Subject: Processors | November 7, 2018 - 11:00 PM | Tim Verry
Tagged: Zen 2, rome, PCI-e 4, Infinity Fabric, EPYC, ddr4, amd, 7nm
In addition to AMD's reveal of 7nm GPUs used in its Radeon Instinct MI60 and MI50 graphics cards (aimed at machine learning and other HPC acceleration), the company teased a few morsels of information on its 7nm CPUs. Specifically, AMD teased attendees of its New Horizon event with information on its 7nm "Rome" EPYC processors based on the new Zen 2 architecture.
Tom's Hardware spotted the upcoming Epyc processor at AMD's New Horizon event.
The codenamed "Rome" EPYC processors will utilize a MCM design like its EPYC and Threadripper predecessors, but increases the number of CPU dies from four to eight (with each chiplet containing eight cores with two CCXs) and adds a new 14nm I/O die that sits in the center of processor that consolidates memory and I/O channels to help even-out the latency among all the cores of the various dies. This new approach allows each chip to directly access up to eight channels of DDR4 memory (up to 4TB) and will no longer have to send requests to neighboring dies connected to memory which was the case with, for example, Threadripper 2. The I/O die is speculated by TechPowerUp to also be responsible for other I/O duties such as PCI-E 4.0 and the PCH communication duties previously integrated into each die.
"Rome" EPYC processors with up to 64 cores (128 threads) are expected to launch next year with AMD already sampling processors to its biggest enterprise clients. The new Zen 2-based processors should work with existing Naples and future Milan server platforms. EPYC will feature from four to up to eight 7nm Zen 2 dies connected via Infinity Fabric to a 14nm I/O die.
AMD CEO Lisa Su holding up "Rome" EPYC CPU during press conference earlier this year.
The new 7nm Zen 2 CPU dies are much smaller than the dies of previous generation parts (even 12nm Zen+). AMD has not provided full details on the changes it has made with the new Zen 2 architecutre, but it has apparently heavily tweaked the front end operations (branch prediction, pre-fetching) and increased cache sizes as well as doubling the size of the FPUs to 256-bit. The architectural improvements alogn with the die shrink should allow AMD to show off some respectable IPC improvements and I am interested to see details and how Zen 2 will shake out.
Subject: General Tech | November 6, 2018 - 03:42 PM | Jeremy Hellstrom
Tagged: AMD Radeon Instinct, MI60, MI50, 7nm, ROCm 2.0, HPC, amd
If you haven't been watching AMD's launch of the 7nm Vega based MI60 and MI50 then you can catch up right here.
You won't be gaming with these beasts, but for those working on deep learning, HPC, cloud computing or rendering apps you might want to take a deeper look. The new PCIe 4.0 cards use HBM2 ECC memory and Infinity Fabric interconnects, offering up to 1 TB/s of memory bandwidth.
The MI60 features 32GB of HBM2 with 64 Compute Units containing 4096 Stream Processors which translates into 59 TOPS INT8, up to 29.5 TFLOPS FP16, 14.7 TFLOPS FP32 and 7.4 TFLOPS FP64. AMD claims is currently the fastest double precision PCIe card on the market, with the 16GB Tesla V100 offering 7 TFLOPS of FP64 performance.
The MI50 is a little less powerful though with 16GB of HBM2, 53.6 TFLOPS of INT8, up to 26.8 TFLOPS FP16, 13.4 TFLOPS FP32 and 6.7 TFLOPS FP64 it is no slouch.
With two Infinity Fabric links per GPU, they can deliver up to 200 GB/s of peer-to-peer bandwidth and you can configure up to four GPUs in a hive ring configuration, made of two hives in eight GPU servers with the help of the new ROCm 2.0 software.
Expect to see AMD in more HPC servers starting at the beginning of the new year, when they start shipping.
Subject: General Tech | October 25, 2018 - 11:32 AM | Ken Addison
Tagged: turtle beach, seasonic, Samsung, podcast, Intel, Core i9-9900K, amd, 7nm
PC Perspective Podcast #519 - 10/25/18
Join us this week for discussion on the Core i9-9900K, Changes at PCPER, and more.
The URL for the podcast is: http://pcper.com/podcast - Share with your friends!
