AMD on Thursday officially unwrapped two new Threadripper 3 CPUs with 24 and 32 cores, respectively. The company also announced an unlocked Athlon and said its 16-core Ryzen 9 3950X will finally go on sale on November 28.

The two new Threadripper CPUs are the biggest news. AMD will sell a 24-core Threadripper 3960X for $1,399. The chip is built on a 7nm process, features 140MB of cache, and will have a base clock of 3.8GHz and boost of 4.5GHz. The 32-core Threadripper 3970X, priced at $1,999, arrives in the wake of much undue hand-wringing that AMD would only go to 24 cores.

Both CPUs will go into AMD’s new TRX40 platform using the new sTRX4 socket. The use of a new socket essentially means the original X399 chipset, along with the  Threadripper 2000-series, will be stranded.

Insane amounts of PCIe lanes

The consolation may come with an insane amount of PCIe bandwidth. AMD said up to 72 lanes of PCIe are available in the new platform. The block map the company showed off indicates 48 PCIe 4.0 lanes can be dispersed for general use. Another pair of x4 PCIe 4.0 can be used for NVMe, or broken up as two quad-clusters of SATA ports. With the chipset connection at a beefy 8 lanes of PCIe 4.0, you get up to 64 from the CPU.

Much of that sounds the same as the previous X399, which also featured 64 PCIe lanes off the CPU. The key difference here is PCIe 4.0, which doubles bandwidth over PCIe 3.0.

The other difference is additional PCIe 4.0 lanes in the chipset. AMD’s block map shows 8 lanes of PCIe 4.0 off the TRX40 chipset for general use, and 8 more lanes of PCIe 4.0 that can be split off. That’s basically 72 total—AMD touts “88 lanes” of PCIe, but that’s only if you count the reserved lanes for chipset communication.

Quad-channel RAM stays put

The amount of memory channels for the new Threadrippers also appears unchanged, with up to four channels of DDR4 supported on both CPUs. While there had been speculation that AMD would go with an eight-channel memory configuration to coincide with an expected 64-core CPU, AMD didn’t say peep about additional memory channel support (which would require an additional new platform) or a new CPU. So for now, Threadripper sticks with quad-channel memory.

AMD will flex its core count muscles with Threadripper 3000.

AMD says itll be fast

AMD has long had a core count advantage over Core i9, but with the new, more efficient Zen 2 cores in the Threadripper 3000 chips, the company also promises a lot more performance. For example, AMD said its 32-core Threadripper 3970X will be up to 90 percent faster in Cinebench R20 over an 18-core Core i9-9980XE. The 24-core will outpace Intel’s current best Core i9 by 54 percent, the company added. AMD also claims a huge advantage over the top-end Skylake-X chip in Chromium compiles, Unreal engine compiles, Adobe Premiere CC encodes and Chaos Group’s V-Ray renderer.

The Core i9-9980XE isn’t Intel’s top prosumer chip, though. The company announced its new Core i9-10980XE at crazy-low prices in early October. While the older 18-Core i9-9980XE still sells for $1,919 today on Amazon.com, the new 18-core Core i9-10980XE will sell for $979.

In a strange turn of events, with Intel’s crazy price cuts in place, AMD’s new Threadripper will actually occupy the higher ground on pricing. If you use the overly simplistic metric of dollar per thread, Intel actually has an advantage in a Core i9 vs Threadripper price match. Of course, from AMD’s point of view, it probably thinks Core i9 can’t even hang with its Threadripper anymore and would rather see its $750 16-core Ryzen 9 3950X duke it out with Core i9 today.

Intel oddly almost has a price advantage.

Ryzen 9 3950X: Coming soon, with Eco Mode

Heres the other big announcement: The long-awaited, slightly delayed 16-core Ryzen 9 3950X will go on sale on November 25, the company said. Pinkie-swear. The CPU had been expected in September, but was pushed back to November. AMD has taken that time to add feature called Eco Mode.

Eco Mode gets you 77 percent the performance of the CPU at 44 percent of the power consumtion of the standard mode.

