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 : Advanced Micro Devices - Moderated (AMD)
AMD 231.83+1.7%Jan 16 9:30 AM EST
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: wbmw who wrote (252914)6/5/2008 11:43:37 PM
From: pgerassiRead Replies (1) of 275872
 
Wbmw:

BS! The memory controller is on die. What you are saying is that QPI can screw up that badly memory connected on the same die as the cores, then QPI is going to be a true disaster. And multi socket performance is going to be far worse. And don't even talk about reliability, because it needs a perfect MB and perfect DIMMs to work. Perfect balance is required with Nehalem, no mixing allowed at all. One slightly different anything will make it operate like crap.

Frankly as bad as I think Intel gets, even they would not be that dumb. It is far more likely that the prime cause is the individual CPU dies themselves. Some samples were not as good as others. And that isn't unusual at this stage. It will take a few respins to get a reliable yielding CPU.

Look at Opteron or A64 MBs. There is very little variation in the memory performance even between the various grades and manufacturers of MBs. That is true of a robust design that has had a few years to work out the problems. There is more variation on Intel based systems, because the chipsets, FSB layouts and memory configurations are so varied.

Now if Nehalem used off die chipset memory controllers, the QPI and MB quality would be a bigger effect. Why would Intel do this since ODMC is the real reason for QPI? Are you implying that the big bad Nehalem's ODMCs are crap and they are using the chipset memory controllers instead? That would be a huge disaster.

Pete
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2026 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News