SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Qualcomm Moderated Thread - please read rules before posting
QCOM 178.63-1.3%Oct 29 3:59 PM EDT
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: GR8FORM who wrote (196161)10/26/2025 11:41:35 AM
From: QCOM_HYPE_TRAIN20 Recommendations

Recommended By
abcs
BDAZZ
Bill Wolf
Dr. John
GR8FORM

and 15 more members

  Read Replies (4) of 196311
 
One thing that may not be obvious to the non-technical folks here is the difference in memory between the AI200 and say all other products.

AI200 uses LPDDR (usually used in phones & laptops) while other DC products like the newly release AMD MI355X use HBM3E.

HBM memory is very fast, very expensive, very power hungry and very hot. Many of these other GPUs have 'limited' memory:

AMD MI355X 288 GB
Nvidia GB300 279 GB

As you can see Qualcomm is shipping a product with 2.6x the memory. This is important for inference as you can pack more models on this card, or have future models fit on your older cards. They didn't mention memory throughput, but I would expect it to be competitive with the competition. Anything wildly slower would probably be a non-starter for anyone looking to buy these.

Then with AI250 with their "innovative memory architecture based on near-memory computing which Qualcomm said will enable 10 times higher effective memory bandwidth and lower power consumption." If I read this correctly it means these cards will have faster significantly faster BW than HBM, with lower power, lower cost, lower heat. Truly a generational leap.

Along with that you better believe QC has patented every single breakthrough with that "innovative memory architecture". Either QC license it for tons of money to AMD and Nvidia or we keep everything close to the chest in hopes of creating a moat for inference servers. We'll see if they can bring this innovative memory to edge devices as well.
S
B of HBM3E
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2025 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News