No, AMD Isn’t Building a 48-Core Ryzen Threadripper 3980X

Rumors are rocketing around the ‘net that AMD is preparing to launch a 48-core Ryzen Threadripper 3980X, based on a fake image being passed around on Twitter. It’s a superficially tempting thought, as it offers the prospect of a high-core-count CPU with perhaps slightly higher clocks than the 3990X. But there are multiple reasons to believe this screenshot is untrue. Said fake looks like this:

Fake 3980X screenshot.

AMD has already stated it has no plans to launch a 48-core chip and none of the information we’ve uncovered about this screenshot suggests it has any intention of changing them. In the past, AMD has indicated that the mid-tier Threadripper parts don’t tend to sell very well; customers either go for the sweet spot chips or the highest-end parts, but not much in-between. This explains why the company has taken the approach it has with third-generation Threadripper. This time around, there’s an entry-level 24-core part, one “sweet spot” CPU (3970X, 32-cores), and one halo part, the 3990X. The previous entry-level TR CPU, of course, became the 16-core desktop 3950X and bumped down to the top of that product stack rather than introducing the workstation Threadripper family.

The second reason to believe this CPU screenshot is fake is that the author forgot to change the “3990X” moniker in the “Specification” field and left it reading 3990X instead.  This isn’t how engineering chips are badged, either. An ES chip might have a seemingly random code in place of its formal product name or a non-standard entry in another field, but you aren’t going to find a 3980X that’s accidentally been badged as a 3990X. Doesn’t happen that way.

Finally, there’s the fact that AMD currently has no reason to release a 48-core Threadripper. A 48-core version of the CPU will have the same scheduling problems that the 3990X does, because Microsoft hasn’t bothered to fix the Windows scheduler yet to support more than 64 logical CPU cores per group (initial reports that Windows 10 Enterprise would outperform Windows 10 Pro did not survive additional analysis). The 3990X offers some real performance improvement over the 3970X and our sample was a great overclocker, but if you aren’t doing the right kinds of applications to benefit from a 3990X, the 3970X will be a better performer.

Right now, the 3990X is a specialty halo part that really only makes sense for a small number of people with specific workload requirements. It’s a technology demonstration as much as a commercial product, and it’s not a product market we’d expect AMD to build out until Windows 10 is more friendly to these high core count configurations. Any CPU above 32C/64T will have the same Processor Group limitation as the 3990X.

None of this means AMD won’t ever release a 48-core chip, but I don’t think we’ll see the firm buffing up its consumer core counts in quite that way just at the moment. A lower-cost 48-core chip is the sort of thing I’d expect AMD to either reserve for a new Threadripper debut or as a competitive response to something Intel had built. Intel isn’t going to be launching any HEDT CPUs with that many cores in the near future either, and AMD has little reason to introduce one now.

Right now, in fact, it looks as if the big fight between 10th Gen Core and AMD’s 7nm Ryzen will be happening in the lower-end and midrange market. AMD now has 4C/8T chips in-market for $100, while Intel has new Core i3 CPUs in a similar configuration starting for $122. The gains from moving to 4C/8T from 4C/4Tmay not be as large as the improvement from 2C/4T to 4T/4T, but this will be the second significant thread-count upgrade the Core i3 has gotten in a relatively short period of time, courtesy of AMD’s willingness to push the envelope at every point in the desktop market.

Now Read:

from ExtremeTechExtremeTech

from Blogger

New Fossils Prove Spinosaurus Was an Underwater Terror

We’re all familiar with Tyrannosaurus Rex, a massive theropod dinosaur from the Cretaceous period and star of several movies about dinosaurs eating people. However, there were even larger, potentially more terrifying beasts on Earth all those millions of years ago. Spinosaurus was even bigger than the T-rex, and new discoveries indicate you wouldn’t have been safe even in the water. Spinosaurus, it turns out, was an excellent swimmer thanks to its large, paddle-like tail. 

Spinosaurus was a theropod like the Tyrannosaurus — that just means it had hollow bones and three-toed limbs. The descendants of theropods most likely evolved into modern birds, but Spinosaurus was more dangerous than any bird. Adults could weigh as much as 7.5 tons and grow to more than 50 feet in length, making them among the largest theropod dinosaurs. 

