Soldato
Like all previous generations of DDR don't fit in each other sockets... how is this news to people?
-
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.
Alder Lake-S leaks
- Thread starter Grim5
- Start date
Like all previous generations of DDR don't fit in each other sockets... how is this news to people?
your thinking in pre pandemic terms the global supply chain is screwed atm
everything I sell is going up month on month with massive shortages which will impact everything we buy as end users a shortage of lorry drivers is even worse if they have a shortage or lorrys and trailers to move it all in
and its only going to get worse the US goverment is warning of massive shortages leading upto christmas on food and consumer products is just one example
This clears up any remaining ambiguity
Intel has released a comprehensive guide for developers for Alder Lake. it's quite detailed and explains exactly how the architecture works, it goes into a lot of detail about how big.little works and how Thread director works
https://www.intel.com/content/www/us/en/developer/articles/guide/alder-lake-developer-guide.html
Interestingly, it mentions that if software is not optimized for big.little, that Thread director can sometimes incorrectly allocate an important process or task to one of the Little cores, resulting in lower performance.
Intel says this issue is very common in software that uses middle ware - such as a videogame that has built in DRM, lol. Intel suggests the developer asks the DRM maker to optimise their DRM for Thread director to avoid a situation where a game has lower performance because of it
Intel has released a comprehensive guide for developers for Alder Lake. it's quite detailed and explains exactly how the architecture works, it goes into a lot of detail about how big.little works and how Thread director works
https://www.intel.com/content/www/us/en/developer/articles/guide/alder-lake-developer-guide.html
Interestingly, it mentions that if software is not optimized for big.little, that Thread director can sometimes incorrectly allocate an important process or task to one of the Little cores, resulting in lower performance.
Intel says this issue is very common in software that uses middle ware - such as a videogame that has built in DRM, lol. Intel suggests the developer asks the DRM maker to optimise their DRM for Thread director to avoid a situation where a game has lower performance because of it
Last edited:
Also unlike the leaks have being saying, AVX-512 is NOT fused off the board.
Alder Lake does indeed support AVX-512, Dave2150 can rejoice! But it's only supported on the Big cores. If the system boots up and little cores are enabled, AVX 512 is blocked from running. AVX512 is only enabled when the little cores are disabled in the Bios.
Alder Lake does indeed support AVX-512, Dave2150 can rejoice! But it's only supported on the Big cores. If the system boots up and little cores are enabled, AVX 512 is blocked from running. AVX512 is only enabled when the little cores are disabled in the Bios.
Let the hand holding and finger pointing begin.
Caporegime
Intel haven't got the power envelope. They already hit 300W with an 8 core CPU.
Yeah the more I'm reading the more it's sounding like the first thing gamers will want to do on alder lake is head into the Bios and disable the little cores and overclock the big cores.
Doing this seems like it will ensure more consistent performance and likely higher gaming performance anyway since games don't yet need more than 8 cores. The only downside I can see is the potential loss of multi tasking performance as those little cores could used to run background tasks like YouTube, Spotify, screen recording, streaming etc - maybe after a few months the issues will be cleared up and disabling the little cores won't be needed anymore unless you must have AVX512.
For AMD, leaks suggest they are also planning to move to Big.Little for Zen5, so it can be a case of Alder Lake being the beta test - and software developers will spend the next year or so optimizing their software for Big.Little and then Raptor Lake and Zen 5 arrive all the bugs and niggles are worked out
Last edited:
Soldato
A hard one to call. keeping in mind the 5950x already has 16 full cores and already wipes the floor with anything Intel has at the moment. The question in my mind is, although AMD are planning Big.Little..........................just how many Big cores are they planning. Because clearly with ADL Intel are not exactly overstretching the core count that's for shure, and i think all of us know even with ADL 300w is almost a certainty.
I wouldn't be overly surprised if the Zen5 = big.LITTLE is 'just' rumour, or limited to mobile where it actually makes sense...
