Last year, AMD released its Ryzen line of processors, but it didn’t stop there. Later in August 2017, AMD launched three impressively overclockable high-end desktop CPUs – named ‘Threadripper’.
However, AMD didn’t just sit on its laurels after releasing the fastest processor we’ve used. AMD has also suggested that bitcoin miners use Threadripper CPUs for cryptocurrency mining, to save money on future upgrades – likely due to the out-of-this-world compute power of Threadripper.
And, now that AMD Ryzen Threadripper 2nd Generation is on its way later this year, we’re probably going to be seeing improved performance, similar to what we saw when Ryzen 2nd Generation succeeded AMD Ryzen.
So, here’s all the information we have on AMD Ryzen Threadripper, from the release date to price evaluation and other useful information.
AMD Ryzen Threadripper first two processors the 1950X and 1920X arrived on August 10, 2017. While AMD’s entry level Ryzen Threadripper 1900X came out a few weeks later on August 31, 2017.
We’re still waiting on any news regarding the AMD Ryzen Threadripper 1900, 1920 and 1950 chips, but given their existence has only been leaked through motherboard manufacturers – a concrete release date is difficult to come by, assuming they ever release at all.
It doesn’t end there, though, as AMD plans to support its Threadripper line of CPUs with new releases until at least 2020, according to some leaks.
At long last, we now have firm details on what price the AMD Ryzen 9 Threadrippers will release at. They’re not cheap, but you can expect to pay less for more performance than you would get with a comparable Intel chip.
Whereas the 16-core Ryzen Threadripper 1950X retails at $999 (around £770, AU$1,295), the Threadripper 1920X will boast 12 cores and sells for $799 (around £615, AU$1,035). Lastly, the Threadripper 1900X comes with a $549 (about £420, AU$690) price.
Meanwhile Intel’s closest competitor to the 1950X, the Core i9-7960X, sets users back a whole $1,699 (around £1,315, AU$2,200) while the 12-core i9-7920X is $1,189 (around £920, AU$1,500). The Threadripper 1900X's closest octa-core competitor from Intel is the Core i7-7820X that retails for $599 (£529, AU$869).
We officially know the AMD Ryzen 9 Threadripper series features "up to 16 cores and 32 threads", with the Ryzen Threadripper 1950X serving as the posterchild for these specs. The 1920X, on the other hand, is limited to 12 cores and 24 threads. Lastly, the 1900X serves up 8 cores and 16 threads.
As for how fast those cores will run, the Ryzen Threadripper 1950X can be found running at a base clock of 3.4GHz and a boost clock of 4.0GHz. At the same time, the Threadripper 1920X runs at a base/boost clock of 3.5/4.0GHz. Finally, the 1900X runs at a base/boost clock of 3.8/4.0GHz.
To put this into perspective, the Intel Core i9-7960X has packs the same 32 threads and 16 cores as the Threadripper 1950X and so on and so forth with the Intel Core i9-7920X vs Threadripper 1920X and Intel Core i7-7820X vs Threadripper 1900X we've setup.
Then there are the XFR-less AMD Ryzen Threadripper 1900, 1920 and 1950. First up, the 1900 will run at 3.1/3.7GHz base/boost clock speeds while the 1920 will operate at 3.2/3.8GHz. Though the 1950’s boost speeds have yet to surface, this chip will run at a base clock speed of 3.2GHz. All of these contain the same amount of cores as the 1900X, 1920X and 1950X.
The Threadripper series will use AMD’s SMT (Simultaneous Multi-Threading) technology found in other Ryzen chips. Finally, the Ryzen Threadripper series processors will use a TP3 socket, which is a modified variant of the SP3 socket originally designed for AMD’s beefy server chips that go up to 32 cores.
Future Intel processor lineups have included almost a half dozen rumored codenames, but at long last we could finally have the details pinned down thanks to a leaked Intel roadmap.
Xfastest posted slides of what it claims to be an official Intel roadmap outlining the processor release schedule for the coming year. Going by the schedule, and disappointingly, we shouldn’t expect 9th generation Intel processors until the first few months of 2019. The Intel roadmap lists that the Core i9-9900K, Core i7-9700K and Core i5-9600K will all release in the first quarter of 2019.