Researchers first proposed that Spinosaurus was primarily an underwater predator several years ago, but the scientific community was unconvinced. Donald Henderson, a paleontologist at Canada’s Royal Tyrrell Museum, noted that Spinosaurus was probably top-heavy with its distinctive back sail and would not have been able to dive underwater. Nazir Ibrahim, lead author of the study, believed the answer would be found in fossils. Previous excavations had only uncovered a few Spinosaurus tail sections, but the team uncovered an almost full set of tail bones at a fossil site in Morocco between 2017 and 2018. 

The newly reconstructed Spinosaurus was undeniably at home in the water. Rather than having a tapered whip-like tail, Spinosaurus had a giant fin attached to its backside. Some of the fossil bones were 12-inches thick, indicating the tail would have been a powerful mode of underwater propulsion. The team speculates Spinosaurus might have spent most of its time in the water. 

The team created a computer model to assess the capabilities of Spinosaurus’ tail, comparing it with modern land-dwelling dinosaurs and semi-aquatic creatures like crocodiles. Unsurprisingly, the Spinosaurus tail fin was about 2.6 times more efficient in the water than the tails of other theropods. 

Museums around the world will have to update their Spinosaurus models in the wake of this discovery, but that’s nothing new. The fossil record is incomplete, and sometimes we get details wrong when trying to reconstruct an entire animal from partial remains. The Tyrannosaurus Rex skeleton at the American Museum of Natural History in New York stood in the incorrect upright posture until 1992 before adopting the correct parallel position. Oh, the developers of Animal Crossing will have to update their inaccurate Spinosaurus fossils, too.

Top image credit: Kumiko/Flickr/CC BY-SA 2.0

Now read:

from ExtremeTechExtremeTech

from Blogger

Overclocking Results Show We’re Hitting the Fundamental Limits of Silicon

Update (4/30/20): The formal unveiling of Intel’s 10th Generation Core i9 family is an excellent opportunity to revisit the points made in this November 2019 article. As of this writing — some five months later — Silicon Lottery is out of 9th Gen chips and waiting on Comet Lake CPUs to arrive. The handful of 7nm AMD CPUs show very similar patterns to what we identified back in November. A 3950X @ 4GHz is just $750, but an all-core 4.1GHz is $850 and a 4.15GHz chip is $999.

Now, with its 10th Gen Comet Lake, Intel has adopted strategies like die-lapping and adding copper to its IHS to improve thermal transfer off the core, while allowing much higher levels of power consumption. It’s not that there’s something wrong with the parts from either company — manufacturers increasingly have no additional firepower to leave on the table for enthusiasts to enjoy.

Original story below:

Silicon Lottery, a website that specializes in selling overclocked Intel and AMD parts, has some 9900KS chips available for sale. The company is offering a 9900KS verified at 5.1GHz for $749 and a 9900KS verified at 5.2GHz for $1199. What’s more interesting to us is the number of chips that qualify at each frequency. Thirty-one percent of Intel 9900KS chips can hit 5.1GHz, while just 3 percent can hit 5.2GHz. The 5.2GHz option was available earlier on 11/4 but is listed as sold-out as of this writing.

The 9900KS is an optimized variant of Intel’s 9900K. The 9900K is Intel’s current top-end CPU. Given the difficulties Intel has had moving to 10nm and the company’s need to maintain competitive standing against a newly-resurgent AMD, it’s safe to assume that Intel has optimized its 14nm++ process to within an inch of its life. The fact that Intel can ship a chip within ~4 percent of its apparent maximum clock in sufficient volume to launch it at all says good things about the company’s quality control and the state of its 14nm process line.

What I find interesting about the Silicon Lottery results is what they say (or said, as of November 2019) about the overall state of clock rates in high-performance desktop microprocessors. AMD is scarcely having an easier time of it. While new AGESA releases have improved overall clocking on 7nm chips, AMD’s engineers told us they were surprised to see clock improvements on the Ryzen 7 3000 family at all, because of the expected characteristics of the 7nm node.

AMD and Intel have continued to refine the clocking and thermal management systems they use and to squeeze more headroom out of silicon that they weren’t previously monetizing, but one of the results of this has been the gradual loss of high-end overclocking. Intel’s 10nm process is now in full production, giving us some idea of the trajectory of the node. Clocks on mobile parts have come down sharply compared with 14nm++. IPC improvements helped compensate for the loss in performance, but Intel still pushed TDPs up to 25W in some of the mobile CPU comparisons it did.