- iTunes - Subscribe to the podcast directly through the iTunes Store (audio only)
- Google Play - Subscribe to our audio podcast directly through Google Play!
- RSS - Subscribe through your regular RSS reader (audio only)
- MP3 - Direct download link to the MP3 file
Hosts: Jim Tanous, Allyn Malventano, Jeremy Hellstrom, Josh Walrath, and Ken Addison
Peanut Gallery: Alex Lustenberg, Ryan Shrout
Program length: 1:48:01
Podcast topics of discussion:
Week in Review:
0:07:25 Beginning of a new journey
News items of interest:
Picks of the Week:
1:35:35 Jeremy: It’s about damn time, and of course it is Motorola
1:37:30 Josh: Bigger is better… so is 144/FreeSync
1:40:05 Ken: T-Force Delta RGB SSD 500 GB
1:42:20 Allyn: Something to watch - M.2 Optane 905P coming soon
Subject: Editorial | October 24, 2018 - 09:13 PM | Josh Walrath
Tagged: amd, quarterly results, Q3 2018, ryzen, EPYC, Polaris, Vega, 7nm, 12nm, Intel, nvidia
This evening AMD announced their Q3 2018 results. Things were at the lower end of the guidance scale from last quarter, but the company still had some solid results. Q3 revenue was $1.65B as compared to Q3 2017’s $1.58B. It is down from the previous quarter’s high of $1.76B. At first glance this seems troubling, but the results are not as negative as one would assume. GAAP net income was a healthy $102M. Q3 2017 was at $61M while Q2 2018 was up at $116M. Profits did not fall nearly as much as one would expect with a decrease of $110M revenue quarter over quarter.
Probably the largest factor of the decrease was the negligible sales of GPUs to the crypto market. AMD had expected such a dropoff and warned about it in their Q2 guidance. That particular drop off was sudden and dramatic. AMD looks to continue to lose marketshare in add-in graphics due to their less competitive offerings across the spectrum. GeForce RTX sales of course did not impact AMD this previous quarter, but with no new AMD offerings on the horizon users look to have been waiting to see exactly what NVIDIA would release.
Ryzen sales have been steady and strong, making up some of the shortfall from the graphics market. Desktop chips are moving briskly for the company and continues to be a strong seller historically for the company. AMD is also starting to move more mobile processors, but it seems that the majority of parts are still desktop based. AMD looks to continue moving older inventory with aggressive pricing on those and manufacturing of the new 2000 series parts has been relatively smooth sailing for the company.
Enterprise, Embedded, and Semi-Custom had a strong quarter, but with less growth as some analysts had been hoping for. Semi-Custom was weaker this quarter, but IP revenue is up. Console chips are weaker at the moment due to the platforms being relatively mature and not exhibiting the sales of the previous two holiday seasons. To further offset the decrease in Semi-Custom, AMD is reporting that the enterprise products (GPU and EPYC) have seen good growth. Overall this division was down 5% from Q3 2017, but up 7% from the previous quarter.
Perhaps the most interesting figure of this is Gross Margins. AMD was able to improve margins from 36% to 40%. This 4% increase quarter on quarter is a significant jump for the company. This means that AMD continues to keep costs under control for the company and is able to deliver product more efficiently than in the year before. It is still a far cry from Intel and NVIDIA, which typically have magins between 55% to 65%. AMD has a long ways to go before reaching that kind of level. Part of the margin offset was again due to IP licensing. If IP licensing was removed then we would see 38% margins rather than 40%.
So what are the overall lessons of the past quarter? EPYC sales are not as brisk as analysts had hoped for, but they are also not non-existent. It has shown solid growth for the company and has offset shortfalls in other areas of the company. Their IP and Semi-Custom areas are still very solid, even though AMD does suffer from console lifecycles and downturns. GPUs continue to sell, but not nearly at the rate they were due to the crypto market. Their Polaris based options are well suited to compete in the sub-$300 US market. The Vega based products were finally down to MSRP, but they had a harder time going against the mature and well liked GeForce GTX 1070 and 1080 products. This will be further compounded with the introduction of the RTX products in those price ranges.
