct, because nVidia's 350mm² GPUs contain RT cores and RT cores are also inactive in most games.

[majaczos22]

Actually the 2080 is 545 mm², not 350. 2080ti is 754 mm².

[mstfbsrn980]

Not important. Because nVidia now produces with RT cores. CUDA TDP is high when RT is not running, low when running so temperature always is stable. But CUDA number is very high, it is very costly. AMD produces low-cost cards and is more attractive in price. Sales policy. In short, if AMD had sold its products with low voltages and low TDP (by processor architecture) as Intel and nVidia did, they wouldn't have made any money. I can write that much with inadequate English.

[affe-mit-waffe]

Hello guy's, first thank you for your reply.

I know it's 7nm, but with a larger ship and no blower (Btw no one like this solution) you can make something great.

I really feel that people are tired of Ngredia Strategie with big chip a wouping price, it's ridiculous.

I would love to see another radeon 5000 serie and kick the nuts of Ngreida.

[majaczos22]

Thing is - because it's 7nm it's very expensive to make. It's also difficult to make because the larger the chip is - the lower the yield.

But Navi 10 is not going to be the only GPU they make. Drivers pont to two more RDNA1 chips in the making - Navi 12 and Navi 14. It appears that the former one is going to be above Navi 10 (5700) and therefore bigger.

[hardcoregames_]

nVidia is using an older 14nm while AMD is using 10nm (7nm is marketing nonsense) which is about what Intel is doing as well. AMD and Intel both have higher costs at 10nm due to the complexities of the manufacturing process.

A safer and more reliable metric I like is the number of transistors per square millimeter

[affe-mit-waffe]

Yes but number of transistors is related to the die seize too.

[hardcoregames_]

I did an informal study of cpu's from intel and amd which have been around since before the PC existed

The number of transistors per square mm has improved but now the change has slowed for a while, which suggests the cost is more brutal than expected