Tyan Introduces 1U POWER8-Based Server for HPC, In-Memory Applicationsby Anton Shilov on April 7, 2016 8:59 AM EST
- Posted in
- IT Computing
At this week's OpenPOWER Summit in San Jose, California, Tyan has introduced its new IBM POWER8-based 1U servers designed for high-performance computing (HPC) as well as in-memory applications. The new machines are designed to target niche markets that do not require extreme processing performance, but need a lot of RAM. The new Tyan GT75 expands the range of Tyan’s POWER8-based machines and helps IBM’s POWER8 platform to compete against Intel Xeon in niche markets.
Nowadays over 95% of servers are based on Intel’s Xeon microprocessors, which does not make producers of servers and operators of large datacenters happy because they have one supplier of key components that controls platform development and prices. Three years ago, IBM teamed up with Google, Mellanox, NVIDIA and Tyan to establish the OpenPOWER Foundation to develop servers based on IBM’s processors collaboratively. So far, IBM POWER-based servers have not gained a lot of market share, but it looks like they are starting to gain traction as companies like Tyan are expanding their POWER8 hardware offerings.
The Tyan GT75 machines (just like the Tyan TN71-BP012 servers launched a year ago) are based on one IBM POWER8 Turismo SCM processor with eight or ten cores and can be equipped with up to 1 TB of memory. The systems feature 32 R-DDR3L DIMM slots using four IBM Centaur memory buffer chips (MBCs) and support 32 GB modules, which are considerably more affordable than 64 GB or 128 GB memory sticks.
|Comparison of IBM POWER8 Turismo CPUs Compatible with
Tyan's SC75 and TN71 Servers
|POWER8 8-Core||POWER8 10-Core||POWER8 12-Core|
|Frequency (nominal)||2.328 GHz
|L2 Cache||512 KB per core
2 MB per CPU
|512 KB per core
5 MB per CPU
|512 KB per core
6 MB per CPU
|L3 Cache||8 MB eDRAM per core
64 MB per CPU
|8 MB eDRAM per core
80 MB per CPU
|8 MB eDRAM per core
96 MB per CPU
|DRAM Interface||DDR3L-1600, memory connects to CPUs using the IBM Centaur MBCs|
|PCI Express||3 × PCIe controllers, 32 lanes|
While the GT75 servers (BSP012G75V4H) feature only one CPU, each IBM POWER8 core can process up to eight hardware threads simultaneously thanks to 16 execution pipelines. The chips also have massive caches (512 KB L2 per core, 8 MB eDRAM L3 per core as well as 16 MB per MBC), which are particularly useful for memory-intensive applications.
|Comparison of Tyan GT75 Servers|
130 W/169 W TDP
130 W/169 W TDP
130 W/169 W TDP
|Installed RAM||8 × 16 GB R-DDR3L||16 × 16 GB R-DDR3L||32 × 16 GB R-DDR3L|
|RAM (subsystem)||Up to 1 TB of DDR3L-1333 DRAM, 32 RDIMM modules, four IBM Centaur MBCs|
|Storage||2 × 512 GB SSDs||2 × 1 TB SSDs||4 × 1 TB SSDs|
|Tyan Storage Mezzanine||MP012-9235-4I
(4-port SATA 6Gb/s IOC w/o RAID stack)
|LAN||4 × GbE ports||4 × 10 GbE ports||4 × 10 GbE ports|
|Tyan LAN Mezzanine||MP012-5719-4C
Broadcom 1GbE LAN Mezz Card
Qlogic+Broadcom 10GbE LAN Mezz Card-
|MP012-Q840-4F Qlogic 10GbE LAN Mezz Card|
The Tyan GT75-BP012, which is a 1U server that is designed for in-memory computing, HPC and virtualization, can be equipped with up to four 3.5” hot-swappable SAS 12G/SATA 6G HDD/SSDs, four network controllers (1 GbE and 10 GbE) as well as a 750 W PSU. The server uses the PLX PEX8748 PCIe switch and one PCIe 3.0 x8 expansion slot. Tyan did not reveal the price of the system, but said that it would be available already in April.
The new 1U machines complement the Tyan TN71-BP012 launched in March, 2015. The TN71 is a 1-way 2U platform for data analytics and applications, it supports 12 3.5” hot-swappable SAS 12G/SATA 6G HDD/SSDs, four 10 GbE network ports as well as two 1200 W PSUs for redundancy reason. This machine supports IBM POWER8 Turismo SCM processor with up to 12 cores and is generally more powerful than the GT75.
Post Your CommentPlease log in or sign up to comment.
View All Comments
Drumsticks - Thursday, April 7, 2016 - linkThat's two new servers in as many days during a pretty high profile time for it. They won't be getting 50% marketshare anytime soon, but Power8 seems to be a good design. NVLink is a nice advantage too; I hope we see a bigger IBM presence in the future.
Brutalizer - Thursday, April 7, 2016 - linkPOWER8 is not a good design, but you have been tricked by IBM marketing to believe so. For instance, POWER8 in general is slower than the fastest x86 cpu. For instance, in SPEC2006, the E5-2699v3 reaches 715 whereas POWER8 reaches 642. Of course, the SPARC M7 is the record holder with 1200 SPECint2006. Here are all the numbers:
In general the SPARC M7 is 2-3x faster than POWER8 or x86, all the way up to 11x faster. In the link above are 25ish different benchmarks such as databases, SPECjbb2005, Hadoop Terasort, Neural networks, etc etc. Just read the benchmarks.