Ryzen continues to be a very good seller across the board. I had hoped that AMD would break down numbers between Ryzen CPUs and APUs, but I have not seen numbers that hint at what ratio they sell at. In retail the Ryzen 2000 series CPUs look to be some of the most popular products based on price/performance. However, retail is only a small portion of processor sales and Intel still holds the vast majority of marketshare here. AMD is competing, but they have not taken significant chunks from their competition over the past year. They have done enough to achieve several positive quarters in a row, but this is not the slam dunk that the original Athlon 64 was back in 2003/2004.
AMD expects further weakness in their results next quarter. Guidance is for revenue around $1.45B, plus or minus $50M. This is still higher than Q4 2017 results, but it is a significant drop from Q3 results. AMD expects strong Ryzen, EPYC, and datacenter GPU growth during this time. It is expected that consumer GPU and Semi-Custom will continue to drop. There does look to be a 7nm GPU introduction this next quarter, but it is probably the long rumored Vega refresh that will be aimed directly at datacenter rather than consumer.
2018 has so far been a year of solid growth and execution for AMD on the CPU side. Their GPU side has suffered a bit of a slide, but this is to be expected by how much belt-tightening AMD has done in the past several years to get their CPU architecture back on track. The lion’s share of development resources was shunted off to the CPU side while the GPU side had to fight for scraps. I believe this is no longer the case, but when development takes years for new GPUs the injection of new resources will not become apparent for a while.
2019 continues to look better for AMD as they are expecting an early release of 7nm EPYC parts which should compete very well with Intel’s 14nm based Xeon products. AMD is expecting a significant uptick in sales due to the thermals, pricing, and performance of these new Zen 2 based parts. The company also continues to point to the end of 1H for introduction of 7nm Ryzen parts based on Zen 2. These will be showing up quite a few months before Intel’s 10nm offerings will be available. Rumors have it that the new Zen 2 based parts exhibit a significant IPC increase that should make them far more competitive to the best that Intel has on the desktop and mobile markets. Combine these IPC improvements with the 7nm boost in power and clocks for the parts, and AMD could have a very good product on their hands. AMD also is expecting a 1H release of 7nm Navi GPUs which should prove to be more competitive with current NVIDIA products that rely on 16nm and 12nm process nodes from TSMC.
While Q3 was a drop in revenue for the company, their current cost structure has still allowed them to make a tidy profit. The company continues to move forward with new products and new developments.
Subject: General Tech | October 24, 2018 - 01:00 PM | Jeremy Hellstrom
Tagged: navi, amd, rumour, 7nm
The rumours about AMD's new 7nm Navi GPU are continuing to spread, this time via The Inquirer. They have heard tell that the chips are currently being tested, which is good news for both AMD and consumers. We know that new Vega cards will be arriving in the near future but as of now we don't know as much about the release of Navi. If they are currently testing the silicon that could imply some sort of release next year, perhaps not a full lineup but quite possibly something we can take a peek at and see how it stacks up to Turing. Lets hope it is sooner rather than later.
"Touted to be the next graphics accelerators to take the fight to Nvidia with its Turing-based GeForce RTX graphics cards, Navi is set to use a 7nm process which promises to get more power out of GPU silicon."
Here is some more Tech News from around the web:
- 'We Expected VR To Be Two To Three Times as Big', Says CCP Games CEO @ Slashdot
- If you want to rent AMD Epyc bare-metal boxes in the cloud, Oracle hopes you see red @ The Register
- NAND flash prices may see larger drop in 2019 @ DigiTimes
- Zip it! 3 more reasons to be glad you didn't jump on Windows 10 1809 @ The Register
- Another Windows 0-day flaw has been published on Twitter @ Ars Technica
- Apple and Samsung fined over performance-throttling software updates @ The Inquirer
- Grocery Delivery in 2018: Does It Finally Work? @ TechSpot
Subject: General Tech | October 18, 2018 - 01:11 PM | Jeremy Hellstrom
Tagged: Samsung, 7nm
A few short years ago 7nm was a holy grail, something to be sought for but unlikely to be successful. The richest knight in the kingdom is still seeking their goal, while the squires have already got one. AMD is doing well with their process but it is Sir Samsung whom has met with the most success. Their trusty EUV proved up to the task and they are no longer seeking a 7nm process and have moved onto sharing the benefits of their quest with the world. The new chips will be Low Power Plus, and find their way into smartphones, and cell providers, IoT devices and many other applications in the small endian market. The Inquirer posted a look at how they got there, as well as some definitions if this is all Gallic to you.
