• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Rumor: TSMC will manufacture Vega 20

FIVR

FIVR

Diamond Member
http://www.digitimes.com/news/a20170908PD210.html

AMD will continue to contract Globalfoundries to manufacture its next-generation Vega GPUs, and will have Siliconware Precision Industries (SPIL) as its major packager for the chips, according to industry sources.

The supply of AMD's Vega 10 GPUs has fallen short of demand, thanks to robust demand for artificial intelligence (AI), Bitcoin mining and gaming applications since 2017, the sources said.

Globalfoundries uses its 14nm FinFET process technology to fabricate AMD's Vega 10-series GPUs, and is expected to land orders for the next-generation Vega 11 series, the sources indicated.

Packaging specialist SPIL, which has already obtained orders for AMD's Vega 10-series chips, will continue to hold the majority of backend orders for the Vega 11 series, the sources noted.

The popularity of AMD's and Nvidia's high-end GPUs for applications such as AI has drawn much attention to the importance of packaging technologies for 2.5D/3D solutions, the sources identified. Heterogeneous integration for use in high-bandwidth memory (HBM) devices also requires such advanced packaging techniques.

Taiwan Semiconductor Manufacturing Company (TSMC) with its CoWoS (chip-on-wafer-on-substrate) technology has reportedly secured orders for AI chips from Nvidia and Google. TSMC has further enhanced its advanced packaging capability eyeing a bigger presence in the supercomputer field, the sources said.

AMD's rumored Vega 20 series will be applied to supercomputers, according to the sources. AMD is likely to switch orders to TSMC, which will use its 7nm FinFET process and CoWoS packaging to fabricate the Vega 20-series GPUs, the sources noted.
Click to expand...


Perhaps we will finally see exactly how much nvidia's performance advantage is due to the process they use 🙂
 
GTX 1050 and 1050ti do have slower clocks compared to the GTX 1060 and above, The 1050 boost clock is at 1455MHz, while the 1060 has a base clock of 1506MHz, even the 1080 has a 1607MHz base clock. Boost clock for 1060 and 1080 are both 1700mhz+, and can easily reach 2000MHz+ with most 1060's regularly reaching 2100MHz when overclocked.

So it seems to me that Nvidia has had much better luck with TSMC process and that it gives them at least 10% advantage over AMD, if not more. Clearly most of their advantage is architectural, but even just 10% advantage from the process is huge and can mean a win or loss for a particular GPU.
 
FIVR said:
Perhaps we will finally see exactly how much nvidia's performance advantage is due to the process they use 🙂
Click to expand...

We already sort of know this because Apple had some of their SoCs fabbed using both TSMC and Samsung (who's process GF used) and it while there was a difference, it isn't anywhere near enough to make Vega 64 a Titan-killer or anything like that.

I suspect that AMD isn't doing this because they need the performance bump (though it will not doubt be an added bonus) but mainly because their CPU business is now competitive again and eating up all of GF's capacity. It's more economical for AMD to use as much of those wafers as they can for CPUs as they're going to make more money doing that.

They still need to get their driver situation figured out and hope that the price of HBM comes down, otherwise even a better process won't save Vega.
 
Guru said:
GTX 1050 and 1050ti do have slower clocks compared to the GTX 1060 and above, The 1050 boost clock is at 1455MHz, while the 1060 has a base clock of 1506MHz, even the 1080 has a 1607MHz base clock. Boost clock for 1060 and 1080 are both 1700mhz+, and can easily reach 2000MHz+ with most 1060's regularly reaching 2100MHz when overclocked.

So it seems to me that Nvidia has had much better luck with TSMC process and that it gives them at least 10% advantage over AMD, if not more. Clearly most of their advantage is architectural, but even just 10% advantage from the process is huge and can mean a win or loss for a particular GPU.
Click to expand...

most 1050s are the highest power Nvidia cards without the 6 pin connector and have to deal with power constrains, so I'm not sure.
pretty sure the 1050 TIs with 6 pin connector and a little tweaking can go to the driver limit easily (1.91GHz)
 
Switching over to TSMC for 7nm seems a little... dubious to me, when GF's 7nm will be from their ex-IBM team and designed as a pure high performance node. TSMC's 7nm node will be another mobile-oriented node, won't it? And AMD would have to compete for wafers with Apple and the other foundry customers. I could be wrong, but that just doesn't sound like a great idea.

Arachnotronic said:
GTX 1050/GTX 1050 Ti are manufactured on Samsung 14nm LPP, so I don't think process is giving NVIDIA much of an advantage. NV really just has a superior architecture right now.
Click to expand...

From what I remember, the 1050 Ti usually tops out at ~1900 Mhz. That's 100-200 Mhz below its larger siblings, so it seems likely that a portion of the clockspeed difference between Pascal and Vega is process related. Another 200 Mhz of scaling would at least make Vega clearly superior to the 1080.

That being said, "It would be just over the performance of the competition's second-best card" is damning with faint praise. Especially when you consider the disparity in hardware and power use between the two. The process disparity argument might be compelling if Vega was nipping at the heels of GP102, but as it is...

Man, we know Vega's a monster in compute. Why didn't that translate over to rendering?
 
SPBHM said:
the driver is locked at 1911MHz max, it's not the chips real highest clock.
I think it's more due to power limitations from the board design and so on.
Click to expand...

Is it? I'm pretty sure I saw reviews where they maxed out above or below that.

I could be wrong though; I've never had that much interest in the bottom-tier cards.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top