Why I feel that the Chinese will pull the 100 SKU in Mini PC using those unused processors?
And with that, with ALL the flaws corrected.
Why I feel that the Chinese will pull the 100 SKU in Mini PC using those unused processors?I guess, but getting Linux support for it would have been boss. Only real flaws from an end-user point-of-view were:
Wrong version of Win11
Having to use USB-C for display
It was easily the best-cooled, highest-end X Elite machine available. If you really wanted to tinker with X Elite to see what it could do pushed to its limits, it was the best rig for the job. Good luck finding the 100 SKU in laptops. It's practically unobtainium.
? GMEM = Graphics MemoryOnboard GMEM. For DLSS-like upscaling?
Yes, I edited it out.? GMEM = Graphics Memory
Do you mistook it for GEMM (General Matrix Multiply)?
Eh, marketing 5.3 GHz LPDDR5x, so 10.6MT/s. They tried to do the right thing, just the wrong wayView attachment 109850
CPU
• 'Prime' and 'Performance' cores. Interesting.
• 24 MB Total Cache: the article text clarifies that this is 12 MB L2 + 12 MB L2.
GPU
• New Sliced Architecture: the article text says Adreno 830 has 3 slices. So that's 3 slices of four CUs each, for a total of 12 CUs.
• Onboard GMEM. For DLSS-like upscaling?
• They are advertising it can run Unreal Engine, Nanite and Chaos Engine, which suggests that the Adreno 8 GPU architecture brings substantial improvements.
pre-a800 was a really bad uarch for running anything UE5.About UE, Nanite etc, Gen3 wasn't able to run them due to APIs or due to insufficient performance?
Key slides:Looks like the deep-dive is up (includes die shots even):
高通新旗舰来啦！骁龙8至尊版强不强？_哔哩哔哩_bilibili
大家久等了！今天高通终于发布了新一代旗舰手机芯片——骁龙8 Elite，中文名骁龙8至尊版。高通每每改名总是伴随着一些大事发生，那这次骁龙8 Elite自然不例外，作为高通近年来第一款自研CPU的产品，他到底是什么规格？能效有看头吗？我们先拿工程机前瞻一下吧……, 视频播放量 18089、弹幕量 428、点赞数 1962、投硬币枚数 716、收藏人数 283、转发人数 504, 视频作者 极客湾Geekerwan, 作者简介 爱折腾、超硬核的科技数码频道！商务合作请联系：bd@geekerwan.com，相关视频：纸面数据还是真王炸？高通骁龙X到底行不行？，【极客湾】笔记本又卡又慢？...www.bilibili.com
Big win on CPU in performance and efficiency, looks like a big improvement over what we saw from the X-Elite. GPU looks mostly on par with the 9400 (although benchmarks were done w/ slower ram), but is a touch less efficient. (I'd probably still go w/ Adreno due to better driver compatibility...)
Prime Core = Oryon-L = 192 KB L1i ?There’s 192KB L1 cache per Prime core and 128KB per Performance core.
So they are the same microarchitecture after all!We do know that the Prime and Performance cores share the same architecture, they are simply optimized for single and multi-threaded workloads based on clock speed, cache, etc..
Energy efficiency is finally what we expected from the CPU architecture with efficiency cores, and GPU efficiency support for modern rendering loads is finally there as well.Key slides:
View attachment 109898View attachment 109899View attachment 109900
Oryon-L and Oryon-M
View attachment 109901
CPU MT efficiency curve better than D9400 and A18 Pro.
View attachment 109902View attachment 109903
4 MB of cache per slice. Total of 12 MB
That's a lot of cache!
View attachment 109904
Neck and neck with D9400 in Steel Nomad Light
View attachment 109905
124 mm² for 8 Elite die.
The MT performance at iso power to the A18P is pretty solid I must say.
Maybe that’ll encourage Apple to consider two more e-cores.
Well they’ve said they won’t do things just to win benchmarks but then we’ve seen 18.1 beta 5 reverse at least the decline in benchmark results (how is up to interpretation). Theoretically they could have better performance at lower wattage as long as their e-cores are sufficient.Why? To win at benchmarks? Show me the real world smartphone code that is running all cores flat out for any length of time.
To be more efficient, E cores are workhorse of Soc in mobile, P cores are just for burst perfomance, 6 E cores would definetely help reducing power usage compare to 4 E cores.Why? To win at benchmarks? Show me the real world smartphone code that is running all cores flat out for any length of time.