With the recent release of Intel’s Haswell-EP (aka Grantley) range of CPUs we have a new assortment of motherboards to match. One of the Supermicro server offerings is the X10SRH-CLN4F single-socket (UP, as opposed to dual socket – DP) board. Continue reading “Supermicro X10SRH-CLN4F Socket 2011-3 Server Motherboard Review”
farmers only men Intel’s network cards are popular due to their speed and reliability, which is often greater than the onboard chips in devices/motherboards. The Intel Gigabit CT Desktop Network Adapter is a PCI-E x1 add-in card with a single gigabit port, usually selling for around $30AU. It is low-profile and should come with a low-profile bracket – handy for thin HTPCs or servers – and is passively cooled as you would expect. It auto-negotiates – so you don’t need to worry about crossover cables – and is PCI-E v.1.1, which supplies more than enough bandwidth for a single gigabit port and should work fine in V2 and V3 slots. It is also supposed to be compatible with x1, x4, x8 and x16 slots. The network controller is Intel’s 82574L – a design released in 2008, with an expected discontinuance of 2018 – Intel certainly expect to get a lot of mileage out of that chip! The 82574L has a TDP of below a single watt, so this is going to be quite a power-efficient add-in card. Intel state that the typical power consumption is in the range of 1.9W for the entire card. Driver support is excellent across virtually all operating systems – it’s plug and play with many Linux distros and works perfectly well with the provided drivers in Windows machines. It also has support for teaming/bonding/link aggregation and 9K jumbo frames. Physically the card is 11.92cm long and 5.53cm wide. In our tests the card managed an impressive average of approx. 950Mbit/s – very close to the theoretical maximum throughput of a gigabit line. If you are in the market for a reliable, fast PCI-Express network card and only need a single port this card is well worth a look – between the features, low power usage, low profile option and driver support it’s an excellent buy for the price.
As promised, here are some benchmarks of Intel’s SSD: An average of 390MB/s for sequential reads? Nice. IOPS: HD Tune Pro’s File Transfer test: Hovering around 500MB/s there – impressive. Last benchmark for now: Overall this is a well performing drive and is the only Sandforce-controller-based drive that we currently use.
The above two CPUs are often shortlisted for entry/midrange gaming builds that aren’t overclocked; so how do they compare? Both are quad-core CPUs without hyperthreading, so they offer 4 cores and 4 threads. The i5-3470 is clocked at 3.2GHz while the i5-3570 is 200MHz faster at 3.4GHz, and both CPUs feature Intel’s Turbo Boost v.2.0, reaching peak speeds of 3.6 and 3.8GHz respectively. This equates to ~6-7% performance difference if you’re running both CPUs at max. Both feature Intel’s HD2500 graphics – according to Intel’s ARK the i5-3570’s graphics are clocked fractionally faster – 1.15GHz as opposed to 1.10GHz, which is a negligible difference. The i5-3570 is about 10% more expensive than the i5-3470 – 10% more cost for 6-7% more CPU performance isn’t bad, though you can’t really go too far wrong either way – if you’re $20 off a better graphics card you’re probably better off going that way and getting the i5-3470, whereas if you could do with a touch more CPU grunt it’s likely worth your $20 to grab the i5-3570.
We were asked to spec a customer build the other day who was torn between the above two processors. Here’s our thoughts on them. E3-1245V2: Quad core, 8-thread, 3.4GHz -> 3.8GHz Turbo E3-1275V2: Quad core, 8-thread, 3.5GHz -> 3.9GHz Turbo All other specifications are equal apart from the 100MHz clockspeed difference. At this level 100MHz is a mere ~3% increase – not something that is going to be visible for most real-world applications. However, if there’s not much price difference it may be worth that extra bit of cash if you really do need every ounce of performance you can get (or just want the bragging rights). So what do they both cost (AU, our prices)? E3-1245V2: $299 E3-1275V2: $389 $90 difference – or around a 30% premium above the E3-1245V2 for a 3% clockspeed increase. If you have to have the utmost performance from a S1155 server chip with onboard graphics, there’s no other option. For most workstation users, however, the 3% is probably not going to be noticed – whereas the $90 could go towards a 128GB SSD or something similar where you’ll get a tangible speed boost. We would recommend taking a long hard look at the price difference – if you need it it’s $90 well spent but we’re pretty sure that most people will go with the E3-1245V2 at the end of the day and spend their $90 elsewhere.
There’s not a great deal to say about these cards apart from that they allow you some crazy network speeds, if you have the disk speed to keep up. They can certainly alleviate network bottlenecks if gigabit is holding you back! This particular card has a fan to keep the chipset cool; it’s not going to be heard in a server room but if your workstation is quiet the high-pitched whine is probably going to be audible. The card does get reasonably hot, particularly if you’re making good use of it’s capabilities – make sure you have enough airflow in the chassis to keep the card cool. Keep in mind it’s assumed that these cards will be used in an environment where there’s at least 200 linear feet per minute of airflow passing over them. 10-gigabit cards are coming down in price quite significantly, though switches are still out of reach of most enthusiasts/small businesses. Watch this space, however, as 10GBe connections are making their way into high-end server boards more regularly and that will slowly filter down to the consumer level. SSD arrays becoming more commonplace will only help with that, as will the new 12gb/s cards from LSI – it’s hard to make use of all that bandwidth if you’re piping it over gigabit!
Here’s a few shots of one of our recent storage server builds: Continue reading “Customer Build: Luna II”