Personal computing discussed
Moderators: renee, Flying Fox, morphine
chuckula wrote:But you get 12.
AND YOU BETTER FREAKIN' LIKE IT
f0d wrote:looks like adoredtv got everything wrong even just days before this announcement
Waco wrote:Damn. Impressive announcement...but the 3950X is going to be an insta-buy for me at 16 cores. You know they will, the chiplets have the cores...
I'm going to impatiently await benchmarks of the consumer chips. I hope to have Rome into a few testbeds in the next couple months, but those will be gated by NDAs expiring for releasing results.
jihadjoe wrote:Desktop maxing out at 12C strongly hints that AMD is saving all of the 8-core chiplets for EPYC. I bet the desktop 8C are using 4+4.
jihadjoe wrote:Desktop maxing out at 12C strongly hints that AMD is saving all of the 8-core chiplets for EPYC. I bet the desktop 8C are using 4+4.
chuckula wrote:Waco wrote:Damn. Impressive announcement...but the 3950X is going to be an insta-buy for me at 16 cores. You know they will, the chiplets have the cores...
I'm going to impatiently await benchmarks of the consumer chips. I hope to have Rome into a few testbeds in the next couple months, but those will be gated by NDAs expiring for releasing results.
To be perfectly blunt: When a regular 16-coar Threadripper is running at $500, why do you want to pay more money for a desktop board that cuts off half of the memory channels and memory capacity??
If you're so in love with AMD, help them out by by buying a 16 core threadripper and using it, because these little chiplets are more for show than for go.
Bonus: Here's Xeon AP beating a 128-coar Epyc system in literally the exact same NAMD demo that AMD showed off on stage at Computex: https://www.youtube.com/watch?v=eCsIETb ... e=youtu.be
You might say "But Inte's 56 coar Xeons are too expensive!" Well for one thing: No they aren't for the customers who matter. For another thing: I never said that was a 56 coar Xeon AP platform because it wasn't. It was a two-socket 96 coar platform beating AMD's monstrosity that has a full 33% more coars in the exact benchmark that AMD cherry picked for its demo.
When Intel launches Sapphire Rapids, I fully expect everyone here to be overjoyed if a Sapphire Rapids part with 33% MORE coarz than an older AMD part still manages to lose in the exact same tests that Intel shows off on stage at the launch. Fair is fair.
sconesy wrote:...and that AMD is now the certified IPC champ.
chuckula wrote:f0d wrote:looks like adoredtv got everything wrong even just days before this announcement
I wish I had the cult-like following that Jim at AdoreTV had. Maybe I need to 1. Get a Youtube channel. 2. Turn off my brain. and... 3. Be overwhelmingly wrong over any over again to the point where any rational person would know I'm just making stuff up with no sources!!
Krogoth wrote:chuckula wrote:Waco wrote:Damn. Impressive announcement...but the 3950X is going to be an insta-buy for me at 16 cores. You know they will, the chiplets have the cores...
I'm going to impatiently await benchmarks of the consumer chips. I hope to have Rome into a few testbeds in the next couple months, but those will be gated by NDAs expiring for releasing results.
To be perfectly blunt: When a regular 16-coar Threadripper is running at $500, why do you want to pay more money for a desktop board that cuts off half of the memory channels and memory capacity??
If you're so in love with AMD, help them out by by buying a 16 core threadripper and using it, because these little chiplets are more for show than for go.
Bonus: Here's Xeon AP beating a 128-coar Epyc system in literally the exact same NAMD demo that AMD showed off on stage at Computex: https://www.youtube.com/watch?v=eCsIETb ... e=youtu.be
You might say "But Inte's 56 coar Xeons are too expensive!" Well for one thing: No they aren't for the customers who matter. For another thing: I never said that was a 56 coar Xeon AP platform because it wasn't. It was a two-socket 96 coar platform beating AMD's monstrosity that has a full 33% more coars in the exact benchmark that AMD cherry picked for its demo.
When Intel launches Sapphire Rapids, I fully expect everyone here to be overjoyed if a Sapphire Rapids part with 33% MORE coarz than an older AMD part still manages to lose in the exact same tests that Intel shows off on stage at the launch. Fair is fair.
Zen-2 parts have higher clockspeeds and have access to PCIe 4.0 if you really need it. Otherwise, you can wait for Threadripper 3 if you want Zen2 on the HEDT-side. It is part of the reason why AMD is withholding their 16-core desktop Zen2 SKUs.
Power efficiency is much more important metric in the HPC world. Intel has an uphill battle here without their historical foundry advantage. They need move away from monolithic designs ASAP but again they have been slowly doing that since Broadwell-E/Broadwell-EP. The Lake dynasty is at wits end. What were seeing is a repeat of Netburst dyantsy's end days and emergence of the K8.
chuckula wrote:But you get 12.
AND YOU BETTER FREAKIN' LIKE IT
https://www.anandtech.com/show/14407/am ... -coming-77
LoneWolf15 wrote:Although I'd say that Netburst was a far less efficient architecture, so the analogy isn't near as valid. Lake could scale further IMO if Intel could have gotten their die process ducks in a row, but that killed them. The vulnerabilities in their SMT aren't helping either; it looks like there needs to be a real change in how this is done.
The IPC will be the most interesting thing to see from AMD, IMO. For them to take the crown there will be impressive indeed.
Redocbew wrote:LoneWolf15 wrote:Although I'd say that Netburst was a far less efficient architecture, so the analogy isn't near as valid. Lake could scale further IMO if Intel could have gotten their die process ducks in a row, but that killed them. The vulnerabilities in their SMT aren't helping either; it looks like there needs to be a real change in how this is done.
The IPC will be the most interesting thing to see from AMD, IMO. For them to take the crown there will be impressive indeed.
Agreed. I wonder how useful of a metric IPC still is though. It's never been a static figure that always stays the same. Some average value is the best we can do, and now with op-caches, op-fusions, skipping of decoders, and differing numbers of floating point and/or integer instructions accepted for each clock the wiggleroom for this metric has probably increased possibly by quite a lot. It's outside my field, but it always makes me wonder when there's a new announcement like this, and everyone goes bonkers over IPC even though nobody knows what it is.