Do they have HW security fixes ?

HWgeek:

Do they have HW security fixes ?

I'd be curious too. Not that I'd buy a Xeon, but it's about time they work this out. Well, the fixes from 2018 that is, new ones will probably be tackled 2020 if it's going to be that fast even. And the year isn't over yet...

HWgeek:

Do they have HW security fixes ?

You believe in Santa Claus? Probably 14nm+++++++++++++++++ again. All security issues compatible. And i didnt read any info about them.

That cache seems pretty small for a Xeon.

schmidtbag:

That cache seems pretty small for a Xeon.

R9 3900X 70mb L3. 😀

karma777police:

Not really.

Yes... it really is. That's less than 2MB of cache per core, which is sparse even for modern consumer-grade CPUs. There are Xeons from over 10 years ago that have more cache per core than this; there are Xeons from that far back which have proportionately more L2 cache. There are plenty of other Xeons made within the past year that have proportionately a lot more cache. So, no matter what way you look at it, for a product that will cost several thousand dollars, this 28-core Xeon is objectively low on L3 cache.

Intel... CREED...

karma777police:

They are not, these are on socket 3467 design a bit different and they all have 6 memory channel.

Do you not spend more than 5 seconds thinking about what you're trying to say, or are you really just that arrogant? Since when have caches (especially L3 caches) been known to shrink this dramatically over time? It's not like Intel has done anything drastic to their architecture in the past 5 years. Besides, the W-3223 (a CPU mentioned on the same table in the article) has over 2MB per core, which is to be expected. That CPU's design isn't really any different, it just has fewer cores. So, that's already breaking your logic. Surely you wouldn't think that CPU has too much cache... Meanwhile, how exactly does the socket or amount of memory channels have anything to do with the amount of cache? The W-3223 also supports 6 channels, and yet that has more cache than the W3275[M], which again, breaks your logic. To reiterate: we're talking about a multi-thousand dollar product. If this were an i3, I probably wouldn't have even paid attention to the cache.

HWgeek:

Do they have HW security fixes ?

they have ltf1 , meltdown, with spectre hardening for sure, For mds its fixed on newer steppings, but it remains to be seen if there intel will be shipping old steppings for these new chips, like they did with the other cascadelake xeons , To see which steppings are effected you can look here https://software.intel.com/security-software-guidance/insights/deep-dive-cpuid-enumeration-and-architectural-msrs#MDS-CPUID

user1:

they have ltf1 , meltdown, with spectre hardening for sure, For mds its fixed on newer steppings, but it remains to be seen if there intel will be shipping old steppings for these new chips, like they did with the other cascadelake xeons , To see which steppings are effected you can look here https://software.intel.com/security-software-guidance/insights/deep-dive-cpuid-enumeration-and-architectural-msrs#MDS-CPUID

I wouldnt count on it, ever, as they have lied before.

Intel offers 28 cores Xeons..yawn. to me, it feels like the kid who shows off the new shiny ball, just to let him play in the team after the got expelled. for bad behavior. my 2 sour-witty cents

schmidtbag:

Do you not spend more than 5 seconds thinking about what you're trying to say, or are you really just that arrogant? Since when have caches (especially L3 caches) been known to shrink this dramatically over time? It's not like Intel has done anything drastic to their architecture in the past 5 years. Besides, the W-3223 (a CPU mentioned on the same table in the article) has over 2MB per core, which is to be expected. That CPU's design isn't really any different, it just has fewer cores. So, that's already breaking your logic. Surely you wouldn't think that CPU has too much cache... Meanwhile, how exactly does the socket or amount of memory channels have anything to do with the amount of cache? The W-3223 also supports 6 channels, and yet that has more cache than the W3275[M], which again, breaks your logic. To reiterate: we're talking about a multi-thousand dollar product. If this were an i3, I probably wouldn't have even paid attention to the cache.

Actually, my dual xeons from circa 2015 already have 90mb of total L3, it does help to improve performance, especially once turbo hacked, and the efi driver peggs the SA/ cache frequency @3.00 Ghz