I think we can generally expect Intel to improve 10nm clocks with 10nm+ and 10nm++ when those nodes are ready. Similarly, AMD may be able to leverage TSMC’s 7nm node improvements for some small frequency gains itself. It’s even possible that both Intel and TSMC will clear away problems currently limiting them from hitting slightly higher CPU clocks. Intel’s 10nm has had severe growing pains and TSMC has never built big-core x86 processors like the Ryzen and Epyc chips it’s now shipping. I’m not trying to imply that CPU clocks have literally peaked at 5GHz and will never, ever improve. But the scope for gains past 5GHz looks limited indeed, and the 5.3GHz top frequency on Comet Lake doesn’t really change that.

Power per unit area versus throughput (that is, number of 32-bit ALU operations per unit time and unit area, in units of tera-integer operations per second; TIOPS) for CMOS and beyond-CMOS devices. The constraint of a power density not higher than 10 W cm2
is implemented, when necessary, by inserting an empty area into the optimally laid out circuits. Caption from the original Intel paper.

The advent of machine learning, AI, and the IoT have collectively ensured that the broader computer industry will feel no pain from these shifts, but those of us who prized clock speed and single-threaded performance may have to find other aspects of computing to focus on long-term. The one architecture I’ve seen proposed as a replacement for CMOS is a spintronics approach Intel is researching. MESO — that’s the name of the new architecture — could open up new options as far as compute power density and efficiency. Both of those are critical goals in their own right, but so far, what we know about MESO suggests it would be more useful for low-power computing as opposed to pushing the high-power envelope, though it may have some utility in this respect in time. One of the frustrating things about being a high-performance computing fan these days is how few options for improving single-thread seem to exist.

This might seem a bit churlish to write in 2020. After all, we’ve seen more movement in the CPU market in the past 3 years, since AMD launched Ryzen, than in the previous six. Both AMD and Intel have made major changes to their product families and introduced new CPUs with higher performance and faster clocks. Density improvements at future nodes ensure both companies will be able to introduce CPUs with more cores than previous models, should they choose to do so. Will they be able to keep cranking the clocks up? That’s a very different question. The evidence thus far is not encouraging.

Now Read:

from ExtremeTechExtremeTech

from Blogger

Intel Unveils Comet Lake Desktop CPUs: Up to 5.3GHz, 10C/20T, and Some Unanswered Questions

The existence and imminent launch of Comet Lake, Intel’s first 10th Generation mainstream desktop CPU family, has been an open secret for several months. But Intel has stayed quiet about the specs and capabilities of the family even as rumors mounted. Now the company is finally talking about the CPUs it will launch later in May. There are some good reasons for Intel enthusiasts to be excited, but we’ve definitely got questions heading into the launch.

Let’s hit the high points first: As anticipated, the Comet Lake S 10th Generation Core desktop CPU family hauls out all the stops in an effort to push frequencies and core counts a little higher. The new Core i9-10900K is a 10C/20T CPU with a boost clock of up to 5.3GHz. That’s a 1.06x increase in top-line clock over the 9900K combined with a 1.25x core count improvement. That’s not an unreasonable level of improvement over the Core i9-9900K given that both are 14nm CPUs.


Hitting these higher frequencies, however, has required Intel to tweak a number of dials and levers. The company is now lapping its own die to reduce the amount of insulative material in-between the transistors themselves and the TIM. At the same time, Intel has increased the amount of copper in its IHS to improve its thermal conductivity.


Die-lapping has been discussed in overclocking circles as a method of improving thermals and possibly overclocks with the 9900K, but seeing Intel officially adopt it here illustrates how difficult it is for even Intel to keep pushing CPU clocks higher. The reason for the slightly thicker IHS is to keep z-height identical to allow for cooler re-use. The higher percentage of copper in the IHS more than offsets its increased thickness.

Other new features of the 10th Gen family include the ability to set a voltage/frequency curve in utilities like XTU, the use of Hyper-Threading across all Intel Core CPUs (i3/i5/i7/i9), formal support for DDR4-2933, integrated Wi-Fi 6 (802.11ax) support, new 2.5G Intel ethernet controllers based on Foxville, and some software optimizations in games like Total War: Three Kingdoms and Remnant: From the Ashes.

What Intel Hasn’t Said

There are several topics Intel hasn’t clarified and didn’t answer questions about during the conference call. No new details about the Z490 chipset or the status of its PCIe 4.0 support were given, even though multiple motherboard OEMs are claiming support for that standard is baked into upcoming boards. There have been rumors of a flaw in the 2.5G Ethernet controller that haven’t been clarified.

