• 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.

Discussion Mediatek SoC thread

Page 17 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
8 elite gen 5 big core looking better and better with a seemingly considerable area deficit (2.2 mm2 according to Geekerwan). Vs C1 Ultra, we don't even need to consider sme area as both have it.
 
Meteor Late said:
8 elite gen 5 big core looking better and better with a seemingly considerable area deficit (2.2 mm2 according to Geekerwan). Vs C1 Ultra, we don't even need to consider sme area as both have it.
Click to expand...
Yeah, Oryon v3 is the better core for sure. Though the C1 Ultra has a 2MB private L2. Without it, they would be about the same size.
 
CouncilorIrissa said:
Wow the Premiums suck. What on Earth is this:
View attachment 130773
Borked physdes? Same L2 capacity, but the integer pipeline of the Premium should be very similar to the Ultra. And it's not like X4 was implemented by a different company, it's literally last years' MTK design.
Click to expand...
I think this time around ARM only managed to markedly improved the Ultra. I'm sure next gen Dimensity 9600 will come with a better improvement on the Premium and the Pro. Hope the next gen Ultra would get the same big improvement too.
 
Tigerick said:
Even with half of memory bus, D9500 managed to shown improvement up to 2.7X.
Click to expand...
1 - You're comparing a cut-down version of the Snapdragon Purwa die to the full Dimensity 9500. At least you'd need to compare the full die version of Purwa which has a 2.1 TFLOPs GPU.

2 - 2.7x difference in theoretical FP32 throughput in GPUs from different architectures doesn't mean much, if at all.

3 - It makes more sense to compare the D9500 to the Elite Gen 5 which are indeed SoCs with similar market positions and same process node. In that case, the Elite Gen 5 gets better geekbench results for both ST and MT.
 
Just bought Vivo X300 (global version with 6040 mAh battery, from France with EU power adapter inside the box) and D9500 sure does feel snappy. 899 € for 16/512GB model - probably right time to buy that considering the RAM prices.
 
MerryCherry said:
wccftech.com

A Wild Rumor Suggests Intel Has Roped In MediaTek's Dimensity Chips For Its 14A Process

After Apple tentatively tapped its 18A process, Intel appears to have secured MediaTek as a major customer for the bleeding-edge 14A process.
wccftech.com wccftech.com

View attachment 138106
Really, Wccftech?

Why would Mediatek, who is Taiwanese, and deeply partnered with TSMC, switch to Intel?
Click to expand...
Wait... remember that nVIDIA and Mediatek partnered for notebooks? Mediatek despite is Taiwanese, nVIDIA is from USA, so yeah, they could use the process to get those notebook processors without issues (remember that TSMC .
 
LightningZ71 said:
That would have gone bankrupt long ago.
Click to expand...

GF quit the 7nm race in 2018. A year later there was massive demand for mining ASICs on 7nm, and a year after that we had the COVID shortages leading to massive profits to TSMC and Samsung, and a couple years later we have the AI race.

Not to say they would be making a lot more money than now, since they'd still be lagging behind TSMC, but it's not like the market didn't demand a whole lot of additional EUV nodes than it has had since 2018.
 
They didn't have the operating capital to purchase any EUV machines. They didn't have the profit to even finance completing the development of their 7nm class node. Their "12nm" class node, which they essentially bought from IBM and Samsung, lagged behind Intel's 14nm node. They were behind in every measure, and even GF12+ wouldn't, and didn't command enough of a price to convince them to finish GF7.

To show how bad they are, 22FDX was supposed to be released half a decade ago, but missed that goal badly, and I don't know if they ever hit HVM on it's supposed follow on 12FDX node. They have some specialty niches, but their R&D is just not good.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top