If it goes for the prices there I think they are HIGHLY possible. Especialy 8c 16t cpu. That would be only 1 CCX not 2 so cheap as chips to produce also I think it will be best gaming option and highest speed on ddr and core itself. Less problems with GLUE tm when only 1ccx is used
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***
- Thread starter Jamin280672
- Start date
If it follows Epyc design, as you would expect, you get more glue as the memory controllers are off-die... Pretty sure Zen 3 will be an improvement though, question is how much and how hard are Intel going to spin
But should be better than RAMBUS IMC in zen and zen+If it follows Epyc design, as you would expect, you get more glue as the memory controllers are off-die... Pretty sure Zen 3 will be an improvement though, question is how much and how hard are Intel going to spin
If it follows Epyc design, as you would expect, you get more glue as the memory controllers are off-die... Pretty sure Zen 3 will be an improvement though, question is how much and how hard are Intel going to spin
I think the guys at Intel sniff glue on the daily.
It won't be 1 chiplet, it'll be 2 chiplets with 4 cores disabled in each. All of the fully-functioning, 8-core chiplets will be used in EPYC Rome. This also precludes a 16-core Ryzen 3000 too, unless AMD do a limited Black Edition and charge a hefty whack for it to recoup revenue lost by not making an EPYC Rome.
This of course assumes Ryzen 3000 will have the same independent I/O controller design as EPYC.
^^ I don't see them treading on TR with desktop Ryzen, 8 cores are plenty for DT for a while yet. And it'll be either one core chiplet and one IO die or two chiplet die with fused cores and the IO die, unless they design a desktop specific die, which goes against the Zen philosophy so far...
This is my issue.
These type of slots work well in confined spaces, like laptops and NUC devices.
But they horrible in a desktop pc, awkward to access, drives get overheated and as you mentioned limited PCB space. I think pci-e NVME drives are far superior in that the slots and space already exist, cooling will be better, and easier installation. But the problem is also as you said that the drive manufacturers would need to make the drives in pcie format when they already make in this format.
So to me a compromise solution could be used. You buy a pcie card where you attach the m.2 drive to. Also this means the valuable lanes can be fully routed to pcie slots instead of shared to m.2 ports that most people wont use.
https://en.wikipedia.org/wiki/U.2
Yeah U2 is deffo superior to m.2 as well, they basically chose the worst solution out of the 3 for me.
Ill put some coal in to the Hype Train
intel are dead and buried with that line up. 5ghz cpus, tons of cores, everyone is a winner. Probably explains the crazy intel 9 pricing, intel going out with a last bang.
So you link to a Youtube video, with the absolutely worst sort of guessing using an "unnamed source" to provide the information which if things don't turn out as stated he just states he was mislead by the so called source etc. His technical knowledge with regards to power delivery is also terrible, and lends zero credibility towards his information. Why do people even remotely believe this sort of tripe?
Is he GavINTEL in disguise sent to troll AMD
Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,336
- Location
- Planet Earth
I hope my ASUS ROG Strix B450-I is fine as it is a £160 motherboard,and almost identical to the X470 version,and has a decentish 6+1 phase VRM.
I hope there is Black edition top clocker. I paid extra 80 pounds for silicone lottery 2700x for extra 100mhz. I'm willing to pay extra 100 over normal chip for better clocker.
GIEV BLACK EDITION !!!!
Two separate cut down single CCX chips and separate I/O chip would be serious double mess for desktop performance:
- Jump between two CCXes has performance penalty for communication between threads already in single die.
So having cores on two separate dies would be even bigger challenge for minimizing that penalty. - Also moving memory controller to separate I/O die is challenge to core-IMC latency and hence memory latency...
Which is very important in desktop use and already currently behind Intel!
And more dependant on memory bandwidth than best latency.
But do both on desktop and it would be lots of work to prevent gaming performance from dropping!
Separate core chiplet and I/O die would be challenge enough.
Putting those crippled and halved CPU core chiplets to some console MCM would be more sensible, than giving Intel "free ammunition" to use against AMD on desktop.
I also think so ^^^^
- Joined
- 12 Feb 2014
- Posts
- 2,997
- Location
- Somewhere Only We Know
Ill put some coal in to the Hype Train
If this turns out to be true, id defo pay £500 for that black edition 3800X, 16 cores, 32 threads boosting to 5ghz nearly...……..Mommy, may I ?
As for fake, pretty sure it is, no one knows yet except for AMD.
If this turns out to be true, id defo pay £500 for that black edition 3800X, 16 cores, 32 threads boosting to 5ghz nearly...……..Mommy, may I ?
As for fake, pretty sure it is, no one knows yet except for AMD.
There are still 7 months in which some characteristics of the process will be improved, so AMD can't know so early neither the frequencies which should be at the to-be-yet-decided phase, nor the pricing.