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 | October 9, 2018 - 09:46 PM | Tim Verry
Tagged: Trident Z RGB, samsung b-die, G.Skill, double capacity DIMM, ddr4, DC DIMM, 64GB
G.Skill has joined forces with ASUS to release a new series of Trident Z RGB DC DDR4 memory modules aimed at ASUS’ Z390 motherboards and take advantage of “double capacity DIMM” technology that uses taller form factors to allow twice the memory ICs per stick.
The new Trident Z RGB DC memory modules feature 16 Samsung B-die memory ICs for a total capacity of 32 GB when using 8Gb chips. Initially, G.Skill will offer the new double capacity modules in 64GB kits (32GB x 2) clocked at 3000 MHz and 3200 MHz. As part of the company’s Trident Z RGB family, the new DC series continue to support RGB LEDs which can be customized by software including Asus Aura Sync. The 3000 MHz kit comes with 14-14-14-34 timings. There are two 3200 MHz kits (both 64 GB) that come either with 14-14-14-34 or 14-15-15-35 timings. All three kits operate at 1.35V out of the box.
The double capacity DIMMs will work with select ASUS motherboard based around the Intel Z390 chipset including the ROG Z390 Maximus XI APEX, ROG Maximus XI Gene, and ROG STRIX Z390-I Gaming.
The Mini ITX Strix Z390-I Gaming board would benefit the most from the double capacity DIMMs at they will allow enthusiasts to pack more than the 32 GB limit of today’s JEDEC standard UDIMMs into the only two memory slots on the board. Meanwhile, the larger Z390 boards will be able to host even more memory enabling workstation workloads to be run (or a big ass home virtual lab environment heh).
G.Skill has not yet released pricing or availability information for these new memory kits. I am curious whether the double capacity DIMM standard will catch on and if it will be adopted by other motherboard manufacturers or if it will stay an ASUS exclusive feature. At least on paper, it appears the only tradeoff is having to accommodate taller modules when considering which CPU cooler to purchase.
Subject: General Tech | September 29, 2018 - 10:48 AM | Tim Verry
Tagged: Ampere, arm, armv8-a, datacenter, ddr4
Ampere recently announced the availability of its first ARM-based server processor dubbed eMAG. The new chips use 16 or 32 custom CPU cores built upon the X-Gene 3 (once pioneered by Applied Micro) compatible with the 64-bit ARMv8-A instruction set. Ampere, in partnership with Lenovo (and several smaller unspecified ODMs), has started shipping eMAG to its customers and partners. Current eMAG processors are based on TSMC 16nm FinFET+ and Ampere plans to move future eMAG processors to TSMC’s 7nm node while adding support for multi-socket servers as soon as next year.
Ampere’s eMAG processors are designed for the datacenter with big data computing workloads in mind that benefit from large amounts of memory and cores including big data analytics, web serving, and in-memory databases. The new ARM server CPU entrant is designed to compete with the likes of Intel’s Xeon and AMD’s EPYC X86-64 processors as well as other ARM-based offerings from Cavium and Qualcomm. Early reports suggest that eMAG is no slouch in performance, but where it really excels is in price to performance, performance per core per dollar, and total cost of ownership metrics.
Today’s eMAG processors feature either 16 or 32 custom ARM cores clocked at 3.0 GHz base and up to 3.3 GHz turbo with 32KB I-cache, 32KB D-cache (L1) per core, 256KB L2 cache which is shared between two paired cores, and a global shared 32MB L3 cache. There are eight DDR4 memory controllers (up to 1TB DDR4-2667 using 16 DIMMs for up to 170.7 GB/s memory bandwidth) as well as 42 lanes of PCI-E 3.0 I/O. The CPU cores, cache, and controllers are connected using a switch that is part of a coherent fabric. Additional I/O support includes four SATA 3 and two USB 2.0 along with 10GbE. The eMAG processors have a 125W TDP.
Perhaps most interesting is the pricing which Ampere has set at a rather aggressive $550 for the 16-core chip and $850 for the 32-core processor. The Ampere chips are interesting especially following Qualcomm’s seeming loss of interest in this space as it dialed back its Centriq efforts earlier this year. With a new ARM entrant that reduces the datacenter barrier to entry for workloads that need lots of acceptable performance cores paired with lots of memory and AMD’s renewed datacenter push on all fronts, Intel is going to have its work cut out for it when it comes to maintaining its datacenter dominance. At the very least it may shake up server CPU pricing. Further, perhaps beyond its intended use, these ARM-based offerings may also introduce some new server platforms that are accessible to enthusiast virtual lab-ers and small HPC developers (small shops, universities, etc) that can use lower cost systems like these for testing and research into developing highly parallelized code that will eventually be run on higher end servers in the “hyperscale” data center.
