Fishing for Ice Lake rumours

Subject: General Tech | August 18, 2017 - 01:16 PM |
Tagged: rumours, Intel, ice lake, coffee lake, 9th generation core

It's Friday so why not engage in some speculation with us about Intel's upcoming new chips?  We will start off by confusing the issue with a post detailing Intel's naming conventions that The Inquirer found.  It would seem that not only is the "Ice Lake processor family is a successor to the 8th generation Intel Core processor family" but it is also described as an "“8th generation Intel Core Processor Family” and available only to early access users.  One can only hope that there is a typo in Intel's decoder ring as the current naming schemes are already confusing enough between AMD and Intel without adding more levels of complexity.

View Full Size

That makes the above a little more interesting than unannounced low power parts usually are.  AnandTech recently learned of these two new families of 8th gen chips, the i7-8xxx and i5-8xxx, both of which offer double the amount of cores as their 7th gen processors.  The base frequencies are lower than the previous generation, perhaps to remain inside the 15W TDP with double the amount of cores, with the turbo frequencies remaining a mystery for now.  With the aforementioned confusion, it is possible these could be Ice Lake based, though it is far more likely that they are indeed caffeinated instead.

View Full Size

The final rumour for you to look at this morning is the above screenshot from Chiphell.  You will need to zoom and enhance to get the full story, however there are some interesting reveals in the legible parts of the slide.  Enjoy.

"More news from Intel this morning, this time published directly on their website. With the upcoming announcement of the 8th Generation Core next week to which Intel has already posted teasers to the media, it would seem that someone at Intel decided to add processor details and pricing into Intel’s official Price List today."

Here is some more Tech News from around the web:

Tech Talk

Source: AnandTech

August 18, 2017 | 01:38 PM - Posted by John H (not verified)

I'm fairly certain once Ryan has signed an NDA then all rumor articles *must* be posted by other staff on here :).

I hope we'll see some real performance numbers on Monday but I'm guessing Monday is just an announcement before an actual release date with reviews...

2600K --> 8700K should be worth it finally.

August 18, 2017 | 05:57 PM - Posted by Jeremy Hellstrom

We're weird that way.

August 18, 2017 | 03:10 PM - Posted by UnderPlasticFantastic (not verified)

Hey some actual U series i7's with quad cores, but look at those clock speeds 1.6GHz to 1.9GHz and at the price of $409 and I'll still be rocking my Quad Core IvyBridge i7 3632QM(35W/6MB L3), $378 at intro pricing, at 2.20 GHz/3.20 GHz boost. Also at a 15W TDP Ice lake is really not going to have the thermal headroom for any heavy laptop workloads. But that 8MB L3 looks nice on the 8650U, 8550U SKUs, hopefully the boost on the quad core i7 Ice lake variants will get higher. and even the Ice lake i5s are getting 4 cores/8 threads(that's new).

Let's see how Raven Ridge stacks up with its 4/8 Zen cores/threads and I hope at least a few SKUs with 35+ Watt thermals so the Vega Graphics can breathe freely.

The Big question is What will Raven Ridge/APU's Vega Graphics use for HBC if there is no HBM2 for the HBCC to use as HBC(High Bandwidth Cache)! Maybe eDRAM in place of HBM2(?). Well hopefully the OEMs will use dual channels to DDR4 with any Raven Ridge APU SKUs because even Vega graphics is not going to do well if there is no HBM2 and the laptop's OEM only provides a single channel to DDR4 memory.

August 19, 2017 | 03:10 PM - Posted by remc86007

Vega APUs with an eDRAM setup could be really interesting. I'm interested to see more Vega undervolting/underclocking coverage as it is clear the current 64 and 56 are running way past their peak efficiency.

August 19, 2017 | 09:39 PM - Posted by SomeHBCforTheHBCC (not verified)

If AMD's Raven Ridge does not have HBM2 to act as the HBC then AMD will not be able to take advantage of that Vega HBCC/HBC IP as much and I'm hoping that AMD does eventually come up with an APU SKU that offers some HBM2, even at a single stack/2GB of HBM2 on an APU. This is because even a single stack of HBM2 provides a 1024 bit interface(8 128 bit independent channels) that provides up to 256GB of memory bandwidth if the HBM2 is clocked to the full JEDEC standard clock speed.

That's a lot more memory bandwidth than 2 regular memory channels to DIMM based DDR4 can provide. Also with regards to Vega's HBCC/HBC even 2GB of HBM2, if it is treated by the HBCC as a last level cache, can make use of a laptop's regular DIMM based DRAM as a virtual VIDEO memory pool allowing the integrated Vega graphics to have a much larger pool of texture/data memory.