"Normally argon fluoride immersion tech is used for chip lithography. But Samsung is touting the advantages of EUV in cutting down the number of masks needed to allow for the stencilling of transistors on silicon in a certain pattern, and thus cuts down on the time and costs of producing chips."
Here is some more Tech News from around the web:
- Microsoft will let users bin more of Windows 10's pre-loaded junk @ The Inquirer
- TSMC reports 3Q18 EPS of NT$3.44 @ DigiTimes
- Microsoft's Surface Pro 6 is 'a pain in the butt' to repair @ The Inquirer
- Video: What to expect from the Oculus Quest @ Ars Technica
- Last year, D-Link flubbed a router bug-fix, so it's back with total pwnage @ The Register
- Apple To Announce New iPads on October 30 @ Slashdot
- Once more with feeling: Windows 10 October 2018 Update inches closer to relaunch @ The Register
- The Supercon Badge is a Freakin’ Computer @ Hack a Day
Subject: General Tech | October 16, 2018 - 01:44 PM | Josh Walrath
Tagged: UMC, TSMC, Samsung, Neoverse, cosmos, cortex, arm, Ares, A76, 7nm, 7+nm, 5nm
Subject: General Tech | October 2, 2018 - 01:24 PM | Jeremy Hellstrom
Tagged: zen 3, rumours, amd, 7nm
There is something interesting going on at AMD according to the rumours coming out of The Inquirer today. It seems that they will be moving from their current 12nm process node directly to a 7nm node for the next generation of Zen processors, both consumer and enterprise. AMD has confirmed the next EPYCs will be on that node, but the news that Ryzen 3 will be was previously unknown. The chip itself is expected to have a base clock of 4GHz with 4.5.GHz top boost, eight cores with 16 threads and performance close to Coffee Lake S. The performance comparison is not all that useful without more information about what scenarios this would refer to, perhaps single thread performance which would be a nice jump.
As always, grab your salt shaker before heading over to follow the links back to the source as we will not know much more until next year.
"THINGS ARE ALL ZEN at AMD but probably not in the way you'd think, as leaked information has spilled the beans on what shape the chip maker's second-gen Zen architecture could look like."
Here is some more Tech News from around the web:
- Intel Core i9-9900K pops up on Amazon US for $582.50 @ The Inquirer
- Apple hardware-centric policy may no longer be viable @ DigiTimes
- HP's Spectre Folio is a 2-in-1 laptop coated in, er, leather @ The Inquirer
- Free ARM cores for Xilinx FPGAs @ Hackaday
- Hackers Are Selling Facebook Credentials on the Dark Web For $3 @ Slashdot
- Microsoft donates 30-year-old obsolete MS-DOS to GitHub @ The Inquirer
- Bitcoin’s Double Spending Flaw Was Hush-Hush During Rollout @ Hackaday
- Some iPhone XS, XS Max Devices Are Experiencing Charging Issues @ Slashdot
Subject: Mobile | September 12, 2018 - 04:24 PM | Sebastian Peak
Tagged: SoC, smartphone, mobile, iPhone XS Max, iPhone XS, iPhone XR, iphone, ios, apple, A12 Bionic, 7nm
Apple’s event today included expected (and previously leaked) iPhone announcements for the faster “S” variant of the iPhone X, as well as a new, larger iPhone XS Max, and finally the new, lower-cost iPhone XR. All three phones include Apple’s latest mobile processor, the A12 Bionic, as well as new cameras and other improvements.
The design is unchanged, but the 6.5-inch form-factor is new (image via Apple)
Beginning with the primary announcement, the new 5.8-inch and 6.5-inch iPhone XS and XS Max phones both feature Super Retina OLED displays which Apple says now offer wider dynamic range, and the glass protecting them is “the most durable glass ever” in a smartphone. The new XS Max offers the same 458 ppi density as the iPhone XS with its 2688x1242 resolution (the iPhone XS has the same 2436x1125 resolution as the iPhone X), and both phones are now IP68 water and dust resistant and dual-SIM capable (using eSIM).