The slides also confirm that 9th-generation Coffee Lake Refresh processors will be compatible with Intel 300-series chipsets, so you’ll supposedly be able to plug any of these next-gen CPUs into something as lowly as a H310 motherboard.
What will be releasing within the year are Intel’s next high-end desktop processors, if this purportedly official roadmap is to be believed. Interestingly, the slides refer to Intel’s new HEDT family as Basin Falls Refresh, rather than Skylake-X Refresh or Cascade Lake-X as has previously been rumored.
The roadmap barely reveals any details, but we do see Basin Falls Refresh dominating Intel's entire release schedule from later this year to mid-2019. It also appears as if Basin Falls Refresh will replace every Core X and Extreme processor from the Core i5-7640X to the defunct Kaby Lake-X family to Intel’s highest-end consumer CPU, the Core i9-7980XE.
While these slides look more authentic than previous rumors, there’s no way to confirm their validity. There’s also no date attached, so even if they are genuine, the information contained within could be outdated. Previously we heard that Intel may even announce Coffee Lake Refresh processors tomorrow (August 1), although this seems unlikely at this point.
Via Segment Next
Anyone looking at a high-end Intel system has three choices: Core i9, Xeon W, or the larger socket Xeon Scalable. Those first two both use the LGA2066 socket, and have identical core/frequency configurations, but are in effect different platforms with locked motherboards for each. The benefits of the Xeon W and Xeon Scalable lie in the ability for ECC memory, vPro management features, and with some processors there are different cache variants.
For gaming PCs that push the pretty pixels on the screens, the video card is the most important component. And given the sheer amount of custom options, choosing the right graphics card for your budget can be very difficult. In our Video Cards for Gaming guides, we give you our recommendations in terms of GPU models and current prices representative of an affordable non-blower custom card. Our guide targets common gaming resolutions at system-build price points similar to our CPU guides.
If you’re curious about how well Intel’s upcoming Core i9-9900K flagship 9th-generation processor will perform, then a new leaked benchmark will definitely be of interest, as it shows a very nippy CPU indeed.
Bearing in mind the usual caveat of this potentially being fake – which is always a possibility with these sort of pre-launch rumors – the result from 3DMark (as highlighted by TUM APISAK) shows the Core i9-9900K hitting an overall score of 9,862 in the Time Spy benchmark. The chip recorded a graphics score of 9,725 and a CPU score of 10,719.
For the benchmark, the processor was nestled in an Asus Z370-F Strix Gaming motherboard – which incidentally confirms backwards compatibility with existing Z370 boards, as previously speculated – and paired with a Gigabyte GeForce GTX 1080 Ti graphics card.
The result represents an impressive turn of speed, and as Wccftech.com, which spotted this leak, points out, it beats out AMD’s Ryzen 7 2700X by more than 1,500 points.
This leak also confirms the spec – to a large degree – and like the Ryzen 7 2700X, Intel’s flagship is an eight-core (16-thread) processor. It also suggests that the Core i9-9900K is capable of boosting to 5GHz out of the box, although apparently that will only be across two cores (4.7GHz will supposedly be the maximum boost across all cores by default).
What’s different with this leak is the base clock isn’t 3.6GHz as previously rumored, but 3.1GHz. That could throw some doubt on the validity of the benchmark for some folks, but as this early leak is likely from an engineering sample of the CPU, it’s not really surprising to see such a relatively low figure for the base clock.
What it also underlines is that the final production chip will be even faster when it comes to benchmarking. Exciting stuff, particularly when you consider all this comes in a package with an alleged TDP of 95W (again beating out the Ryzen 7 2700X, which weighs in at 105W).
If all this speculation pans out, that is. We’re keeping our fingers firmly crossed.
ASUS has published a list of its upcoming motherboards powered by Intel’s yet-not-announced Z390 chipset. As expected, the company will offer several families of platforms featuring over a dozen of models aimed at different market segments and supporting different features.
