News Intel Rocket Lake, Alder Lake CPU Photos Leaked Online

[Admin]

VideoCardz shares alleged CPU shots of Intel's future Rocket Lake and Alder Lake chips.

Intel Rocket Lake, Alder Lake CPU Photos Leaked Online : Read more

 

[everettfsargent]

"Slated for a first quarter release in 2021, Rocket Lake leverages the Cypress Cove microarchitecture, and Intel has already touted an instruction per cycle (IPC) gain up to double digits. ... While an exact timeframe is unknown, it's plausible that Alder Lake won't touch down until the second half of 2021. ... It's evident from the Alder Lake photograph that the processor will require a new socket."

So a very late Q1'21 launch for RL followed by a potential Q2'21 launch of AL on a new socket. Yeah right, that is s-o-o-o-o-o-o-o-o-o-o going to happen. In other words, it is very plausible that you don't have an bleeping clue. There is an Easter Bunny.

 

[nofanneeded]

Big little is only usable in mobile and not desktop . Intel should have gone three or four threads per core instead of this ,

 

[JayNor]

"Rocket Lake might finally bring PCIe 4.0 support to Intel's desktop platform..."

I think pcie4 has been confirmed by an Intel spokesperson already, so why the "might".

Alder Lake "might" be bringing pcie5 to a desktop platform before AMD, based on the leaks in notebookcheck.

When might AMD "finally bring" pcie4 to a laptop platform?

 

[JayNor]

Big little is only usable in mobile and not desktop . Intel should have gone three or four threads per core instead of this ,

Have to disagree. Intel stated the Lakefield small cores were more performant than multi-threading. In addition, Gracemont add avx2, and simd operations are normally not as performant in combination with multi-threading ... so the Gracemont simd contribution may be a nice win.

 

[InvalidError]

Big little is only usable in mobile and not desktop.

Seems perfectly fine to me: most programs have a handful of threads that require high performance and dozens of others that just do background or otherwise non-performance-critical stuff, so it makes sense to have some more power-efficient cores to take care of those and leave the high-performance cores unencumbered by low-priority tasks. The game I play most is WoW and ~90% of its CPU usage comes from three of its 30ish threads.

You also have all of Windows' background processes, all of your Chrome/Firefox/Edge/Opera/whatever tabs, all of your own background stuff and opened applications, etc. that could use low-power cores to leave performance cores available for games or whatever the foreground process is and its dependencies.

 

[everettfsargent]

So, in summary, two generations on LGA1200 and who knows how many generations on LGA1700 other than at least one. Meanwhile, AMD will have had four generations on AM4. Looks to me like someone told Intel to kill itself in the desktop space and that Intel is determined to do so. BTW who the heck does LGA1150, LGA1151, LDA1155 and LGA1156 anyways, four sockets for a total difference of six pins. Now you know why Intel continues to makes so much money ... dittoheads who have just enough money for annual CPU/MB upgrades. Oh look. I'm still on a 22nm Haswell as I decided to wait until 10mn arrived on the Intel desktop. Bought an AMD 3700X in the meantime.

 

[spongiemaster]

"Slated for a first quarter release in 2021, Rocket Lake leverages the Cypress Cove microarchitecture, and Intel has already touted an instruction per cycle (IPC) gain up to double digits. ... While an exact timeframe is unknown, it's plausible that Alder Lake won't touch down until the second half of 2021. ... It's evident from the Alder Lake photograph that the processor will require a new socket."

So a very late Q1'21 launch for RL followed by a potential Q2'21 launch of AL on a new socket. Yeah right, that is s-o-o-o-o-o-o-o-o-o-o going to happen. In other words, it is very plausible that you don't have an bleeping clue. There is an Easter Bunny.

Weird phrasing by the article. Alder Lake has never been rumored to launch in the 1st half of 2021. According to Intel, the platform will support DDR5 and PCIE5 which means it was never going to get released H1 of 2021. It will likely be late in 2021 to early 2022 when it is released.

 

[spongiemaster]

So, in summary, two generations on LGA1200 and who knows how many generations on LGA1700 other than at least one. Meanwhile, AMD will have had four generations on AM4. Looks to me like someone told Intel to kill itself in the desktop space and that Intel is determined to do so. BTW who the heck does LGA1150, LGA1151, LDA1155 and LGA1156 anyways, four sockets for a total difference of six pins. Now you know why Intel continues to makes so much money ... dittoheads who have just enough money for annual CPU/MB upgrades. Oh look. I'm still on a 22nm Haswell as I decided to wait until 10mn arrived on the Intel desktop. Bought an AMD 3700X in the meantime.