Apple says the A12 Bionic chip will be the first to market at 7nm (Hauwei's 7nm Kirin 980 was previously announced but not shipping until mid-October), and the move to this smaller process should allow for lower power consumption and increased performance.
The A12 Bionic has a 6-core CPU design as we saw with the A11, and uses the same Apple-designed Fusion architecture. Apple says its two performance cores are “up to 15% faster and 40% lower power”, and the four efficiency cores offer “up to 50% lower power” with no stated increase in performance. Other than stating that it is a proprietary design little was revealed about the GPU other than it is now a 4-core design, which Apple says is “50% faster” than before.
The camera system on the new phones offers a new “advanced bokeh” feature which allows for f-stop adjustment after the photo has been taken, and during the presentation this feature appears to work in a very realistic way comparable to dedicated lenses with a DSLR. Other features include improved speakers, stereo audio recording with video, and "Gigabit-class" LTE.
The iPhone XR is an LCD variant with lower cost (image via Apple)
The “one more thing” at the even was a new lower-cost iPhone based on the iPhone X design, but with an LCD display that Apple is calling “Liquid Retina”. This 6.1-inch device has a display resolution of 1792x828 (326 ppi), uses the new A12 chip, and while it is a single-camera phone like the iPhone 8 it uses the latest wide-angle camera from its “S” model siblings.
The display also features “120 Hz touch-sensing” - which may be independent of display refresh, but that is unknown at this point - a wide color gamut, and is a True Tone display like the iPhone X. The phone drops 3D Touch, using instead what appears to be a long-press detection with haptic feedback. The phone does not offer the "Gigabit-class LTE" of the XS/XS Max, is IP67 rather than IP68 water and dust resistant, but does retain the new “most durable glass” from the "S" models.
Pricing for the new lineup is as follows:
- iPhone XS 64GB - $999
- iPhone XS 256GB - $1149
- iPhone XS 512GB - $1349
- iPhone XS Max 64GB - $1099
- iPhone XS Max 256GB - $1249
- iPhone XS Max 512GB - $1449
- iPhone XR 64GB - $749
- iPhone XR 128GB - $799
- iPhone XR 256GB - $899
The new iPhones XS and XS Max will be available next week, with a September 21 launch day (pre-ordering begins on Friday, September 14). The iPhone XR launches on October 26 (pre-order October 19).
Subject: Editorial | July 29, 2018 - 07:49 PM | Josh Walrath
Tagged: TSMC, Skylake, ryzen, Results, Q2, Intel, amd, 7nm, 2018, 10nm
The day after AMD announced their quarterly results, Intel followed up with a very impressive quarter of their own. Intel has reported another record quarter with $17B in revenue and $5B net. The business is extremely healthy and they continue to provide a lot of value and returns to shareholders. Typically Q2 is the second slowest quarter of the year, but Intel was able to improve their revenues by $900M over Q1. In certain quarters a 5% increase may not be all that large, but it is a significant jump from Q1 to Q2.
Intel reported that nearly all areas of the company have grown. Client Computing Group showed a 6% increase year over year, which is good news for the industry in general as many have (often) predicted that the PC market is in decline. This is also in the face of renewed competition from AMD and their Zen architecture based products. AMD also has grown steadily over the past year in terms of shipping products, so that further reinforces the impression that the PC market continues to grow steadily.
The data-centric business is steadily closing the gap between it and the PC centric group. CCG posted $8.7B in revenues while the data groups combined came in at around $8.1B. The Data Center Group was $5.5B of that result. It is up a very impressive 27% yoy. Intel has what seems to be a juggernaut in the data center with their Xeon products, and that growth is quite likely to continue growing as the need for data processing in our information rich world seemingly knows no bounds.
Intel raised their outlook for the year by nearly $2B to an impressive $69B in revenues. This is easily 10x that of their primary competitor. 2018 has certainly been a very profitable year for Intel and it looks to continue that trend throughout the last two quarters. Intel continues to improve upon their 14nm processes and it has allowed them to achieve a 61.4% margin. Compare this to AMD’s 37% margin and we can understand why 2018 is looking so good. Intel has lost a little bit on margin as compared to last year, but the amount of products being shipped is simply stunning as compared to its rival.
