AnandTech Benchmarks Memory for Servers

Subject: General Tech, Memory | January 1, 2014 - 03:37 AM |
Tagged: RDIMM, ram, LRDIMM

In all honesty, outside of on-die graphics solutions, memory speed and latency are often neglected. My only requirements for RAM beyond the recommended specs for my motherboard and processor has been a heat spreader of some sort (and that is just because I have bad luck with several DOAs on unshielded RAM which I assume was handling problems).

But this story is for the enterprise users.

View Full Size

Johan De Gelas of AnandTech decided to test a few different configurations of RAM including both RDIMM and LRDIMM modules. LRDIMMs are significantly more expensive than the cheaper RIMM modules but, especially if you could reduce server count (and active licenses of software running on them) they wanted to investigate whether it could be cheaper overall. This would not be the case if software is completely CPU-limited... but, again, when is memory the limiting factor?

That is where the benchmarks come in. Among the handful of measurements performed, they simulated thousands of users accessing a CDN with between one-to-three-quarters of a terabyte of memory. In both cases, 768GB of LRDIMM memory had significantly higher throughput and significantly lower latency.

As always, check out the review if you are interested.

Source: Anandtech

January 1, 2014 | 04:01 AM - Posted by arbiter

IMO less you are going form like 1333mhz ram to 2133 or 2400mhz ram, there isn't really a boost. If you use built in gpu (god forbid) faster ram helps but could just use money you were gonna dump in to faster ram and put it towards a gpu.

January 1, 2014 | 10:40 AM - Posted by Anonymous (not verified)

With gaming console APUs moving the CPU cores on die with the GPU, and sharing the the GPUs fat bus with GDDR5 memory, and AMD placing GPU accelerators on die of some future server chips(Berlin and Warsaw server APUs for x86, Seattle for ARM based) how much longer will it be before motherboard memory on a narrow and bandwith constrained motherboard will no longer be a factor. Future gaming Console type APU based complete gaming systems on a PCI card, appears to be the way to completely bypass the motherboard bandwidth constraints, and even though latency on GDDR5 is higher, these future PCI based complete systems APUs will probably have large on die RAMs for gaming engines, or core server OS functions(for server variants), to reside in and bypass most of the GDDR5 higher latency for faster bandwith issues. AMD appears to be favoring server APUs with weaker, x86 and ARM, based CPUs offloading more general purpose workloads to on die GPU general purpose compute accelerators, for servers, instead of server chips based just on more powerful CPUs only.

January 1, 2014 | 06:53 PM - Posted by arbiter

Um, only ps4 does that with GDDR5 ram. Xbox1 uses 2133mhz DDR3 ram in quad channel setup. Ps4 does stomp xb1 with 160 GB/s memory bandwidth vs 60 GB/s for xb1.

January 1, 2014 | 08:24 PM - Posted by Anonymous (not verified)

The post describes future APU Consoles on a PCI card based systems that will appear, and their APUs will have on die RAM as well as share wide GPU buses with the on die GPUs, these whole compting platforms will share GDDR5 memory and be all on a PCI card! They will in essence be gaming APUs(combining Xbone style one die ram, and PS4 style GDDR5 memory), and will relegate the motherboard CPU to a support role! These new PCI based complete APU systems on a PCI card will have all the advantages of wider shared BUSES with the on die GPU, the GDDR5 high bandwidth, and a large On Die Memory to host the Gaming engine/gaming OS core functions(to obviate the GDDR5's higher lantency issues with respect to CPU type code execution). This is the future, and Nvidia will also be offering with its Maxwell ARMv8 based gaming CPU/GPU "APU" type systems to compete with AMD's x86 complete APU based gaming systems on PCI card based products! AMD and Nvidia will also take the same approach with their ARMv8 based server options, And in AMD's case x86 as well as ARM based complete server clusters on a PCI card. These server based product's APUs(ARM and x86) will be GPU accelerated, and have all the advantages of the CPU/GPU fat memory bus, high speed GDDR5 memory, and On die RAM, for high preformence home and HPC based computing! HSA and PCI based complete Gaming/computiting "APU" type platforms are the logical next step for the gaming desktop/home server and SMB markets. Future gaming GPUs will no longer be just a GPU, they will become GPU/CPU complete systems, that host specilized OSs tuned to their Gaming workloads all contained on a PCI card formfactor.

January 1, 2014 | 10:25 PM - Posted by Anonymous (not verified)

BigChiefRunAmok is that you?

January 1, 2014 | 11:39 PM - Posted by Anonymous (not verified)

More like a cut and paste, but there is much astroturfing going on with the M$ and Intel fanboys out there, but who can tell these days, and BigChief is just reading the same articles thet everyone else is reading, could you be one of those tech report terfers that are taking swipes at others who do not worship at the Church of WinTel?
APUs, especially the ones on the consoles, could be easily made in PCI form, and are powerfull CPUs really needed for gaming, or is it just a bandwith issue and not necessarly the need for a powerfull CPU that will allow the likes of Nvidia and AMD to produce entire Gaming systems on a PCI card and cut Chipzella out of the equation, Nivida's Maxwell will begin Nivida's movement towards this very end, forced by AMDs sucess in the console gaming market! APUs have this potential being both CPU and GPU enabled, and the Gaming GPUs with their large data buses and fast GDDR5 memory, could truly become complete systems with a few cpu cores and some on die RAM, Nivida sure accelerated their Maxwell roadmap, after looking what AMD did with the gaming console APU, BigChief was not the first to notice the future potential of the gaming APU, with respect to having a Descrete APU PCI based gaming system, that slots into a PCI formfactor, just read the roadmaps for the Big 3, and others with regards to such APU gaming beasts landing on the PCI slots of many a motherboard, running Steam OS distos in a PCI gaming Nirvana complete ecosystem. AMD in the past has started many an innovative firestorm in the computing world, and the gaming APU with HSA and other innovations, is just the beginning, of the fight for the Gaming/server/computing market where the CPU will merge with the GPU, and free itself from the constraints of the classic motherboard narrow bus formfactor. Who would not like a system with motherboard or daughterboard made with loads PCI 3.0 x8 or x16 slots and a gaming cluster of these PCI based APU complete gaming systems! Linux in the HPC market is more than capable of handeling such workloads with its already available software developed to run the supercomputers, and server clusters. Things are about to get really interesting!

January 1, 2014 | 11:59 PM - Posted by Anonymous (not verified)

arbiter is that you?, or whoever you are you appear to be a veteran damage control specialist! trying to tie posts with posters on other websites! there is Much fear and loathing out there in these rapidly changing markets, that the Big Technology Trusts did not do a good job predicting, and are now trying to play catch up, with the old school marketing tactics, that worked in the past, but no longer are working!

January 2, 2014 | 06:46 PM - Posted by Anonymous (not verified)

Looks like they went to the full-size truck grille school of design.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.