Intel has been doing 2 generations per socket for over a decade. It's not news, and not likely to change. Just because a CPU can be dropped into a socket doesn't mean it will work. As seen in the chart below there is not AM4 chipset that supports all 4 generations of Ryzen CPU's. The most any chipset supports is 3, with the B550 only supporting 2, just like Intel.

 

[everettfsargent]

Asus confirms AMD Ryzen 5000 support on X470, B450, B450 II

Asus confirms AMD Ryzen 5000 support on X470, B450, B450 II

Company forced to issue a clarifying statement after Product Support exec slip-up.

hexus.net

Oops!

 

[InvalidError]

So, in summary, two generations on LGA1200 and who knows how many generations on LGA1700 other than at least one.

LGA1156 is the only mainstream Intel socket from the last ~25 years to have supported only one CPU generation instead of two. Unless Intel makes a significant design mistake with LGA1700, it is practically guaranteed to be two generations as usual.

 

[Thunder64]

LGA1156 is the only mainstream Intel socket from the last ~25 years to have supported only one CPU generation instead of two. Unless Intel makes a significant design mistake with LGA1700, it is practically guaranteed to be two generations as usual.

Nice try. Socket 423. Besides, when did two generations become acceptable? How many people were upgrading from Sandy Bridge to Ivy Bridge, or Skylake to Kaby Lake?

AM4 has technically 5 generations. AM2 had many. You could go from a single core to a quad core. Intel wants to sell more chipsets to keep their fabs busy. AMD wants to sell chipsets as well, but they aren't as obvious about it.

 

[nofanneeded]

Seems perfectly fine to me: most programs have a handful of threads that require high performance and dozens of others that just do background or otherwise non-performance-critical stuff, so it makes sense to have some more power-efficient cores to take care of those and leave the high-performance cores unencumbered by low-priority tasks. The game I play most is WoW and ~90% of its CPU usage comes from three of its 30ish threads.

You also have all of Windows' background processes, all of your Chrome/Firefox/Edge/Opera/whatever tabs, all of your own background stuff and opened applications, etc. that could use low-power cores to leave performance cores available for games or whatever the foreground process is and its dependencies.

Why not go more threads per core instead of adding up heat and size from lower cores ?

I still think that moving to more threads per core is the better option.

 

[nofanneeded]

Have to disagree. Intel stated the Lakefield small cores were more performant than multi-threading. In addition, Gracemont add avx2, and simd operations are normally not as performant in combination with multi-threading ... so the Gracemont simd contribution may be a nice win.

you need to put in mind that the size and more power used for the little cores can be used to add one or two more of the Big core , add to it more threads thats for sure a better gain , because once you need more bigger cores for your apps the lower ones will not be used . unlike when all are big cores with multi thread

 

[InvalidError]

Why not go more threads per core instead of adding up heat and size from lower cores ?

I still think that moving to more threads per core is the better option.

For most consumer and office software, single-threaded performance is paramount to good user experience. Having more threads sharing a single core's execution resources hinders single-threaded performance, which is why lightly threaded games usually perform worse with SMT enabled when there are more than enough cores to afford disabling SMT.

If you want to go from SMT2 to SMT4 without losing single-threaded IPC from having four threads competing for access to execution units and other resources, you need to add enough execution units to accommodate shoving the same number of average instructions per thread as you previously did with two, which would require making most of the CPU core 50+% wider and in turn make it impossible to maintain the same clock frequencies simply due to signals needing more time to propagate, so you may lose something like 20% on clocks.

Power-efficient cores are simpler, smaller, use less power and don't bloat the high-performance cores like SMT4 would. SMT4/8 is something that only makes sense for applications where you can afford to sacrifice 20+% in single-threaded performance to gain 30+% throughput per-watt, per-area or other efficiency metric.

 

[shady28]

"Slated for a first quarter release in 2021, Rocket Lake leverages the Cypress Cove microarchitecture, and Intel has already touted an instruction per cycle (IPC) gain up to double digits. ... While an exact timeframe is unknown, it's plausible that Alder Lake won't touch down until the second half of 2021. ... It's evident from the Alder Lake photograph that the processor will require a new socket."