ASUS intends to offer Intel Z390-based motherboards across five families of its products. The top-of-the-range Z390 offerings aimed at enthusiasts and overclockers will be marketed as the ROG Maximus XI series. For gamers and enthusiasts ASUS will offer four ROG Strix Z390 mainboards in different form-factors. For gamers with lower budgets the company will release eight Z390 motherboards under the TUF and Prime brands. In addition, there will be a single Z390 Dragon mainboard that will be available only in China.
| ASUS' Upcoming Z390 Motherboards | |||||||
| Form-Factor | ROG Maximus | ROG Strix | TUF | Prime | Dragon | ||
| ATX | ROG Maximus XI Apex | ROG Strix Z390-E Gaming | TUF Z390-Plus Gaming | Prime Z390-A | Z390-Dragon | ||
| ROG Maximus XI Code | ROG Strix Z390-F Gaming | TUF Z390-Plus Gaming Wi-Fi | Prime Z390-P | - | |||
| ROG Maximus XI Extreme | ROG Strix Z390-H Gaming | TUF Z390-Pro Gaming | - | ||||
| ROG Maximus XI Formula | - | - | |||||
| ROG Maximus XI Hero | |||||||
| ROG Maximus XI Hero Wi-Fi | |||||||
| Micro-ATX | - | TUF Z390M-Pro Gaming | Prime Z390M-Plus | ||||
| TUF Z390M-Pro Gaming Wi-Fi |
- | ||||||
| Mini-ITX | ROG Strix Z390-I Gaming | - | |||||
The list of ASUS motherboards was published in a description to a new BIOS feature that was introduced on July 23. The advertised BIOS feature enables to create user profiles in BIOS and therefore better tailor PCs for particular users or use cases. The capability is supported not only on the upcoming Intel Z390-based motherboards, but also on existing Intel 300-series platforms from ASUS.
The fact that the world’s largest maker of motherboards accidentally publishes the list of its Intel Z390-based platforms indicates that the product family has been formed and the company is probably nearly ready to ship the products. Meanwhile, it is highly likely that Intel will launch its Z390 chipset along with its upcoming eight-core Coffee Lake processor*, so it is going to take a while.
*Unless Intel decides to repeat its 2014 launch tactics and release its new enthusiast-class class chipset a month ahead of a new enthusiast-class CPU (Z97/H97 were released in early June, whereas the Devil’s Canyon CPUs were launched in late June). The rationality of such tactics poses questions because Intel’s upcoming eight-core chips in LGA1151 v2 packaging are likely to be supported by various Cannon Point-based chipsets.
Related Reading:
Source: ASUS (thanks to our reader SH SOTN for the tip)
Intel released the new "Bean Canyon" NUC family featuring Coffee Lake-U CPUs and Iris Plus Graphics 655 GPUs at a higher 28W TDP and performance.
Acer prepares its new gaming PC powered by AMD's upcoming Ryzen 5 2500X processor.
ASUS has introduced its first Intel C246-based motherboard for Intel’s Xeon E processors. The WS C246 PRO platform is aimed at rather serious machines that use up to four PCIe 3.0 x16 expansion cards, up to eight SATA storage devices, and two M.2 SSDs. For those who plan to use SAS storage with the motherboard, ASUS will also offer optional eight-port SAS 12G RAID cards.
The ASUS WS C246 PRO motherboard is based on Intel’s C246 chipset (codenamed Cannon Point-H WS) and therefore supports Intel’s Xeon E-2100-series processors for entry-level servers and workstations as well as Intel’s 8th Gen Core i3/i5/i7 chips along with their low-cost Pentium Gold and Celeron counterparts. The mainboard is outfitted with an eight-phase CPU VRM that uses solid-state inductors and capacitors to maintain steady operation during protracted workloads and also to ensure a long lifespan for the product. The WS C246 PRO is not equipped with a TPM module, but there is a connector for it, so workstation OEMs can install it themselves. As for remote management, the platform supports the WFM 2.0 spec as well as ASUS’ proprietary ACC IT management platform.
Memory support is something that all Intel C246-based motherboards are going to have in common. Just like the rest of the C246 pack, the WS C246 PRO has four DDR4 DIMM slots supporting up to 64 GB of DDR4-2667 DRAM with or without ECC, depending on the CPU and modules used.