So that Vega HBCC/HBC-HBM2 IP in Vega in that using the HBM2 as a last level Cache mode can extend virtual VRAM not only to the laptop's/PC's available system DIMM based DRAM but also extend that virtual VRAM to include memory paged onto a SSD/Hard-Drive for up to 512TB of virtual VRAM adressability. I'm not saying that someone gaming on a laptop would actually need more than the laptop's available DIMM based DRAM size for most games, but there are other programs and OS/Services that need to use the Laptop's/PC's DIMM based DRAM also so that ability to use virtual VRAM memory paging to an SSD/Hard-Drive comes in handy.

Most laptops are coming with at least the ability to have 8 or 16 GB of memory installed and most games have not been fully optimized for Vega's new feature set yet, but VEGA's HBCC/HBC IP is the most interesting part of that new design.

There has been very little information discussed by AMD concerning Raven Ridge, what with all this current Vega hubbub that is currently going on with supplies and pricing. So once that stabilizes then the next big release from AMD is Raven Ridge APUs for the mobile market only towards the end of this year, and desktop APUs in 2018.

The Vega undervolting/underclocking coverage is starting to come in now that there are folks getting theirs, but there also needs to be increased Vega Benchmarking over the long term as that games/gaming engines begin to make good use of Vega's HBCC/HBC-HBM2 IP and more games are converted over to being fully optimized for DX12/Vulkan so that the games can make better use of Vega's extra compute(Ditto for Polaris on that DX12/Vulkan with compute usage).

I think that most of the top 2-5% of the best thermally performing/clocking Vega Die stocks are going to the Radeon Pro WX/Radeon Instinct MI25/other variants. And those professional SKUs can command a higher price point so maybe soon enough there will be better more stocks of better performing Vega 10 dies that will perform a little better than these first release versions so the more Vega production the better from AMD and its fab partners.

The more HBM2 based GPU SKUs being sold on the market the loser that HBM2 price will eventually become so Samsung and Sk Hynix will get the needed revenues to ramp up more HBM2 production. Really Samsung is so big that they can afford to ramp up early on in advance to get more HBM2 on the market before the HBM2 prices become to high and choke off demand. AMD is probably intentionally using HBM2 when it could probably get by with using GDDR5/6 in an effort to create more use of HBM2 so that economy of scale can kick in sooner for HBM2 production and HBM2 prices can stabilize for the long term once the production capacity for HBM2 is there.

August 19, 2017 | 09:44 PM - Posted by SomeHBCforTheHBCC (not verified)

Edit: loser
to: lower

August 19, 2017 | 12:54 AM - Posted by Aparsh335i (not verified)

Jeremy, you should really proofread your work at least once before posting to the whole world. You literally mentioned some making a typo one sentence after you said "as described as described." It just looks lazy and unprofessional.

August 19, 2017 | 10:01 AM - Posted by Anonymouse (not verified)

I think what Intel is saying is that this is the last (or second to last, penultimate) CPU for this Socket and then you'll need a new Socket but the 300 Series Chipset will still be OK - so what, you'll need a new Motherboard for the Socket, it not like you can unsolder the Chipset and find a new Socket Motherboard to wave solder it on to.

Then, you'll be OK (won't need to buy a new Motherboard) for a couple of small incremental improvements (which you're not going to toss out your CPU to upgrade).

Then it's on to a new Series after the 300 to actually reap the next CPUs (unnamed, after Tigerlake) benefits; so you don't actually require a new Motherboard but you'll probably want one.

Then it's out of the 300's and into the 400's but don't worry you won't need a new Socket, but wait you'll need a new Motherboard because whether it's Socket change or Chipset change you'll need one - duh, now we're up to speed.

It's these small incremental improvements, and the Chipset for latest features before a Socket change smacks you, that's wearing mighty thin.

Just keep both Socket and Chipset for as long as reasonable and change BOTH at the same time, no need for Intel to peddle double the Motherboards - must be a Backroom Deal.

August 21, 2017 | 11:01 PM - Posted by Evan (not verified)

So many people were coming to the defense of the poor mistreated motherboard manufacturers when X299 was launched and lamented how they would have to work extra hard at printing literally 2 extra pages in their manual that say "use these slots only if you have these CPUs" and now that Z370 means MSI or Gigabyte get to sell a new $100 product this year, it's suddenly a shady backroom deal Intel has done. At least be paranoid consistently you nutters.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.