Eco Mode will let you scale the CPU back from its rated 105-watt TDP to 65 watts. There is a performance cost: AMD said Eco Mode will get you about 77 percent of the performance of the standard mode, but at 44 percent of the power consumption and also about 7 percent cooler. AMD said Eco Mode can be turned on and off from within a new version of the Ryzen Master utility.

The company said the mode was added by request from users who want to use Ryzen CPUs inside home theater PCs and other environments with limited thermal and noise headroom. AMD is casting a wide net, saying the new Eco Mode will apply to other 105-watt Ryzen-based chips. Eco Mode will also allow 95-watt chips to drop to 65 watts, and 65-watt CPUs to drop to 45 watts.

AMD took the opportunity to drop more hints about how much of a thrashing the CPU will give to its competitors. Of course, the comparison is fairly close against Intel’s $500, 8-core, Core i9-9900K, but AMD said it’ll thrash the 12-core Core i9-9920X, which sells for $1,000.

According to AMD, the Ryzen 9 3950X will outrun the Core i9 in some games.

Content creation will be the 16-core Ryzen 9 3900Xs forte. The $500 Ryzen 9 3900X already had the Core i9-9900KS on the ropes for multi-threaded and most content creation tasks, so the 16 core? Yeah, it’s just a victory lap at this point for AMD.

Content creation though, is a cake walk.

An overclockable Athlon, too

Not everyone can afford a $2,000 32-core Threadripper 3970X or a $750 Ryzen 9 3950X, so AMD is throwing budget PC users a bone: an overclockable Athlon 3000G.

The Athlon 3000G will feature two cores with symmetrical multi-threading, and Radeon Vega 3 graphics, for $50. The chip will have a clock frequency of 3.5GHz and, yes, you can overclock it.

AMD envisions the Athlon 3000G going up against Intel’s $60 dual-core Pentium Gold G5400. Although AMD didn’t detail its architecture, the Athlon 3000G is likely a Zen+ chip built on a 12nm process. That’s still enough, AMD claims, to put it ahead of the Pentium Gold G5400. AMD said its tests show the Athlon outperforming the Pentium in boring desktop use, and easily outrunning the Pentium Gold in gaming. AMD said the new Athlon 3000G will go on sale November 19.

Struggling microprocessor maker Advanced Micro Devices said Rory Read, its chief executive, had stepped down and will be replaced by Lisa Su, the companys chief operating officer, effective immediately.

The company said that the transition was part of a succession plan that Read and the companys board had developed together. In a somewhat unusual move, Read appeared with Su on a conference call announcing the transition.

"Shes a semiconductor professional," Read of Su. "She loves this space. She knows this space." Read praised Su extensively, saying that AMD had hired Su, trained her, and apparently groomed her for the position. Shes the "right leader" for AMD, he said, and is uniquely positioned for the job, he said.

Intel’s eleven new Ice Lake chips are betting big on graphics. Intel claims that its new 10th-gen Core family, launched Thursday, will offer the best integrated graphics experience of any thin-and-light notebook PC—about equaling what some discrete chips offer in PCWorld tests.

Intel began setting up Ice Lake expectations in May, positioning its first mass-market 10nm parts as improvements for the notebook PC as a whole, including AI, communications, and I/O. The overarching message is “new:” a new process technology, a new CPU core, and a new graphics engine, along with “new” advancements in communications and I/O. 

Here are Intel’s mobile Ice Lake chips, in a nutshell: 11 processors, most sporting 4 cores and 8 threads, in either a 9W, 15W, or 28W power tier. Base frequencies begin at 700MHz and climb to 2.3GHz, while single-clock turbo frequencies top out at 4.1GHz. Graphics frequencies will hover right around 1GHz, with 32, 48 and 64 EUs differentiating performance. We also have real-world testing data to demonstrate how Ice Lake delivers graphic performance on a par with the lowest tier of discrete GPUs.

Ice Lakes "sizzle reel:" everything that Intel believes that Ice Lake can offer, along a number of performance axes.