Things start to get interesting with PCIe slots configuration of the WS C246 PRO. The motherboard has four PCIe 3.0 x16 (electrical) connectors that can work in x16/x8, x8, x4 and x4 modes (logical). The first two are controlled by the CPU, whereas the remaining two rely on the C246 PCH. One of the reasons why ASUS outfitted the motherboard with so many electrical PCIe 3.0 x16 slots is the fact that it will supply the motherboard with its PIKE II 3008-8i 8-port SAS 12G RAID or PIKE II 3108-8i 8-port SAS 12G HW RAID controllers to parties interested in using SAS storage devices. Both HBA cards use a PCIe 3.0 x8 interface, so Intel needed to ensure that there are enough slots not only for SAS controllers, but for whatever other cards a customer may need to use (a high-end PCIe SSD, for example). In addition to PCIe 3.0 x16 slots, there are also two PCIe 3.0 x1 slots.
Moving on to other storage capabilities of the WS C246 PRO. The motherboard comes with eight SATA connectors as well as two M.2 slots supporting an up to M.2-2280 and an up to M.2-22110 SSDs. ASUS does not detail how the latter work and share resources with PCIe slots as well as SATA ports, but keep in mind that C246 PCH only supports eight SATA ports in total.
When it comes to general I/O interfaces, the WS C246 PRO features a fairly standard set of connectors, including two GbE ports, four USB 3.1 Gen 1 Type-A headers (there are internal ones too), USB 3.1 Gen 2 Type-A & Type-C receptacles, four display outputs (DP, HDMI, DVI, D-Sub), 3.5-mm jacks for a 7.1-channel audio subsystem, and so on.
ASUS did not touch upon pricing details and availability timeframe of the WS C246 PRO motherboard in its announcement, but keeping in mind market specifics, expect the first workstations based on this platform to arrive in the coming months. As for pricing, this is a specialized premium product and will be priced accordingly, especially when bundled with SAS HBAs.
| ASUS Intel C246 ATX Motherboard | |||
| WS C246 PRO | |||
| Supported CPUs | LGA1151 v2 CPUs Xeon E 8th Gen Core i3/i5/i7, Pentium, Celeron |
||
| PCH | Intel C246 (Cannon Point-H WS) |
||
| Graphics | Intel UHD Graphics from CPU PCIe 3.0 x16 slot |
||
| Display Outputs | 1 × DisplayPort 1.2 1 × HDMI 2.0a 1 × DVI-D 1 × D-Sub |
||
| Memory | 4 × DDR4 DIMM Up to 64 GB of DDR4-2667 with or without ECC, depending on CPU |
||
| Slots for Add-In-Cards | Electrical | Logical | |
| 1 × PCIe 3.0 x16 1 × PCIe 3.0 x8 2 × PCIe 3.0 16 1 × PCIe 3.0 x1 |
x16/x8 x8 x4 x1 |
||
| Ethernet | LAN 1: Intel I219LM GbE PHY LAN 2: Intel I210AT GbE PHY |
||
| Storage | M.2 | 1 × M.2-22110 (PCIe 3.0 x4 or SATA?) 1 × M.2-2280 (PCIe 3.0 x4 or SATA?) |
|
| SATA | 8 × SATA 6 Gbps | ||
| SAS (Optional) |
ASUS PIKE II 3008-8i 8-port SAS 12G RAID ASUS PIKE II 3108-8i 8-port SAS 12G HW RAID |
||
| Audio | 7.1-channel audio with analog and S/P DIF outputs | ||
| USB | 4 × USB 3.1 Gen 1 Type-A 1 × USB 3.1 Gen 2 Type-A 1 × USB 3.1 Gen 2 Type-C |
||
| Serial Ports | 1 × internal COM port | ||
| Other I/O | 1 × TPM header 1 × internal USB 3.1 Gen 1 Type-A (for BIOS flashing) 2 × internal USB Type-A (for front panel) |
||
| Monitoring | CPU Temperature Fan RPM |
||
| Management | WfM 2.0, DMI 3.0, WOL by PME, WOR by PME, PXE | ||
| Additional Features | ASUS Control Center (ACC), an integrated IT management platform for remote monitoring, controlling, and updating ASUS commercial products (servers, workstations, etc). | ||
| Form-Factor | ATX (305 mm × 244 mm | 12" × 9.6") | ||
Related Reading:
All across the hardware industry 2018 is shaping up to be a banner year for earnings, and the 800lb gorilla that is Intel is no exception. Reporting their Q2’2018 earnings yesterday afternoon, Intel booked $17.0B in revenue for the quarter, setting another new Q2 revenue record for the company. Meanwhile net income, while not being noted as a record for the company, was still very high at $5.0B. This puts Intel well ahead of itself both on a year-over-year and sequential basis, topping Q1’2018 and Q2’2017, both of which were also record quarters. All told, Intel’s both Intel’s revenue and net income for the quarter increased by $2.2B year-over-year, while against Q1 they’re up $0.9B and $0.5 respectively.
