Discussion in 'Frontpage news' started by Hilbert Hagedoorn, May 22, 2020.
Just on nand density and perf/watt ...... nothing important for CPU's right ?
Hmmm, another 14nm from Intel? Not a fan.
Wow, if this is true then Rocket Lake is DOA.
This guy looks at a rumored engineering sample of one model uploaded to a website for a product not coming out till maybe the end of the year and decides its DOA. I love the internet.
That's why I am still stuck at the 8 th gen . But i am just itching for new toy !
At least we are getting more cores now , the same cores for many years but more , thats good i was sick of getting the same 4 core parts with a small increase in clocks sold as a new gen .
He started in 2018? if they design and implement a new architecture in 3 years it will be a first.
Care to show me where any engineering sample ever differed dramatically from the final product six months away? Besides, I said "if this is true", so I really didn't decided anything there, only speculated.
Again, if this is true, then it means Rocket Lake won't have any meaningful IPC increase compared to Comet, and considering the 14nm process is already on its limits, either it has to be a lot more power efficient (which is much needed and maybe not possible), or it will be just a rebrand... of a rebrand, of a rebrand, of a rebrand, of a rebrand (yes, five times already). So, DOA, especially if ryzen 4000 shows up at the same time frame.
Where's the "Intel"? / You have the "Intel"?....in movies..though in REAL LIFE also...... you can hear these....
but i've never heard Where's the "AMD"?
Rocket Lake is based on Willow Cove, which is improved on Sunny Cove. Sunny Cove was an increase in IPC of around 15% over Skylake. And Willow Cove will have another IPC increase over Sunny Cove. So yes, Rocket Lake WILL have IPC increase over Comet Lake. My guess would be a 15-20% IPC increase over Comet Lake. And remember, IPC increases everything, single and multi-thread. Rocket Lake may "still be 14nm" but it is not going to just be a simple refresh like we've been seeing. Increased IPC, cache redesign, PCIE 4.0 to name a few.
And best not to believe Zen 3 rumours, as we all know how well that went when people kept saying Ryzen would do 5ghz.
Well first you have to define what a new architecture even is. Zen for example wasn't a completely new architecture - the entire front end was borrowed from Bulldozer, large parts of the FPU were copied from FlexFPU, bunch of other parts were similar - it was definitely a radically shift though and took roughly 4 or so years from design to implementation (End of 2012 to early 2017) but AMD was actually demonstrating it in August 2016.
Golden Cove isn't rumored to be a complete top to bottom redesign either - it's just supposedly going to have very large IPC increases. Alderlake (what Golden Cove is shipping in) is rumored to launch end of 2021, early 2022 - so roughly 4 years after Keller took his position.
So if you're talking about a radical architecture shift - then yeah, I think Golden Cove is the first one done entirely by Keller at Intel. If you're talking about a complete redesign, top to bottom - I'm not sure that would ever actually happen or ever did happen. Even Pentium to Core shared huge swaths of the same underpinnings. Same from Bulldozer to Zen.
Where did you see reported that Rocket Lake would have 15-20% IPC gain? Haven't seen anything like that on typical news sites, including this one.
And regarding Zen, they have been consistent on delivering what has been promised, or very close. People dreaming of 5Ghz were delusional and that was never even hinted at by AMD. But their IPC gains have been very real, to the point where clock for clock they're already faster than intel's sky/endlesslake by around 10%. So yeah, I would bet on them, as they seem to have a good hand right now.
clock for clock isn't the ultimate measure when you have one cpu clock at 4.5ghz and the other at well over 5ghz.
Clock for clock means 1hz vs 1hz. Or 4.5GHz vs 4.5GHz. Not 5 vs 4.5.
Errm, yes, we all know that. I'm not even comparing intel to AMD here, it's more an intel 10th gen vs 11th gen and how the latter will be pointless for the following reasons:
1) 14nm process from intel is very likely at its peak performance, due to age and laws of physic;
2) due to 1, clocks will probably not go too far from where they are now - single core 5,3ghz seems to be the practical hard wall of intel's 14nm with lakes architecture, as seen from overclocks from even 6th and 7th gens;
3) considering that 5,3ghz already spends a LOT of energy and needs a LOT of cooling, intel can't go any further in that department, which means...
4) IPC increase is their only option to gain performance in 14nm now, or else 11th gen will be pointless / the same as 10th gen plus pci-e 4
That is already a bleak scenario, but let's put AMD in the mix now:
1) Zen 3 is being done in TSMC's N7P process, which in itself brings a confirmed +7% performance gain or 10% reduction in power consumption when compared with the process used to make Zen 2 - we can safely guess which option AMD will pick due to Zen 2 already being very power efficient;
2) On top of that, AMD claims "typical increase in IPC from a generational leap", which industry common sense has translated to 15~20% gain in IPC. Even if that is not true and the IPC increase is in a very conservative 10% range, that still puts a lot of performance into those processors when combined with 1
3) All of that will supposedly launch at the same time frame as intel's 11th gen, meaning that intel won't have any other response for at least another 6 months
TLDR: I wouldn't bet on intel now.