A bit of sticker shock, however, may affect how some perceive Ice Lake. Most 10th-gen Ice Lake base clock speeds fall short of their existing 8th-gen Whiskey Lake counterparts. Some of the new Ice Lake low-power chips offer base clocks of less than 1GHz. Versions of Ice Lake designed for tablets now consume 9 watts, rather than the 5 watts of their predecessors, Amber Lake.

Intel says that more than 35 notebooks built around Ice Lake microprocessors will be sold alongside existing Whiskey Lake and Amber Lake notebooks, beginning this holiday season. There’s still more to come: Intel’s next microprocessor disclosure will be made in the second half of August, where the “second part of our mobile family” will target “high-frequency applications” and form the foundation for Intel’s commercial offerings, said Ran Senderowitz, vice president of mobile platform marketing for Intel’s Client Computing Group.

Will that be Intel’s next-gen Tiger Lake chip, due in 2020? Or the Comet Lake chip that was part of a recent leaked roadmap? Things are getting messy.

The basic specifications of Intel’s Ice Lake processor, summed up.

How to read Intel’s new Ice Lake model numbers

Intel is promising to ship new products to every PC segment over the next 12 months—nothing particularly new, as product refreshes occur throughout the year. But with a mixture of 14nm and the new 10nm products co-existing in the market, things may get confusing—especially with Qualcomm’s Snapdragon 8cx and AMD’s third-gen mobile Ryzen chips also in play.

Before you can parse Intel’s new Ice Lake products, however, you’ll need to describe them—and yes, Intel has a new model number scheme, too.

Intel still offers Core i3, i5, and i7 processor families, differentiating their good-better-best status. Each Ice Lake model number also begins with a 10 prefix to designate its 10th-generation status. 

Now, however, there are good-better-best suffixes, too: the G1, G4, and G7 descriptors signal the various tiers of graphics performance, with both the G4 and G7 tiers falling under the Intel Iris Plus branding. Also, Intel has officially done away with the U-series and Y-series labels for notebooks and tablets. Instead, the fourth digit of the model number now designates the power tier: “5” for a U-series chip, and “0” for a Y-series processor. A single 28W (or 8-tier) chip exists: the Core i7-1068G7.

Still confused? We’ve boiled it down for you, here:

Here’s how to make sense of the new Intel model numbers.

Intel’s eleven Ice Lake chips, explained

Now that you know what to look for within Intel’s Ice Lake model numbers, here are the eleven new chips.

As the slide notes, all of the new Ice Lake processors will include Wi-Fi6 (Gig+) Wi-Fi and Thunderbolt 3 capabilities: four ports this time around, with two ports available on either side of the PC. We go into far more detail in our earlier discussion of the Ice Lake architecture, which is predicated on the Sunny Cove CPU design, a key feature of which is "magic"  AI inferencing using Intel’s DLBoost instructions. 

Intel’s eleven new Ice Lake microprocessors. The ConfigUP TDP numbers indicate that a laptop maker is permitted to power the chip up to the higher rated wattage, to enable additional performance. Battery life can suffer somewhat as a result, however.

We also now know that Intel’s Gen11 graphics have three different performance tiers, each with different numbers of EUs: 32, 48, and 64. (EUs are Intel’s programmable shader units, with 3D rendering, compute, programmable and fixed-function media capabilities.) Based on what we’re seeing here, Intel appears to be using the number of EUs as a performance differentiator, rather than the clock speed. Remember, Intel has previously made the claim that Ice Lakes integrated graphics were as good as what AMD offers, keeping pace with the Ryzen 3700U on a variety of different games. 

According to Senderowitz, Intel won’t be adding to the three EU graphics tiers (at least for now, we can assume). The eleven Ice Lake chips represent the entirety of Ice Lake products that Intel plans to introduce this year, he said.

Nevertheless, one of our Ice Lake performance tests shows how well Intels Ice Lake matches up to the downclocked Nvidia GeForce MX100 discrete GPU that was featured in a recent release of an Huawei Matebook. Intels claims that its Ice Lake graphics capabilities equal those of a discrete GPU seem to be borne out. 

Note that Huawei used two versions of the GeForce MX100 discrete GPU in successive Matebook notebooks; the slower chip was used in the Huawei Matebook X Pro.