Meanwhile Intel’s all-important gross margin has held relatively steady at 61.4%, which is down two-tenths from Q2’2017, but up from the previous quarter.
| Intel Q2 2018 Financial Results (GAAP) | |||||
| Q2'2018 | Q1'2018 | Q1'2017 | |||
| Revenue | $17.0B | $16.1B | $14.8B | ||
| Operating Income | $5.3B | $4.5B | $3.8B | ||
| Net Income | $5.0B | $4.5B | $2.8B | ||
| Gross Margin | 61.4% | 60.6% | 61.6% | ||
| Client Computing Group Revenue | $8.7B | -3% | +6% | ||
| Data Center Group Revenue | $5.5B | +6% | +27% | ||
| Internet of Things Revenue | $880M | +5% | +22% | ||
| Non-Volatile Memory Solutions Group | $1.1B | +10% | +23% | ||
| Programmable Solutions Group | $517M | +4% | +18% | ||
Breaking down the numbers by Intel’s various business groups, all of Intel’s groups saw revenue growth for the year. This ranges from 6% for the client computing group, up to 27% for the data center group. And even on a less-comparable quarterly basis, every group except CCG has seen some growth.
Overall Intel has been making a concentrated effort to pivot towards (or at least play up) what the company sees as the segments primed for growth – which Intel is calling its “data-centric” business – and isolate it from the “PC-centric” client computing group. Looking at revenue growth it’s easy to understand why; client computing revenue has been slow to move for years now, while Intel’s other businesses have grown much faster, and as an added kicker are generally more profitable too. None the less, Intel can’t ignore the fact that the client computing group alone makes up more than half of the company’s revenue. However at the rate things are going, this might be one of the last quarters where that is true.
On the client computing front, Intel is citing the commercial and enthusiast segments as being the driving factors for the growth, which is unsurprising as the two have been consistent performers with higher-than-average spending. Beyond that, Intel hasn’t broken down client computing revenue by form factor or such this quarter, so unfortunately we don’t get any insight into how the recently launched Coffee Lake mobile parts have done.
Besides being a more stagnant space, the biggest challenge Intel faces with the client computing group right now is in fact the company’s biggest challenge, which is the stalled development of their 10nm process. As part of today’s earnings call the company has announced that they are still on-track for 10nm mass production in 2019, however they aren’t expecting complete systems to reach store shelves until the holiday shopping season. Which means Intel will be relying on their 14nm technology and resulting products for the next 12 to 15 months, which is going to make it harder to entice already fickle consumers into upgrading.
The news is of course happier with Intel’s other business, all of which have grown at a much greater rate. Intel continues to sell Xeon processors to server vendors and cloud server providers at a good clip, with revenues there improving by 27% year-over-year. However Intel did miss their own projections here, as the unit didn’t grow by quite as much as the company once predicted. There’s clearly still very high demand for server processors, but Intel is going to be facing off with AMD and EPYC, which is finally finding its footing. And in the longer term, according to Intel’s own accidentally published roadmaps, the company won’t have 10nm-class Xeons read until 2020, which means Intel will likely be playing defense against AMD starting in 2019.
