Hardware stub

HW News - AMD Launches 64-Core CPUs, RX 5700 XT THICC, & Intel 10nm Shipping

Posted on August 10, 2019

Hardware news this week is largely focused on new product launches, or rumors thereof, with additional coverage of Intel's plans to launch 10nm Ice Lake CPUs in some capacity (for real, this time) by end of year. The XFX RX 5700 XT "THICC" was leaked -- yes, that's a real name -- and it's accompanied by other partner model cards coming out in the next week.

Show notes continue after the embedded video.

AMD AGESA Update Removes PCIe 4.0 Support

AMD’s latest AGESA update, AGESA AM4 1.0.0.3 ABB, will axe any PCIe 4.0 functionality on non-X570 AM4 boards. This is due to a handful of rogue motherboard vendors implementing PCIe 4.0 support on older motherboards, albeit in limited form. In addition to removing any added PCIe support for older motherboards, the new AGESA update also includes fixes for Destiny 2, which has been a problem for early adopters of Ryzen 3000.

AMD officially told us at Computex that it intended to eventually kill PCIe 4.0 support. We’ve had some motherboard manufacturers tell us that, theoretically, they would be able to reverse engineer the AGESA binary provided by AMD and re-enable PCIe 4.0, but it’s just a question of if anyone actually does that. The code is provided as binary and isn’t explained or detailed beyond what’s provided, so a company would have to go through and toggle bits until they find the PCIe 4.0 support.

Using an older AGESA would allow you to continue using PCIe Gen4 on boards where it was temporarily supported, but then you’d lose other fixes in AGESA code, like the Destiny fixes.

Intel Ships 10nm Ice Lake For 4Q19 Availability

While discussing Q2 earnings, Intel revealed the company began shipping 10nm Ice Lake chips in volume during 2Q19. During a conference call, CEO Bob Swan confirmed 10nm volume production.

“We began shipping Ice Lake clients in the second quarter supporting systems on the shelf for the holiday selling season and expect to ship Agilex, our first 10-nanometer FPGA later this year.

We now have two factories in full production on 10-nanometer. We are also on track to launch 7-nanometer in 2021. With a roughly 2x improvement in density over 10-nanometer, our 7-nanometer process, which will be comparable to competitors' 5-nanometer nodes, and will put us on pace with historical Moore's Law scaling.

We're also making steady progress increasing CPU supply. Through our investments, focused execution and tighter customer collaboration, we expect our PC CPU supply will be up mid single-digits this year, while we expect the PC TAM to grow slightly. We'll continue to work with our customers to meet their required product mix and ramp additional capacity to ensure we are not a constraint on their growth.”

Intel intends to have its new chips in mobile PCs by the holiday season. The new chips, presumably Ice Lake-U and Ice Lake-Y, are based on the new Sunny Cove architecture and are reported to bring an 18% raw IPC increase over Skylake. Noticeably absent from the earnings call was when we might see Sunny Cove and 10nm for desktop.

Dell already has a refreshed model from its XPS line that is the first notebook (or convertible, rather) offering new 10nm Ice Lake chips. The XPS 13 2-in-1 is configurable with either the i3-1005G1, i5-1035G1, or i7-1065G7. The new XPS 13 2-in-1 model will start at $1000, and orders will ship in early September.

Source: https://www.anandtech.com/show/14679/intel-begins-commercial-shipments-of-10-nm-ice-lake-cpus-to-oems

https://www.intc.com/investor-relations/investor-education-and-news/investor-news/press-release-details/2019/Intel-Reports-Second-Quarter-2019-Financial-Results/

https://hothardware.com/news/dell-xps-13-2-in-1-now-on-sale-with-intel-10th-gen-ice-lake-cpus

UserBenchmark Addresses Controversy, Sort Of

UserBenchmark courted no shortage of enthusiast ire when the website recently made some bizarre changes to its algorithm for weighing average CPU scores. The new score weighting for average scores is as follows: 40% single-core/thread, 58% quad-core, and 2% multi-core. This is in contrast to the previous system, where the score weight consisted of 40% single-core, 50% quad-core, and 10% multi-core.

The new scoring weights affect all processors, but impact AMD’s new Ryzen 3000 in particular. UserBenchmark claimed the changes were necessary to offset the “unrealistic overestimating” for processors with core counts beyond 8. This seems to skew results in favor of lower core count CPUs with higher frequencies, and has especially upset ardent AMD fans.

In an attempt to address the outcry, UserBenchmark updated its FAQ section with a special note regarding the “AMD community.” UserBenchmark has since retraced that statement -- of which Tom’s Hardware still has (see source linked below) -- and replaced it with this one.

“Following the July 2019 cohort of new CPUs, we noticed that our CPU gaming and desktop indices were overestimating all CPUs with core counts higher than eight so both indices were rebalanced. Neither the underlying data (single, quad, multi-64) nor the workstation ranks were impacted.

The Ryzen 3000 gaming rank changes were net positive: 3900X -2, 3800X +7, 3600X +14, 3600 +13. The 2000 series 32-core Threadripper 2990WX moved from first position to 48th and the 8-core Intel 9900K took the top position in the gaming index up from its previous rank of 7.