So a very late Q1'21 launch for RL followed by a potential Q2'21 launch of AL on a new socket. Yeah right, that is s-o-o-o-o-o-o-o-o-o-o going to happen. In other words, it is very plausible that you don't have an bleeping clue. There is an Easter Bunny.

Actually makes sense if you think about the situation Intel is in. Consider :

1. Intel doesn't have the 10nm capacity to meet demand for both mobile and desktop chips, not to mention server chips.
2. Intel wants to skip further 10nm capacity expansion and go to 7nm, starting very late 2021 for early 2022 launch(es).
3. This leaves Intel with a lot of 14nm capacity throughout 2021

So how do you release some new chips and stay relevant with that setup?

1. Have your basic desktop lineup on improved uArch using 14nm - aka "Rocket Lake"
2. Put your high end offering on 10nm Alder Lake shortly after the low end Rocket Lake line is released.
3. Keep most of your mobile high power and low cost laptops on 14nm Comet Lake
4. Release a 10nm Tiger Lake, but hobble it with only 4 cores and low power < 30W

This basically helps them stay relevant with Alder Lake and Tiger Lake, but tread water using enhanced architecture 14nm products too. I think the strategy is very risky as most consumers will be confused by this lineup, and when they buy a lesser chip than what they were expecting after reading say a Tiger Lake or possibly Alder Lake review, they'll feel cheated.

 

[Thunder64]

blah

Completely ignored my post? I guess because you didn't want to admit you were wrong?

 

[jkflipflop98]

OMG it has an IHS and pads on the other side?!? Shocking photos.

 

[everettfsargent]

Hey, let us make up a new rumor, right here, right now. LGA1700 will support 666 generations. Please remember, you heard that new rumor, here 1st.

 

[everettfsargent]

Actually makes sense if you think about the situation Intel is in.

I have and no amount of wordsmithing or dreaming will change the situation Intel is in.

 

[JayNor]

you need to put in mind that the size and more power used for the little cores can be used to add one or two more of the Big core , add to it more threads thats for sure a better gain , because once you need more bigger cores for your apps the lower ones will not be used . unlike when all are big cores with multi thread

Intel is already adding avx512 and dlboost to Alder Lake big cores, so the extra space you imagine may already be used.

By adding small cores with avx2 and no multi-threading, these may be equivalent to AMD avx2 for simd. Intel puts a lot of emphasis on AI processing, so this likely looks like a big win for their goals. AMD has yet to add any dlboost.

 

[JayNor]

This basically helps them stay relevant with Alder Lake and Tiger Lake, but tread water using enhanced architecture 14nm products too.

Rocket Lake doesn't look like a water tread to me. Looks like they want to continue dominating single thread with IPC update + 14nm boost clock + faster DDR + 20 lanes PCIE4 + improved GPU.

 

[Makaveli]

"Rocket Lake might finally bring PCIe 4.0 support to Intel's desktop platform..."

I think pcie4 has been confirmed by an Intel spokesperson already, so why the "might".

Alder Lake "might" be bringing pcie5 to a desktop platform before AMD, based on the leaks in notebookcheck.

When might AMD "finally bring" pcie4 to a laptop platform?

Laptops can stay on PCie 3.0 for now. Why do you want less battery life and more power consumption to support PCIe 4.0 in a laptop?

 

[everettfsargent]

Laptops can stay on PCie 3.0 for now. Why do you want less battery life and more power consumption to support PCIe 4.0 in a laptop?

Intel does what it wants! Apple left IBM PowerPC because they were too slow and too hot and too many watts. Apple is leaving Intel Core because they are too slow and too hot and too many watts for all of Apple's many, many toys. A PCIe 4.0 laptop, now that one is rich, with irony and sarcasm. I think Intel should do a PCIe 6.0 laptop, like yesterday even. And replace that battery with a V12 engine/25 gallon gas tank. Backport that process node!

 

[InvalidError]

Laptops can stay on PCie 3.0 for now. Why do you want less battery life and more power consumption to support PCIe 4.0 in a laptop?

One PCIe 4.0 lane uses less power and die area than two 3.0 ones. Intel may also have more power-efficient PCIe 4.0 PHY/MAC designs than AMD's - Intel has been doing high speed networking for a long time, developed high-speed transceivers for Thunderbolt for a while, acquired a portfolio of multi-gigabit transceivers by buying out Altera and probably a few others. Xilinx which AMD is currently in the process of acquiring likely has better PCIe 4.0/5.0 designs than AMD/TSMC too.