The rest of the “data-centric” business is a similar upbeat story. FPGA sales are up, non-volatile memory sales are up, and even IoT sales are up. The reasons are varied, but overall NAND prices have been good and the drive for neural networking hardware is helping sales on both the FPGA (cloud) and IoT (edge) sides.
Overall then, based what has now been two strong quarters of growth for Intel, the company is expecting the rest of the year to go similarly. And indeed at this point they’re expecting 2018 as a whole to be their third consecutive record year. Which is especially impressive at a time where Intel’s core manufacturing advantage has increasingly slipped away. Conversely however, there remains the lingering question of whether Intel can sustain this kind of growth for the next year without their next-generation fabrication technology.
Intel has confirmed that its long-awaited 10-nanometer Cannon Lake processors will be arriving in volume in the second half of 2019, with consumer PCs carrying the CPUs to be on sale before the year is out (just in time for the holidays, the firm says).
The news came via a conference call with the company following its Q2 fiscal results. If you’ve been following the saga of when Intel will drop down to 10nm Cannon Lake (from the current 14nm processors), you’ll recall that the launch of these chips has been delayed considerably – they were first expected to ship in 2016 – and the latest we heard in April was that they had been delayed to 2019.
So, this isn’t another delay as such, just a confirmation of exactly when in 2019 the launch will happen – but it’s kind of a disappointment seeing as the timeframe sounds pretty much as late as it can be (with consumer PCs launching by December).
Intel isn’t worried though (naturally), and as PC World reports, the firm is confident it can still get more out of its 14nm products in the meantime – Whiskey Lake and Amber Lake are due to land later this year using a fresh spin on the 14nm++ process to eke out some more gains. Intel is confident that these efforts will help it maintain “system leadership”, company executives asserted (as they are duty-bound to).
Note that very small amounts of 10nm chips have started shipping – in the form of low-end processors for Asian laptop manufacturers – but volume production is what really counts, and this is what won’t happen until later on in 2019.
As we’ve heard before, one of the main reasons for the Cannon Lake delay has been the difficulty in refining and honing 10nm yields to be able to hit mass production, which has evidently been a tough task.
It’s also likely that security worries have been laying heavy on the minds of Intel’s engineers, given the whole Spectre and Meltdown debacle, and doubtless there’s been a lot of tightening done on the security front with these processors.
Earlier this year Intel confirmed that it would delay mass production of 10nm CPUs to 2019 due to issues with yields, but did not elaborate on when in 2019 that would be. Late on Thursday Intel finally made a long-awaited clarification: the first systems powered by 10nm CPUs will hit store shelves in the second half of 2019. But there is a catch: 14nm products will still continue to play a big role next year.
Intel is set to start mass production of processors using its 10nm process technology in 2019, and while the company isn't explicitly saying when mass production will begin, it's sounding like that will be in the Q2/Q3 time frame. This being based off of what Intel is saying, which is that they expect to have PCs based on these chips to arrive during the 2019 holiday shopping season. Given the kind of lead-time required to get complete OEM systems on to store shelves, this would mean that mass production on chips needs to start 3 to 6 months sooner than that.
As for said finished systems, what's typically considered the holiday shopping season starts in October or early November. So systems based on Intel’s 10nm CPUs would finally be available in mass quantities starting in Q4 of 2019. Consequently, for more three quarters of next year Intel will continue to rely almost entirely on its products made using its various 14nm process technologies.
It remains to be seen how much time it will take Intel to ramp up production of its 10nm CPUs and when the volume crossover between 14nm and 10nm chips occurs. At present, Intel seems to be optimistic about what it has today and will have tomorrow — a variety of 14nm products. In fact, there is a lot of room for growth here: shipments of Intel’s Xeon Scalable processors accounted for a little less than 50% of Intel’s datacenter revenue in the most recent quarter, so the company will certainly continue to ramp up shipments of these products for many quarters to come, increasing its ASPs and revenue. Keep in mind that Intel’s upcoming Cascade Lake-SP platform will bring support for 3D XPoint-based Optane DC Persistent Memory DIMMs, which will bring in huge earnings opportunities to the company.