We frequently tune our effective speed indices and expect to add an octa core component to the index in due course.”

For the record, we believe UserBenchmark is useless, anyway, as it basically amounts to an SEO spam site that distributes an uncontrolled benchmark to run on random PCs rather than actually doing work testing the hardware, so no real loss here.

Source: https://www.tomshardware.com/news/userbenchmark-benchmark-change-criticism-amd-intel,40032.html

https://cpu.userbenchmark.com/Faq/What-is-the-effective-CPU-speed-index/55

AMD Launches EPYC 7002 Series with New Customers

AMD officially unveiled its second-generation EPYC Rome processors, marking the world’s first 7nm datacenter CPU, and the second time AMD has wrestled a process node lead from Intel -- the latter of which cannot be overstated. EPYC Rome, coming in the form of the EPYC 7002-series, pairs TSMC’s 7nm manufacturing process with AMD’s Zen 2 architecture and chiplet-based design.

The chips will use a nine-die approach, with eight 7nm compute dies, and one 14nm I/O die. AMD calls this “Hybrid Multi-die Architecture.” Through AMD’s chiplet design, EPYC 7002 scales up to 64 cores and 128 threads, with the 7nm chiplets being connected through a second-gen Infinity Fabric. EPYC 7002 also bears the distinction of being the first x86 processor for the server space that’s PCIe 4.0 ready, with all chips offering 128 PCIe lanes by default, though custom servers can expose up to 168 lanes. Additionally, the new EPYC chips support eight DDR4 memory channels, beating Intel’s six.

AMD is also maintaining its aggressive price/performance ratio with competitive prices for the new EPYC 7002 product stack. The EPYC 7232P base model starts at $450 for 8 cores and 16 threads. Meanwhile, at the top of the stack is the 64C/128T EPYC 7742, listed for $6,950. That may not sound like much of a bargain, until weighed against Intel’s 28C/56T Xeon 8280 that goes for $10,000.

AMD has spent a lot of time cozying up to various customers in the industry, many of the same ones Intel has been courting for years. With the launch of EPYC 7002, AMD was joined by several big players such as Google, Twitter, Microsoft, Dell, HPE, Cray, and Lenovo, all of which are adopting EYPC in varying capacities. This serves to illustrate just how far AMD has come in fostering a broad partner ecosystem for EPYC.

Source: https://www.amd.com/en/press-releases/2019-08-07-2nd-gen-amd-epyc-processors-set-new-standard-for-the-modern-datacenter

Power Outage Costs WD $339 Million

In past episodes, we mentioned a 13 minute power outage that affected both Western Digital and Toshiba Memory, and that short term NAND prices could see price hikes as a result of several exabytes of NAND being lost.

Now, it seems Western Digital and Toshiba have recovered from the unexpected power interruption and fabs have resumed normal operation. However, the damages will come in at $339 million for Western Digital. The damages include not only wafers, but equipment as well. Toshiba did not disclose any details about losses, but we can assume the company is in a similar boat.

Source: https://www.anandtech.com/show/14702/western-digital-nearly-all-nand-capacities-resumed-normal-operations

Cryorig Still In Business

We’ve seen many comments on Reddit and on our HW News videos regarding the status of Cryorig, mostly due to the radio silence of the company’s social media accounts and Cryorig products being out of stock at online retailers. Many have feared the company had gone under; however, thanks to an exclusive over at TechPowerUp, TPU was able to confirm that’s not the case.

TPU’s most prolific news writer, btarunr, wrote the following:

“We reached out to Cryorig and one of their representatives was kind enough to respond to us with an update on what has happened at the company. Cryorig is impacted by the U.S.-China trade-war, as the high import tariffs affected the viability of its products. The company would earlier directly access the U.S. market through exclusive stores on Amazon and Newegg,” says TechPowerUp.

While Cryorig has been biding its time waiting for relief from the US-China trade war, it has maintained an active presence in other markets such as Asia and Europe. And while TechPowerUp reports that Cryorig will not exit the US market, the company will have to access the market through an intermediary.

Cryorig will sell its products through Outlet PC, who will also handle support and RMAs. TechPowerUp reports that the transition to Outlet PC should be completed in early August.

Source: https://www.techpowerup.com/257192/cryorig-not-dead-but-the-us-china-trade-war-hurt-it

XFX RX 5700 XT THICC BOI - Yes, Really

If you remember the XFX RX 590 FATBOY video card, it should be no surprise that XFX has decided to promote the 13-year-old in charge of naming and brand its next card the THICC II. If you’d forgotten, XFX is the company that fought a losing, public argument with us on twitter that it was selectively enforcing “warranty void if removed” stickers in countries where it could get away with them, a response we got when we publicly asked why they use the illegal stickers on cards in the US. The XFX 5700 XT THICC II will likely launch next week along with the glut of other RX Navi cards coming out. It’s supposed to be a dual-fan partner model 5700. We’ll look at some cards, but probably not this one.

https://videocardz.com/newz/xfx-radeon-rx-5700-xt-thicc2-pictured

Editorial: Eric Hamilton
Host: Steve Burke
Video: Andrew Coleman