There were some expecting AMD to be taking up more of Intel’s marketshare, but that has not been the case. If anything, while AMD’s bottom line has improved, Intel appears to have actually taken more share in an expanding market. Unlike 2003 when AMD had the superior product with the Athlon 64 over Intel’s Pentium 4, the current Ryzen CPUs are “merely” competitive. While the performance and efficiency jump for AMD’s architecture is impressive considering the previous “Bulldozer” based generation, they now offer comparable performance with a price/core count advantage over Intel. This has not been enough to convince people and organizations to change en masse to AMD’s offerings. In 2003 a 2 GHz Athlon 64 was outperforming a 3.2 GHz Pentium 4. AMD was able to continue outperforming Intel even though they were at a serious process disadvantage.
While Q3 and Q4 look to continue Intel’s string of record quarters, things do not look as rosy when we get into 2019. Intel has had an endless stream of problems getting their advanced 10nm process up and running. It was originally expected to replace Intel’s 14nm process around two years after that particular process had been introduced. Then it turned into three years. Now we are five years into Intel using a 14nm variant for their latest generation of products. Intel used to have a 18 to 24 month lead over the competition when it comes to process technology, but now that advantage has all but evaporated. In theory Intel’s 10nm process is superior to what TSMC is offering with its 7nm in terms of die size, power, and transistor performance. However, those advantages do not amount to anything if it is unworkable. Intel has been very tight lipped with analysts and shareholders about the exact issues it is facing with the direction they set on with 10nm. It seems the combination of materials, tolerances, and self-aligned quad patterning is problematic enough that Intel cannot get consistent results with yields and bins.
In the conference call Intel said that 10nm parts will be available on shelves by the holiday season of 2019. This means that Intel expects to hit high volume manufacturing near the end of 1H 2019. Intel further stated that data center parts will be shipping shortly after desktop and mobile, so most expect the first products to hit in Q1 2020. The problem that Intel will is that TSMC will be starting volume manufacturing of their 7nm parts shortly, if not already. AMD has 7nm EPYC sampling to partners and has spoken of a 1H introduction of those parts in volume. AMD will be introducing the Zen 2 architecture in that time on both server and desktop, and they are hinting at a significant IPC uplift with these parts.
If Intel is able to hit its 10nm goal in late 2019, AMD will have around a nine month window where they theoretically could have a superior product than Intel. AMD will surely come ahead from a density standpoint. If we combine this with the potential IPC improvement and a small uplift in transistor performance, then Zen 2 products should be able to outclass anything Intel comes out with. If AMD is really on the ball, then their EPYC processors could have a year to themselves without a comparable product from Intel.
This type of competition does not mean that Intel will simply shrivel up and die, but it is causing investors to rethink holding onto the stock after the pretty impressive run up over the past several years. Intel still has more fab space available to it than AMD could dream of at this point. There will be a lot of competition for 7nm wafer starts that will be shared by AMD, NVIDIA, Qualcomm, and Apple (not to mention dozens of other fab-less semi firms). AMD could very well sell as many chips as it can make, but it simply cannot address the needs of all of the markets that it is competing in. If GLOBALFOUNDRIES 7nm process is similar to TSMC’s, then we will see AMD be able to supply far greater amounts of product to the market, but GF is at least six months behind TSMC when it comes to ramping up their next generation process line. I would not expect GF based CPUs to hit anytime before Q2 2019, if not towards the end of that quarter.
Does this mean that Intel expects nothing except doom and gloom throughout 2019 and possibly into 2020? I do not think so. Intel will retain its market dominance, but it looks to be experiencing a situation that is a combination of a competitor hitting its stride as well as some bad luck/poor planning with manufacturing. This should open the door for AMD to make significant advances in marketshare and allow the company to make some serious money by improving their ASPs as well as shipping more parts.
2018 will undoubtedly be a record year for Intel. It is 2019 that is giving pause to investors and shareholders. If Intel can clean up its 10nm process in a timely manner they will close the door on any advances from AMD. If the company continues to experience issues with 10nm and never in fact gets it out the door, then it will be a long couple of years til Intel gets out their 7nm process. The rumor is that engineers have been pulled off of 7nm to fix 10nm. If this is the case, then I hesitate to even think when we will be seeing that upcoming node coming to fruition.