Speaking of datacenter products. Murthy Renduchintala, Intel’s chief engineering officer, promised that 10nm “datacenter products” will follow “shortly after” availability of client systems running 10nm CPUs in the late 2019. He did not give any exact timeframes, but stressed that its next-gen Xeon will follow its next-gen client parts quicker usually. Essentially, Intel confirmed this week’s report claiming that the codenamed Ice Lake-SP CPUs may hit the market in Q3 or Q4 of 2020.
Note: Image of a wafer is for illustrative purposes only.
Related Reading:
Yes, the Acer Swift 7 is the world’s thinnest laptop, and it manages good battery life. But its svelte styling comes with a last-gen CPU, a touchpad with no click, and an oddly jumbled keyboard.
Along with AMD’s epic quarterly results that were announced yesterday afternoon, there was also an interesting nugget buried in the earnings call for the results. We’ve known for some time now that AMD was going to be working with both TSMC and fraternal fab GlobalFoundries to produce 7nm products, similar to what they’ve done for their current-generation 16nm/14nm products. However until now it hasn’t been clear how those products would be allocated – whether TSMC will once again just produce semi-custom APUs, or if they’ll do more. But now we have an answer thanks to AMD’s earnings call, and the answer is indeed “more”.
Specifically, AMD CEO Lisa Su has announced that second-generation “Rome” EPYC CPU that the company is wrapping up work on is being produced out at TSMC. This is a notable departure from how things have gone for AMD with the Zen 1 generation, as GlobalFoundries has produced all of AMD’s Zen CPUs, both for consumer Ryzen and professional EPYC parts.
So, Matt, on your first question relative to the manufacturing of the second generation of EPYC, so as I said earlier, we are working with both the TSMC and GLOBALFOUNDRIES in 7-nanometer. As for the 7-nanometer Rome that we're currently sampling, that's being manufactured at TSMC.
As it stands, AMD seems rather optimistic about how things are currently going. Rome silicon is already back in the labs, and indeed AMD is already sampling the parts to certain partners for early validation. Which means AMD remains on track to launch their second-generation EPYC processors in 2019.
Unfortunately, the call offered no real insight as to why AMD has shifted to using TSMC for some of their CPUs for this upcoming generation. AMD’s hands are somewhat tied on fab choice in as much as they have quotas under their wafer share agreement with GlobalFoundries that they must fulfill. How AMD wants to split that up is up to them, but beyond that AMD hasn’t offered any information as to why they’ve made the choices they have.
Ultimately however if they are meeting their order quota from GlobalFoundries, then AMD’s situation is ultimately much more market driven: which fab can offer the necessary capacity and performance, and at the best prices. Which will be an important consideration as GlobalFoundries has indicated that it may not be able to keep up with 7nm demand, especially with the long manufacturing process their first-generation DUV-based 7nm “7LP” process requires.
What a difference a year makes. Yesterday afternoon AMD announced their second quarter results for their fiscal year 2018, and the news is nothing short of stunning, with AMD having one of their best quarters in seven years. AMD already started off the year profitable, but things have continued to ramp up in Q2, marking the 1 year anniversary of the Zen CPU architecture launch, and the inflection point for AMD’s badly-needed recovery.
| AMD Q2 2018 Financial Results (GAAP) | |||||
| Q2'2018 | Q1'2018 | Q2'2017 | |||
| Revenue | $1756M | $1647M | $1151M | ||
| Gross Margin | 37% | 36% | 34% | ||
| Operating Income | $153M | $120M | -$1M | ||
| Net Income | $116M | $81M | -$42M | ||
| Earnings Per Share | $0.11 | $0.08 | -$0.04 | ||
For Q2 AMD booked $1.76B in revenue, their greatest revenue in quite some time and beating not just their seasonally strong Q4 from last year, but also the cryptocurrency-fueled Q1 of this year. This is a massive 53% jump in revenue year-over-year, while even on a quarterly sequential basis that’s a 7% increase. As a result, both AMD’s profits and gross margins were up. Net income rose by $158M year-over-year to $116M – a far cry from the money-losing situation of AMD a year ago – while gross margins have risen to 37%.