The additional pins added to LGA1200 are reportedly for power circuitry and we know the board TDP has bumped to 125W, but that number seems fairly meaningless in light of what we know about power consumption on modern high-end Intel CPUs. Unless you specifically program them to draw no more than their rated TDPs, high-end chips like the 9900K draw far more than 95W while boosting under load. Sustained power draw is also much higher. Neither AMD nor Intel sticks to TDP as a measure of actual power consumption, but the rumors concerning the 10900K have implied it could draw as much as 250-280W.

Ignore the “Up to’s” in the base clock column in the image above. Intel has informed ExtremeTech that these frequencies are meant to be listed as static clocks. “Up to” only applies to the boost clock frequency. Overall, these new CPU configurations are an improvement over what Intel has had in-market with 9th Gen.

The Core i9-9900K is an 8C/16T with a 3.6GHz base clock and 5GHz all-core boost, with an official price of $500. The Core i7-10700K is an 8C/16T CPU with a 3.8GHz base clock and a 5.1GHz boost clock, with the same 4.7GHz all-core boost as the 9900K. Price? $375.

I’m not going to speculate on how the 10700K will compare against CPUs like the 3700X or 3900X until we have silicon in-hand, but the 10700K is much better-positioned against AMD than its predecessor was. It isn’t clear how much performance will improve from the 9900K to the 10700K, but the 10700K should offer at least 100 percent of the Core i9’s performance for 75 percent its price.

The price cuts and performance adjustments continue down the stack, to good overall effect. The bottom-end Core i3, the Core i3-10100, is a 4C/8T CPU with a 3.6GHz base clock and 4.3GHz turbo for $122. The equivalent 9th Gen CPU, the Core i3-9100, is a 4C/4T CPU at 3.6GHz/4.2GHz. The addition of HT should translate to a 1.15x – 1.2x improvement across the board.

Comet Lake and LGA1200 will definitely deliver some improvements over 9th Gen, but we want to see exactly how these chips and platforms compare before we say more. One thing we are sure of — anyone planning to play at the top of the Comet Lake stack will want a high-end CPU cooler to make certain they squeeze every last bit of performance out of the chip.

Now Read:

from ExtremeTechExtremeTech

from Blogger

Toshiba Clarifies Which of Its Consumer HDDs Use Shingled Magnetic Recording

After news broke that the major hard drive companies have all been shipping SMR drives into various consumer products, some of the manufacturers involved have been clarifying which of their own HDDs are actually SMR drives instead of CMR. Toshiba is the latest company to release this information, though there are limits to its report that make it potentially less useful than we’d like.


Seagate has also deployed shingled magnetic recording to boost areal density of its drive platters, though the technology isn’t a great fit for consumer products.

As a reminder: SMR stands for Shingled Magnetic Recording and refers to the placement of tracks on the HDD platter itself. With conventional recording, a gap is left in-between each track, allowing the track to be individually read and written. With SMR, the tracks are layered directly next to each other, rather like shingles. This means that writing data to the drive requires reading and writing multiple tracks at once.

Data and graph by Anandtech.

The impact on read speeds is small-to-nil, but the write speed impact for using SMR can be significant. There’s not a ton of data on how this hits consumer use-cases because, up until now, reviewers haven’t been treating SMR drives as if they were likely to wind up being used for primary hard drives. I’d be surprised if we don’t start seeing more reviews on this in short order.

In any event, here’s what Toshiba has to say. The P300 6TB HDWD260UZSVA and P300 4TB HDWD240UZSVA are both SMR desktop HDDs that Toshiba only ships to bulk OEMs — which means any laptop you buy with a 4TB or a 6TB Toshiba HDD has to be checked to see if it uses one of these two models. Retail P300 drives top out at 3TB.

Now, unlike the desktop family, multiple bulk and retail L200 laptop drives also use SMR, including:

HDWL120UZSVA (2TB, bulk)
HDWL120EZSTA (2TB, retail)
HDWL120XZSTA (2TB, retail)

The 1TB drives at 7mm thick are also impacted (HDWL110UZSVA, HDWL110EZSTA, and HDWL110XZSTA). The first is a bulk drive, the second two are retail products.

Unfortunately, because Toshiba is selling these drives in bulk, it may be difficult to make certain you aren’t buying one. ExtremeTech does not recommend using an SMR drive in a consumer system as primary storage unless you are specifically aware of its likely performance characteristics and do not mind them. The dramatically lower write performance that SMR offers in some instances is of less concern for personal backup drives or similar applications, but hard drives are already poor solutions for storage speed compared with SSDs, and SMR drives are lower than CMR (conventional magnetic recording) in several additional aspects.