I am curious to see if the eMAG will live up to its performance claims and expectations of competing with the big players in this space. According to ExtremeTech, Ampere claims the 32-core eMAG is able to match the Intel Xeon Gold 6130 (16 core / 32 thread, 2.1-3.7 GHz, 22MB L3, and 125W TDP) in SPEC CINT2006 benchmarks. The company further claimed earlier this year that eMAG would offer up to 90% performance per dollar versus Xeon Silver and 40% higher performance per dollar compared to Xeon Gold processors from Intel.
What are your thoughts on eMAG and ARM in the server space?
- ARM Reveals First Public CPU Roadmap - Targeting Intel Performance
- ARMing the Cloud; Qualcomm's Centriq 2400 Platform will power Microsoft Azure instances
Subject: General Tech | June 11, 2018 - 11:25 PM | Tim Verry
Tagged: Trident Z RGB, samsung b-die, overclocking, msi, LN2, liquid nitrogen, Intel, G.Skill, ddr4, computex 2018, computex
G.Skill held its annual extreme overclocking competitions (the OC World Cup Competition and OC World Record Stage) at Computex 2018 in Taipei where the overclockers managed to break 13 world records including the two highest DDR4 clockspeeds and the fastest Core i7-8700K clockspeed.
Overclocking teams from around the world using Intel processors, G.Skill DDR4 memory, and motherboards from MSI, EVGA, and ASRock along with extreme cooling methods (de-lidding and loads of LN2) were used to set the world records in 3DMark Fire Strike, SuperPi, Maxxmem, Geekbench 4, GPUPi for CPU, WPrime, and PiFast benchmarks along with hardware records of DDR4 5543 MHz and an Intel Core i7-8700K at 7409.03 MHz.
On the memory front, G.Skill notes that Toppc is now the world record holder with the DDR4-5543 MHz overclock achieved using an Intel i7-8700K, MSI Z370I Gaming Pro Carbon AC, and G.Skill Trident Z RGB memory. Following Toppc’s overclock Kovan Yang managed to achieve the second highest DDR4 clockspeed record at DDR4-5541 MHz on the MSI X299 Gaming Pro Carbon AC motherboard and Intel Core i7-7740X processor which is an interesting feat on the HEDT platform.
Other notable benchmark world records include a 3DMark Fire Strike Extreme Single score of 20,320 (i9-7980XE and EVGA X299 Dark platform), Geekbench4 Single Core score of 9842 points (i7-8700K on an ASRock Z170M OC Formula), WPRIME -32M score of 1.937 seconds, and a SuperPi 32M score of 4 minutes and 8.922 seconds.
Interestingly, G.Skill’s video coverage (embedded below) shows both manual full pot cooling as well as the automated Roboclocker LN2 cooler being used. The video jumps from scene to scene quickly but it does give you some glimpses at the process and the pots/heatsinks used with the RAM and processor to keep things cool even when cranking up the voltage and clocks!
Subject: Memory | June 6, 2018 - 08:06 PM | Tim Verry
Tagged: gskill, G.Skill, ddr4, Trident Z RGB, RGB LED, computex, computex 2018
G.Skill teased new DDR4 memory modules at Computex this year including the mysterious and stylish Trident Z RGB Royal memory. The new memory, of which G.Skill had a single stick on display, features a metallic heatspreader with a mirror finish as well as a crystalized RGB LED light diffuser that manages to make RGB look awesome.
Unfortunately, other than admiring the aesthetics, I have not been able to find any other information on this new RGB Royal series of memory. Hopefully G.Skill will be more willing to spill the beans after the craziness of Computex is over and the memory gets closer to fruition. One thing that is almost certain is that these DIMMs are not going to be cheap!
Subject: Memory | May 3, 2018 - 04:06 AM | Tim Verry
Tagged: hyperx, gaming, ddr4, ddr4-2933, RGB, RGB LED
Kingston’s enthusiast-focused HyperX brand recently launched a new set of RGB-equipped DDR4 memory modules that use IR transceivers to sync up the LEDs across all the DIMMs. The aptly named Predator DDR4 RGB memory kits feature stylized angular black aluminum heat spreaders and RGB LEDs along the top edge. The DIMMs use eight 1GB chips along a single side.
HyperX’s new Predator DDR4 modules are compatible with Asus Aura Sync, Gigabyte RGB Fusion, and MSI Mystic Light Sync RGB LED control software. The new kits are available in 8GB, 16GB, and 32GB capacities using one, two, or four 8GB modules. HyperX rates (PDF) the modules at DDR4 2933 MHz with CL15-17-17 timings at 1.35V when using the Intel XMP profile. Out of the box, the modules run at 2400 MHz (CL17) and 1.2 volts, however.