| AMD Q2 2018 Financial Results (Non-GAAP) | |||||
| Q2'2018 | Q1'2018 | Q2'2017 | |||
| Revenue | $1756M | $1647M | $1151M | ||
| Gross Margin | 37% | 36% | 34% | ||
| Operating Income | $186M | $152M | $23M | ||
| Net Income | $156M | $121M | -$7M | ||
| Earnings Per Share | $0.14 | $0.11 | -$0.01 | ||
Looking at the non-GAAP numbers, which exclude results such as restructuring charges, debt fees, and stock based compensation, which are also strong. AMD continues to pay down its sizable long-term debt and issue stock as compensation, both of which bring down the GAAP numbers, but not extensively. More importantly it shows that AMD’s GAAP numbers themselves aren’t a fluke, and that AMD’s core business is stronger than it has been in quite some time.
AMD is attributing the great quarter to what can best be described as all-around improved demand for products. Ryzen, Epyc, and Radeon shipments were all well up year-over-year, and even on a sequential basis AMD’s CPUs continue to be in high demand. AMD’s specifically citing the ramp of new products here (presumably Ryzen 2) as being especially beneficial towards the gross margin, and ultimately their bottom line.
| AMD Q2 2018 Computing and Graphics | |||||
| Q2'2018 | Q1'2018 | Q2'2017 | |||
| Revenue | $1086M | $1115M | $661M | ||
| Operating Income | $117M | $138M | $7M | ||
Digging into the individual segments a bit more, the larger of AMD’s two segments, Computing and Graphics, booked $1.09B in revenue for the quarter. This is a 64% jump in revenue year-over-year, though a slight decline sequentially. Operating income was $117, which is over $100M better than last year, but again a bit of a drop over Q1. AMD credits the year-over-year shift to the significant growth of Ryzen products – this being the first Q2 where AMD has a top-to-bottom CPU and APU stack in place for desktop and mobile – while Radeon is a bit more of a mixed bag. AMD’s strong Q1 performance was in part built on top of the red-hot cryptocurrency market, which has significantly cooled in the last couple of months. This being why AMD’s numbers are down a bit on a quarterly basis, as overall GPU revenue has dropped.
The April launch of Ryzen 2 CPUs meant that AMD has seen “double digit” shipment growth in Ryzen processors over the previous quarter, on top of growth of Ryzen mobile APUs, where ships actually more than doubled. Consequently desktop processor ASPs are down, but mobile processor ASPs are actually up, presumably as customers shift away from bargain-priced Bristol Ridge parts. Interestingly GPU ASPs are also up, despite the cooling crypto market, with AMD citing datacenter sales as pulling ASPs up.
| AMD Q2 2018 Enterprise, Embedded, and Semi-Custom | |||||
| Q2'2018 | Q1'2018 | Q2'2017 | |||
| Revenue | $670M | $532M | $490M | ||
| Operating Income | $69M | $14M | $16M | ||
Meanwhile for the Enterprise, Embedded, and Semi-Custom market, AMD booked $670M in revenue for the year, up $180M from Q2 of 2017. Operating income is also up, though not as wildly as with computing products, rising from $16M to $69M. Notably, this segment includes AMD’s EPYC processors for datacenters. I’d been expecting first-generation EPYC sales to be soft as AMD needed to prove itself to customers who were burnt on the Bulldozer generation, however AMD has been consistently reporting good news here, as EPYC adoption continues to grow. Meanwhile the semi-custom side has also done better than expected thanks to some non-cancellable purchase orders.
Overall then AMD had a very good quarter, and if the company has its way, this is the start of things to come. As energizing as the company’s success with Ryzen has been, the big money for the company is ultimately in high-margin datacenter products, including both EPYC and Radeon Instinct. The next-generation EPYC processors aren’t due until 2019, but there’s still time to grow the current generation, as EPYC revenue was up 50% sequentially over Q1. And indeed AMD is predicting that gross margins will tick up to 38% in Q3. Which is still a far cry from Intel’s traditionally massive margins, but is moving AMD where they need to be to be consistently and significantly profitable. Otherwise, AMD is predicting $1.7B in revenue for Q3, which would end up being very close to this Q2.