We are glad Toshiba came forward with this information, but we can only recommend buying a system with a Toshiba HDD if you either know exactly what you’ll be getting into or can confirm you aren’t buying an SMR drive.

Now Read:

from ExtremeTechExtremeTech

from Blogger

Chiplets Do Not ‘Reinstate’ Moore’s Law

Ever since chiplets became a topic of discussion in the semiconductor industry, there’s been something of a fight over how to talk about them. It’s not unusual to see articles claiming that chiplets represent some kind of new advance that will allow us to return to an era of idealized scaling and higher performance generation.

There are two problems with this framing. First, while it’s not exactly wrong, it’s too simplistic and obscures some important details in the relationship between chiplets and Moore’s Law. Second, casting chiplets strictly in terms of Moore’s Law ignores some of the most exciting ideas for how we should use them in the future.

Chiplets Reverse a Long-Standing Trend Out of Necessity

The history of computing is the history of function integration. The very name integrated circuit recalls the long history of improving computer performance by building circuit components closer together. FPUs, CPU caches, memory controllers, GPUs, PCIe lanes, and I/O controllers are just some of the once-separate components that are now commonly integrated on-die.

Chiplets fundamentally reverse this trend by breaking once-monolithic chips into separate functional blocks based on how amenable these blocks are to further scaling. In AMD’s case, I/O functions and the chip’s DRAM channels are built on a 14nm die from GF (using 12nm design rules), while the actual chiplets containing the CPU cores and the L3 cache were scaled down on TSMC’s new node.

Prior to 7nm, we didn’t need chiplets because it was still more valuable to keep the entire chip unified than to break it into pieces and deal with the higher latency and power costs.


Epyc’s I/O die, as shown at AMD’s New Horizon event.

Do chiplets improve scaling by virtue of focusing that effort where it’s needed most? Yes.

Is it an extra step that we didn’t previously need to take? Yes.

Chiplets are both a demonstration of how good engineers are at finding new ways to improve performance and a demonstration of how continuing to improve performance requires compromising in ways that didn’t used to be necessary. Even if they allow companies to accelerate density improvements, they’re still only applying those improvements to part of what has typically been considered a CPU.

Also, keep in mind that endlessly increasing transistor density is of limited effectiveness without corresponding decreases in power consumption. Higher transistor densities also inevitably mean a greater chance of a performance-limiting hot spot on the die.

Chiplets: Beyond Moore’s Law

The most interesting feature of chiplets, in my own opinion, has nothing to do with their ability to drive future density scaling. I’m very curious to see if we see firms deploying chiplets made from different types of semiconductors within the same CPU. The integration of different materials, like III-V semiconductors, could allow for chiplet-to-chiplet communication to be handled via optical interconnects in future designs, or allow a conventional chiplet with a set of standard CPU cores to be paired with, say, a spintronics-based chip built on gallium nitride.

We don’t use silicon because it’s the highest-performing transistor material. We use silicon because it’s affordable, easy to work with, and doesn’t have any enormous flaws that limit its usefulness in any particular application. Probably the best feature of chiplets is the way they could allow a company like Intel or AMD to take a smaller risk on adopting a new material for silicon engineering without betting the entire farm in the process.

Imagine a scenario where Intel or AMD wanted to introduce a chiplet-based CPU with four ultra-high-performance cores built with something like InGaAs (indium gallium arsenide), and 16 cores based on improved-but-conventional silicon. If the InGaAs project fails, the work done on the rest of the chip isn’t wasted and neither company is stuck starting from scratch on an entire CPU design.

The idea of optimizing chiplet design for different types of materials and use-cases within the same SoC is a logical extension of the trend towards specialization that created chiplets themselves. Intel has even discussed using III-V semiconductors like InGaAs before, though not since ~2015, as far as I know.

The most exciting thing about chiplets, in my opinion, isn’t that they offer a way to keep packing transistors. It’s that they may give companies more latitude to experiment with new materials and engineering processes that will accelerate performance or improve power efficiency without requiring them to deploy these technologies across an entire SoC simultaneously. Chiplets are just one example of how companies are rethinking the traditional method of building products with an eye towards improving performance through something other than smaller manufacturing nodes. The idea of getting rid of PC motherboards or of using wafer-scale processing to build super-high-performance processors are both different applications of the same concept: Radically changing our preconceived notions on what a system looks like in ways that aren’t directly tied to Moore’s Law.

Now Read:

from ExtremeTechExtremeTech

from Blogger