The RGB modules reportedly offer smooth lighting effects with low latencies thanks to the direct module-to-module IR communication keeping everything in sync.
The HyperX Predator DDR4 RGB modules come with a lifetime warranty and have a MSRP of $257 for the 16 GB (2x8GB) kit and $513 for the 32 GB (4 x 8GB) kit. Fortunately, the kits are going for a bit less than MSRP online with the 16 GB RGB kit going for $245.99 and the 32GB RGB kit going for $491.99 or about $20 to $30 over the non-RGB Predator DDR4 3000 MHz offerings.
Subject: Memory | April 13, 2018 - 10:46 AM | Tim Verry
Tagged: adata, xpg, ddr4, Samsung, overclocking, 5ghz, coffee lake, Z370
ADATA recently announced that it was able to overclock its upcoming XPG Spectrix D41 RGB DDR4 memory to 5 GHz on air cooling. The new Spectrix modules were first shown off at CES 2018 along with phase change cooled Spectrix D80 DIMMs.
Not content to let G.Skill have all the fun, ADATA took its 2132 MHz AX4U470038G19-DR41 memory and pushed it to 5 GHz in dual channel mode with fairly tight timings of 21-26-26-45-2T. They do not mention how much voltage was needed, but the XMP 2.0 profile of 4608 MHz at 19-19-19-39 and 1.45V suggests that likely at least 1.5V was needed. For comparison, G.Skill was able to hit 5007.4 MHz at CL21-26-26-46-2T while ADATA hit 4996.8 MHz at 21-26-26-45-2T (as reported by CPU-z). Both memory manufacturers used a MSI Z370I Gaming Pro Carbon AC motherboard and Intel Coffee Lake Core i7-8700K to achieve their overclocks. ADATA had the processor clocked at 4.3 GHz (100 BCLK x 43x multiplier).
ADATA’s Spectrix D41 memory uses stylized heat spreaders along with RGB LEDs along the top edges. According to ADATA it is using carefully screened Samsung B-die ICs which so far appear to be the best chips out there for DDR4 when it comes to pushing clocks and AMD compatibility. While a retail kit clocked at 5 GHz (at least when XMP is turned on) out of the box is still far off, the increasing number of successful overclocks is promising for enthusiasts that are looking for kits to overclock on their own. I am still waiting for the memory kit makers to demonstrate the 5GHz on air feat with an AMD platform though as so far the attempts have all used an Intel platform. Perhaps once Ryzen 2000 CPUs and X470 motherboards are out we will see what 5 GHz does for Infinity Fabric.
Tom Chan, director at ADATA Technology, was quoted in the press release as stating:
“For us, the next critical step will be working to make this more than just a technological milestone, but something that will be accessible to gamers, overclockers and others, so that they can ultimately benefit from this amazing performance.”
ADATA / XPG have not yet announced pricing for its Spectrix D41 (or D80) kits but hopefully they will be available soon. The Spectrix D41 should be available in up to 16GB per DIMM capacities and up to 4600 MHz with XMP 2.0 profiles. I am curious whether the D80 with its phase change cooler could be overclocked any more than 5 GHz or if that is simply the limits of Samsung’s current generation ICs regardless of cooling method (outside of exotic cooling like lquid helium or liquid nitrogen and needing ludicrous amounts of voltage of course heh).
Subject: Memory | April 11, 2018 - 08:07 PM | Tim Verry
Tagged: xmp, led, Intel, ddr4, Corsair Dominator Platinum, corsair
Corsair is introducing a new special edition DDR4 memory kit called the Dominator Platinum Special Edition CONTRAST. The new individually numbed kits feature a monochrome white a black design with white LEDs. The modules come in 32GB kits comprised of either four 8GB DIMMs or two 16GB DIMMs.
Corsair’s fancy black and pearlescent white special edition memory features 10-layer PCBs, carefully screened Samsung ICs, and DHX (dual-path heat exchange) cooling. The top of the modules hold white LEDs (no RGB here!) to add a bit of glow to your system.
Out of the box, the kits come clocked at 2133 MHz with CAS latencies of 15-15-15-36 and running at 1.2 volts. There is a XMP 2.0 profile that, when activated in the BIOS, bumps things up to 3466 MHz and 16-18-18-36 timings though the speed increase comes at the cost of more power draw at 1.35V.
The Special Edition Contrast memory would look good in most any build, but especially one that eschews RGB for white lights and a simple color scheme. The refined Dominator Platinum memory comes at a premium price though with the 2x16 GB kit having a MSRP of $439.99 and the 4x8 GB kit hitting $479.99. The four DIMM kit is available now from Corsair and the two DIMM kit is coming soon.