I dont really understand why anybody would want to sell POWER8 servers, as they are slower than x86 and more expensive?
-POWER6 was several times faster than x86 and costed 10x more
-POWER7 was 10-20% faster than x86 and costed 3x more
-POWER8 is slower than x86 and still more expensive?
This is bad news for POWER, as IBM only does high margin business and walks away from low margin business. And POWER9 must be cheaper than x86 because it will be slower. Nobody will buy POWER anymore.
Oh BTW, IBM has officially said that AIX will be killed off. AIX runs on POWER, so without AIX, why sell slow POWER servers?
"...Asked whether IBM's eventual goal is to replace AIX with Linux, Mills responded, "It's fairly obvious we're fine with that idea...It's the logical successor."
Better learn Linux boys.
name99 - Thursday, April 7, 2016 - linkYou're comparing Intel's just released to IBM's two years old. And you're assuming that the performance characteristics of IBM's target audience (on the one hand, very large data HPC, on the other hand very large I-footprint business code) match SPEC2006. These are both dubious assumptions.
POWER8+ will probably be released soon, with enough tweaks (frequency and cache, probably) to match x86; while POWER9 will have 24 cores.
Meanwhile Google continues to look into POWER as a possible CPU for some of its servers (perhaps in the context of driving nV GPUs for AI neural networks). Intel kinda sorta has a vague plan to counter things like nVLink through either using Xeon Phi or one day shipping a Xeon with an on-board FPGA, but it's not clear that either of these solutions is absolutely superior to IBM's solutions.
Phil_Oracle - Friday, April 8, 2016 - linkHow is Power8+ going to get frequency & cache improvements when the process isn't changing (its still at 22nm)? While Power8 was "announced" in August 2013, real 12-core Power8 chips delivered in systems didn’t get announced till April 2015, almost 2 years later. If anything, I believe IBM will just trickle down the higher GHZ Power8's @ the higher GHzes down to the lower end systems, where currently the weaker Power8's are today. Maybe even installing 12-core Power8 onto DCM's to double-up the core count per system.
Brutalizer - Friday, April 8, 2016 - link"...POWER8+ will probably be released soon, with enough tweaks (frequency and cache, probably) to match x86; while POWER9 will have 24 cores...."
Well, if POWER8+ can match x86 it is good for IBM. But I doubt that. Intel is not resting, and when POWER8+ arrives, there will be new Xeons, even faster. And cheaper. So, why would anyone buy POWER8+? Oh, if you want to loose money big time.
name99 - Friday, April 8, 2016 - linkPOWER8+ (apparently now renamed "POWER8 with NVLink") will arrive this year. This year's Xeon's have already arrived. So we will probably have parity of the newest Xeons (the largest E7 v4's) with the POWER8's of the next few months.
Next year POWER9 will be on 14nm FF (compare with 22nm today). So huge process improvement, obviously being used to double density, and likely either increase per-core resources or bump up frequency slightly.
Meanwhile next year's Skylake Xeons will be on 14nm FF (just like this year's Broadwell Xeons) so no process improvement, and just a few percentage points micro-architecture performance improvement. Which means that for 2017 POWER9 is likely to be way ahead of Skylake Xeon.
At some point, of course, Intel will eventually move to 10 nm. But the Kaby Lake delays show that their process improvement has slowed down. They might have two years or so of 10nm before GF get there --- but they will likely use those two years shipping first mobile then desktop, so the Xeon's at 10nm will arrive maybe only a few months before 2020's POWER10.
A second interesting aspect, as yet not clear, is whether POWER will have the chance to explode like ARM has done by having Chinese [or even Japanese] third parties design their own SoCs around a basic POWER core. This seems to be what IBM wants to happen, but I don't think the actual plan specifics are public knowledge. Obviously there are a different set of diificulties in translating the design for a core (or whatever IBM sells) into a full SoC from doing the same with an ARM core targeting mobile. But I could imagine that the Chinese government would be willing to subsidize at least one company to do the necessary research. And it's even possible that Fujitsu might decide this is a more useful path going forward than continuing with SPARC.
Of course these are simply theories, but I don't think they're utterly crazy. If IBM could reinvent (part of) itself as something like an ARM for HPC, that might be good enough to keep POWER viable for quite a few more years.
Vatharian - Tuesday, April 12, 2016 - linkIs NVLink is going to be exclusive to IBM arch?
milli - Thursday, April 7, 2016 - linkYou conveniently forget to mention that you're comparing a 32 core (M7) and 18 core (Intel) CPU to a 10 core Power8 CPU.
Also, the Power8 CPU is very close to the Intel chip in SPECfp.
Its per core performance is the highest of the three under SPEC_rate.
Phil_Oracle - Friday, April 8, 2016 - linkAnd for Database workloads like OLTP, running the open source HammerDB benchmark test, based on TPC-C, shows SPARC M7, with its 32-cores, outperforms 4 x Power8's totaling 24-cores by 37%, therefore showcasing higher performance/core. SPARC M7 also surpassed a dual-Xeon E5-2699 v3 with 36-cores total by a whopping 38%, showcasing a 55% performance/core advantage. https://blogs.oracle.com/BestPerf/entry/20160317_s...
Theres several other real-world benchmarks showcasing SPARC M7 better performance/core. And even Power9 won't be 24-cores till atleast 2018!
milli - Friday, April 8, 2016 - linkPhil_Oracle giving me an Oracle link with benchmarks done by Oracle to put the Oracle SPARC M7 in a nice spotlight.