While we dont know what Intel will be charging individual PC makers, Intel published some of its microprocessor prices as part of its ARK database on Thursday morning:

• Core i7-1065G7: $426
• Core i5-1035G7: $320
• Core i5-1035G4: $309
• Core i5-1035G1: $297
• Core i3-1005G1: $281

For reference, Intels August microprocessor price list lists the 8th-gen "Whiskey Lake" i7-8565U at $409, and the Core i5-8265U at $297.

How Ice Lake differs from Whiskey Lake, Amber Lake

What’s interesting, though, is that the low-power (previously known as Y-series) parts consume more power, at 9W, than their 5W Amber Lake predecessors. Ultrabooks and tablets that used the Amber Lake Y-series chips were relatively few—Ultrabook Review counted just seven, of which five were 13-inches or larger. It’s not clear whether the higher power levels will end up attracting more tablet customers. Save for Dells announcement of its Ice Lake-based XPS 13 2-in-1, we really dont know which notebooks will use Ice Lake, period.

On paper, Intels 8th-gen Whiskey Lake chips look more attractive. But Intel says design enhancements and other improvements will boost real-world performance higher. 

The vast majority of Ice Lake chips also start out at extremely low clock speeds, yet can boost to about the levels of their Whiskey Lake counterparts. How long Ice Lake chips can remain in boost mode will help determine their performance. Again, our performance evaluation goes into more detail. 

While Intels eighth-gen Whiskey Lake chips still emphasized performance increases versus the prior generation, those chips introduced Intels gigabit WiFi and two ports of Thunderbolt 3 I/O. Both have been beefed up, to gigabit+ speeds as well as four Thunderbolt ports.

Intel also enhanced its graphics quality as well as its performance, with support for three simultaneous 4K displays at 120Hz, a pair of 5K displays at 60Hz, or a single 8K30 display. The graphics engine includes three display pipes that support DisplayPort 1.4 HBR3 and HDMI 2.08, plus HDR10 and Dolby Vision.

Whats next? Project Athena, faster speeds, probably more cores

What we’ll learn at the end of August about the next chip will also help position Ice Lake. Will that chip introduce an H-series part? Simply flesh out the top end of the “U-series” lineup with a Core i9 offering? Introduce an entirely new lineup? Based on what Senderowitz said about the entirety of the Ice Lake product line, we’d opt for the latter. Apparently the Core i7-1068G7 doesn’t necessarily represent the high end of Intel’s Core offerings for 2019. A six-core, twelve-thread processor almost certainly is in Intel’s future, too. 

Remember, Ice Lake represents Intel’s first mass-market 10nm processor and its first major redesign in several generations. Even though Intel shipped its first 10nm Cannon Lake chip, the Core i3-8121U, in May 2018, it probably never exceeded extremely low volumes. It’s fair to say, then, that Intel has significant latitude to refine the 10nm process and push up future clock speeds. 

“I think that you will be able to see how we optimize the workloads...for more segmented and more targeted designs to meet customer needs,” Senderowitz said of what you can expect of Intels future processor positioning.

Weve been waiting for a year or so for Project Athena concepts and other designs to arrive, though theyve been designed to make a splash when they do.

For now, however, Intel’s betting big on Ice Lake, and you should see numerous thin-and-light and notebook designs built around Intel’s Ice Lake processors this holiday season. Remember, too, that those will include ”Project Athena” designs: flashy “halo” notebooks similar to HP’s Spectre Folio, that try to accomplish bold new things with the next-generation ultrabook form factor. 

If youre in the market for a notebook PC this holiday season, though, we can already say that youll benefit immensely from knowing what youre buying. We see 2nd- and a few 3rd-gen Ryzen notebooks crowding the low end of the notebook PC market, alongside the 7th- and 8th-gen Core chips that will give way to Ice Lake and its future, faster, cousins. Qualcomm claims that its Snapdragon 8cx offers Core i5-class performance, too.

All of that competition can create confusion. But you, as a customer, will ultimately benefit. 

This story was updated at 9:37 AM to include Intels published processor prices for its 10th-gen Core, Ice Lake chips.