Definitely on the expensive side, but it sure looks nice! What are your thoughts?
Subject: Motherboards | April 5, 2018 - 08:27 PM | Tim Verry
Tagged: x470, mini ITX, ddr4, biostar, amd, AM4
Biostar is planning to release an updated Mini ITX AMD motherboard according to leaked images sourced by Videocardz. According to the image, the new board will be called the Biostar Racing X470GTN and will feature the AMD X470 chipset which is a refreshed enthusiast chipset that is supposed to be more power efficient and contain tweaks and optimizations for AMD’s upcoming “Zen+” Ryzen 2000 series of desktop processors and APUs.
The Racing X470GTN looks very similar to the Racing X370GTN that Biostar released last year down to the same black PCB and board component layout though the VRM heatsink has been spruced up a bit and is now in red and white rather than black and white. Further, the X470 chipset heatsink lacks the carbon fiber aesthetic and the PCI-E slot is white instead of black. Lastly, the PCB audio isolation for the onboard audio may have been slightly tweaked. The LED-equipped Mini ITX motherboard is powered by a 24-pin ATX and a 4-pin EPS CPU power connector that feeds the seven phase (4+3) power phases. Unfortunately, the heatsink on the VRMs does not look any larger which may hamper any heavy overclocking attempts on the processor as Hardware Canucks saw rather high temperatures (though not magic smoke bad) at stock clocks on the X370GTN they reviewed. In any event, the AM4 socket sits up top and is surrounded by two DDR4 memory slots, four SATA 6 Gbps ports, and one PCI-E 3.0 x16 slot. There is also a PCI-E M.2 slot hidden behind the board for enthusiasts. The onboard audio codec is a Realtek ALC892 chip (per Videocardz) and while the leaked image does not confirm anything I am guessing the Gigabit Ethernet NIC is also of Realtek origin and is likely the same RTL8118AS used in the previous generation motherboard.
As far as rear I/O is concerned, there is not much to speak of but Biostar does include a decent amount of high speed USB ports with at least two being USB 3.1 Gen 2 and the remaining four being USB 3.0 or USB 3.1 Gen 1 (the X370GTN had USB 3.0 but the X470 update may have bumped these up to USB 3.1 Gen 1 not that it matters in practical terms of speed). In addition to USB, the Racing X470GTN’s rear panel hosts a combo PS/2 port, DVI and HDMI video outputs, Gigabit Ethernet, and six audio outputs (one optical, five analog).
Naturally, being a leak, there is no word on official pricing or availability on this motherboard, but I would guess it will be priced around $120 following the launch of AMD’s Ryzen 2000 series CPUs and 400 series chipsets.
In other Mini ITX X470 news:
Subject: Memory | March 29, 2018 - 12:58 AM | Tim Verry
Tagged: Trident Z RGB, RGB, overclocking, G.Skill Trident Z, G.Skill, dual channel, ddr4, 5000 mhz
A bit over a month ago G.Skill launched a new Trident Z RGB kit that offered up 4700 MHz speeds in a 16GB kit using Samsung B-dies. Now, G.Skill has managed to push the kit to 5,000 MHz on air and the prototype kit is getting closer to fruition as a retail product.
G.Skill managed to overclock its Trident Z RGB 4700 MHz kit by a bit over 300 MHz to hit 5,007.4 MHz in an air cooled system featuring an MSI Z370I Gaming Pro Carbon AC and an Intel Core i7-8700K. The RGB memory kit achieved 5,007.4 MHz with timings of 21-26-26-46 2T (CL, tRCD, tRP, tRAS, CR) and while they did not mention voltage the kit likely required around 1.5V since the base 4700 MHz kit needs 1.45 volts. The 8700K processor was sitting at the default 100 BCLK with a 43x multiplier for a clockspeed of 4.3 GHz. Perhaps more promising is that the overclocked memory was still able to be used in dual channel mode where previous attempts required extreme cooling methods and/or operating in single channel mode.
Tequila Huang, the Corporate Vice President of G.Skill International, had the following to say in the press release:
“Previously, the 5GHz memory speed is only achievable in extreme overclocking and in single-channel. We’re excited to share that we’ve been able to achieve the 5GHz memory speed in not only air-cooling conditions, but also in dual-channels. This is a major milestone for us. We will make every effort to bring this specification onto the consumer market, and bring the experience of extreme performance to worldwide users.”
G.Skill is not quite ready to bring a 5,000 MHz RGB memory kit to market, but they are getting closer and hopefully by the time they do memory pricing will have settled down a bit! It is impressive how far memory speeds have come in the last few years, and I am curious where we will go from here.