As we approach the rumored November 4 announcement for Alder Lake, more information about the hybrid chips has come to light. On this occasion, a Chinese netizen (via momomo_us (opens in new tab)) has shared some insights on Intel's next-generation platform. As with any leak, it’s prudent to exercise caution when approaching the information.
If the rumors are accurate, the Core i5, Core i7 and Core i9 K/KF-series will lead the charge for the Alder Lake army. Since these are unlocked SKUs, motherboard vendors will launch their Z690 products first. According to a BiliBili user (opens in new tab), we can expect the non-K models and the more accessible B660 and H610 motherboards to arrive in January 2021.
Alder Lake purportedly has good headroom for overclocking. The leaker mentioned that a 5.2 GHz to 5.3 GHz all-core boost clock is available if you turn off the Gracemont cores. Heat dissipation may be decent as well thanks to the large integrated heat spreader (IHS). Outside of improving overclocking margins, disabling the Gracemont cores doesn't help lower power consumption, though.
FIVR's Return (Kinda)
As per the tipster's information, Alder Lake's V-Core and Core GT (iGPU) voltages stick to the normal scheme and only the Uncore conforms to Intel's FIVR (Fully Integrated Voltage Regulator) design. Intel seemingly combined the VCCSA and VCCIO voltages, and processors may need around 1.8V to post before FIVR kicks in to step it down.
In terms of best bang for your buck, the user thinks that the Core i5-12400 is the chip to pick up. The Alder Lake processor wields six Golden Cove cores with Hyper-Threading and doesn't have any Gracemont cores. Even though it only has a 4 GHz all-core boost clock speed, it should be a substantial upgrade over the Core i5-11400 or the older Core i5-10400.
As we already know, Alder Lake will only shine on Windows 11 with the new hardware scheduler. In Windows 10, programs and games will only work the Gracemont cores. Games that have been optimized for Windows 11 will put the Gracemont cores to sleep so the Golden Cove cores will have exclusive access to the large L3 cache.
A few months ago, a mysterious document showed that Intel was reportedly pushing the ATX12VO power connector with Alder Lake. However, there's nothing to worry about. The tipster estimated that over 90% of the 600-series motherboards will arrive with the conventional ATX 24-pin 12V power connector.
Take Your Pick: DDR4 or DDR5
Alder Lake natively supports DDR5-4800 or DDR4-3200 memory modules. When it comes to LGA1700 motherboards, we'll either get one or the other. For DDR5, we probably won't see those rare motherboards that support two different memory slots. Remember that DDR5 has its own 12V power management IC (PMIC) for voltage regulation. As a result, LGA1700 motherboards will not longer carry buck converters, but DDR4 still needs those. Therefore, the possibility of both DDR4 and DDR5 coexisting on a LGA1700 motherboard is less than 1%.
Memory will function differently with Alder Lake, depending if it's DDR4 or DDR5. DDR4 can run on Gear 1, but the ceiling is at DDR4-3600. Increasing the uncore voltage could bump it up to DDR4-3866, though. By default, DDR5 will only run on Gear 2, due to the high data rate. In order to see any improvement over DDR4, DDR5 needs to be clocked over DDR5-8000. Gear 4 is present for future DDR5 memory modules that exceed DDR5-10000.
In addition to DDR5 support, Alder Lake will leverage PCIe 5.0, but in a limited way. The processor reportedly offers a PCIe 5.0 x16 connection to the primary PCIe expansion slot whereas the connection to the M.2 slots, for example, is still confined to PCIe 4.0 speeds. AMD has shown us that PCIe 4.0 functionality takes a heavy toll on the chipset to a point where active cooling is required. The Chinese leaker claimed that Intel's 600-series chipsets don't run hot, or at least not enough to warrant a cooling fan.
Lastly, the Bilibili user believes that Alder Lake will be a lot more expensive than Rocket Lake. Early U.S. retailer listings lend credence to his notion. He went on to mention that the supply for the fourth quarter of this year is around 300,000 units, although he didn't specify if that number corresponds to the Chinese or Asian market.
looks like a lot of nonsense to me. 300,000 units at launch? that's a paper launch pure and simple. no one will see those chips till july if that's the initial product launch.
That...can't be right.
Were they trying to say that Games in windows 10 will use only the (up to) 8 big cores, and that Games in windows 11 will use only the 8 big cores plus a bit of extra cache?
If so, intel is making a mistake to give lower performance to Windows 10, given how bad Windows 11 looks.
Also, would that behavior really be limited to games? How would the windows scheduler know the difference between a game and any other program?
Sadly, that's par for the course lately though. You couldn't get your hands on a 5900x or 5950x until summer despite launching in October of last year, even the 5600x took into 21 before you could get it at retail pricing. Even worse if you throw Ampere into the mix, though that dynamic is a little different with mining.
They mentioned the games would have to be updated for Windows 11, if true it's probably an instruction to enter "big core only" mode. Not unlike using SSE instructions.
I think it is odd as well though, I'm not sure why you would completely shut off the low power cores. It's not like Windows pauses background tasks durning gaming, you would think they would want to keep those on, maybe just reduce their power or turn off a few. Although, given this is a first gen of this architecture, nothing would surprise me.
Any game that uses all cores will use all available cores which means also the smaller cores and if a game does that it will get huge syncing issues which will cause huge drops, just like zen had with the cross ccx issues, results would get to the GPU at different times.
I would assume games would only use big cores and the little cores would be for background/windows tasks.
There is already similar behavior with AMD and Intel boosting algorithms, where certain core(s) are identified as being the best ones within a chip, and as such are targeted for lightly threaded loads. The application doesn't need to know anything about it, it's up to the OS/drivers/FW to allocate the threads appropriately for best performance.