AMD討論POST(74) 睇#1(Updating ) &有時間幫手填資料

通常係走線 同 io問題

如果下代x板改單粒chipset 又會改善左io

問題係呢代b650，實際係x670，唔買r9就有啲浪費，但唔係用黎工作買黎係更加浪費

用asus X670 / B650板留意USB4 係 Thunderbolt 4

https://www.chiphell.com/forum.php?mod=viewthread&tid=2459636&extra=page%3D1&mobile=2

5600 $740左右
都算幾抵玩
amazon usd133-14-25(hsbc)

有budget 搵下5800x3d 😃😃😃

唔係一早知嘅咩

有諗過
但係佢次次平個時我都咁啱買唔到
尤其前兩日最平個時$329-25真係好吸引
不過諗諗下又無咩game玩
又齊哂console咁濟
我個mon又得2K144
平玩5600+6700XT住先算數

大家用ryzen master會揀auto oc all core定per core?定係係bios入面揀pbo2再揀curve optimizer ?

暫時只有簡單玩吓pbo2 only
pbo advanced settings
pbo power settings motherboard
scaler + freqency offset睇你想唔想拉高 (我懶得試所以1x + 0mhz就算)
最後curve optimizer negative (all core / per core 都得但要自己試)
我而家只係頭 8 粒 -10 之後CCD1尾8粒就 -5 ~ -8

超ram搞吓timings可能仲比較好

https://videocardz.com/newz/intel-reportedly-working-on-raptor-lake-refresh-amd-ryzen-7000x3d-might-be-limited-to-8-cores-for-now
得八核嘅7000X3D......

3D cache家用應用主要得gaming有提升
雙CCD結構搞到16核對打機都冇咩用(ccd to ccd延遲高)
淨係得單CCD先上X3D都合理

Zen5要24H2先有有啲驚訝，估唔到咁遲

即係頂盡得7800X3D ... 琴日仲喺度諗緊7950X3D條List , 又要失望了 .

單CCD最大問題係用唔盡DDR5嘅高頻寬 , 更新咗GMI去到3都係大概差唔多得35GB/s

好似有7600X3D?

佢roadmap仲有Zen4c 未出

只能夠等zen5或以後睇吓會唔會上advanced packaging
互連唔再走銅線應該可以提高埋FCLK上限

初初以為Epyc發佈會會出曬zen4 zen4x3d zen4c

點知得zen4，又拖長嚟出

zen4c唔知會唔會有desktop產品

如果整個8C z4 CCD+12C z4c CCD感覺上好似已經可以打番牙膏大小核設計轉頭

咁都係V-cache 就係用黎減低對main memory bus 既依賴既

公道講如果出齊全set X3D
non-3D真係唔使賣
期望zen4+/zen5改做全面大cache

AMD is expected to debut its entry-level A620 motherboard chipset in Q2-2023. This chipset reportedly lacks CPU overclocking capability, is expected to lack PCIe Gen 5, and caps memory speed to DDR5-4800.

ram都鎖頻唔出好過出

睇番中文原文，係咪指鎖ram我覺得可圈可點喎

AMD A620 可能要等到 Q2 了，依舊鎖 CPU OC，然後在稍晚一點會有桌面的 AM5 APU，不過內存只會有 DDR5 JEDEC 4800，不排除會有強化型號，目前已知 8C 6C 兩個型號，不過沒有零售的消息。

啲CPU一向有個官方支援DDR speed規格，我覺得只係指個官方規格係4800

同埋就咁睇呢個AM5 APU

可能只係Rembrandt (6nm RDNA2 & Zen3+)

Rembrandt嘅官方DDR5 IMC規格正正就係4800

同埋佢話沒有零售消息 姐係得Pro版

但Ryzen 4xxxG Pro好似都可以超

長遠可能會發展成base die係一大塊L3 cache，喺上面再放CCD IOD GCD etc

出年發佈嘅MI300應該會睇到AMD嘅3D stacking方向係點


之前MLID定RGT有吹過，zen5會改成CCD只有共享L2，L3另外一塊疊喺CCD上/下

不過呢兩條友啲amd料而家睇可以收收皮無可信性


同埋年中分析日roadmap zen5仲喺有分做zen5同zen5 v-cache

某程度上可以推測zen5喺晶片互聯上未必會有升級

Cerebras Systems, the pioneer in accelerating artificial intelligence (AI) compute, today unveiled Andromeda, a 13.5 million core AI supercomputer, now available and being used for commercial and academic work. Built with a cluster of 16 Cerebras CS-2 systems and leveraging Cerebras MemoryX and SwarmX technologies, Andromeda delivers more than 1 Exaflop of AI compute and 120 Petaflops of dense compute at 16-bit half precision. It is the only AI supercomputer to ever demonstrate near-perfect linear scaling on large language model workloads relying on simple data parallelism alone.

With more than 13.5 million AI-optimized compute cores and fed by 18,176 3rd Gen AMD EPYC processors, Andromeda features more cores than 1,953 Nvidia A100 GPUs and 1.6 times as many cores as the largest supercomputer in the world, Frontier, which has 8.7 million cores. Unlike any known GPU-based cluster, Andromeda delivers near-perfect scaling via simple data parallelism across GPT-class large language models, including GPT-3, GPT-J and GPT-NeoX.

一部超級電腦就食咗AMD萬八粒上代EPYC

睇真啲發覺班press應該錯寫18,176 cores 做 18,176 processors
睇返佢個介紹網實情只係用咗284粒64 core zen3 EPYC