It increasingly seems like Alder Lake is big.LITTLE because Intel just can't meet the core counts with just big/performance cores inside any kind of reasonable power envelope (as if 300W is reasonable...), not really because it's the 'right' architectural decision (for desktop at least).
It increasingly seems like Alder Lake is big.LITTLE because Intel just can't meet the core counts with just big/performance cores inside any kind of reasonable power envelope (as if 300W is reasonable...), not really because it's the 'right' architectural decision (for desktop at least).
@SKILL - I agree, why would AMD shoot themselves in the foot by implementing low power, lower spec cores, in addition to the existing large cores?
Surely its more efficient to stick to incremental large core count increases, when power usage allows?
I think going beyond 16 cores (big or small), AMD will very quickly run into diminishing returns, what would be the point for most consumers, when most games and applications struggle to utilize 70-80% of 8 CPU cores?
In general, I think higher clocks, cache per core and IPC are still the main factors that determine CPU performance, not more cores.
With higher transistor density, AMD should be able to beef up those cores without massively increasing their size and power usage.
Surely its more efficient to stick to incremental large core count increases, when power usage allows?
I think going beyond 16 cores (big or small), AMD will very quickly run into diminishing returns, what would be the point for most consumers, when most games and applications struggle to utilize 70-80% of 8 CPU cores?
In general, I think higher clocks, cache per core and IPC are still the main factors that determine CPU performance, not more cores.
With higher transistor density, AMD should be able to beef up those cores without massively increasing their size and power usage.
Last edited:
Caporegime
Build it and they will come, if games only need 8 cores and therefore 8 cores is all need to build games will forever only need 8 cores.
Much like games only need 4 cores, forever, until they didn't right?
Much like games only need 4 cores, forever, until they didn't right?
Isn't that what Intel are doing though by focusing on 8 faster high IPC cores for stuff like gaming while adding the small efficient cores for productivity.
So long as the 8 cores speed keeps increasing by a decent margin every generation then 8 cores will be enough, the reason quads started struggling was due to the cores not getting much faster over many years.
AMD and Intel only started boosting core counts when they couldn't increase the transistor density fast enough between generations. At least, that's how I remember it (this started happening many years ago now)...
In other words, Moore's Law was starting to falter, as the laws of physics became more of an obstacle.
We'll see I suppose. How smooth will the transition to TSMC's 5nm EUV process be? And Intel's 7nm EUV probably by 2023? And beyond that? It seems unknowable
In other words, Moore's Law was starting to falter, as the laws of physics became more of an obstacle.
We'll see I suppose. How smooth will the transition to TSMC's 5nm EUV process be? And Intel's 7nm EUV probably by 2023? And beyond that? It seems unknowable
Last edited:
Soldato
Well, it would be if you can run the whole lot at the same time and still be up there when gaming. Only time, benchies and results will tell us that.
Absolute rubbish, quads have got to well over 5Ghz for years now and havn't got above that..........................i know i'd much rather be running 16 cores at 4.6Ghz than 4 or 8 cores at 5Ghz
Caporegime
Of course but you can't get the biggest leap in performance from Mhz and IPC alone, i can envisage a future of 100 core CPU's, and soon, i wouldn't be surprised if AMD go to 96 cores with Zen 4, 200 core CPU's, 300 core CPU's, hell why not? again if you don't do it no one will care and that's how you get stagnation.
Caporegime
They are not meant for us mare mortals, yet, but it trickles down eventually, you have a 12? 16 core CPU? For less than an 8 core would have cost you just 4 years ago...
We also have the lowest Sku's costing almost as much as the flagships did 5 years ago.
It's the IPC which stagnated so quads fell behind, if Intel had brought 20% IPC every generation for the last 10 years then quads would still be fine.
Caporegime
A £350 5800Xis a much better CPU that a 6700K, or 7700K or 8700K.
Its progress. something we have not had for a decade.