1. Intel Core X CPU review

Intel’s ultra high-end X99 chipset and LGA2011-3 Broadwell E CPUs have been around quite a while now, but this summer, Intel introduced their replacement, the new Core X platform.

The Core X series CPUs use a new socket: LGA2066, a new chipset: X299 and have more cores, cache and frequency than their predecesors. For instance, the launch line up includes quad-core, 6-core, 8-core and 10-core processors with Intel planning on releasing further models with up to 18-cores later this year.

The two quad-core models, the Core i5 7640X and Core i7 7740X use the Kaby Lake X architecture, which means they are basically just LGA1151 Kaby Lake chips repackaged for the larger LGA2066 socket. The only real difference between Kaby Lake and Kaby Lake X is that Intel has disabled the on-board GPU, which I can’t see as a problem as anybody considering buying such a high-end CPU would be planning on using a discrete graphics card anyway.

As the bigger socket means a bigger heatspreader, the Kaby Lake X CPUs run a little faster than the smaller Kaby Lake CPUs, although you’d be hard pressed to see the difference with the naked eye. For instance, the Kaby Lake X Core i7 7740X was only 1% faster than the Kaby Lake Core i7 7770K using Handbrake to encode H.264 videos. On a final note, as with any Core X processor, the Kaby Lake X CPUs have an unlocked multiplier for easy overclocking, with our sample Core i7 7740X happily running benchmarks at 5GHz, which is around 200MHz more than a typical Core i7 7700K. Even so, it's hard to get too exicited about Kaby Lake X as it's so similar to Kaby Lake.

The table below highlights the key differences between the two new Kaby Lake X CPUs.

The next three chips up use a new architecture, Skylake X. The key differences between the Kaby Lake X and Skylake X CPUs is support for quad-channel memory which provides more bandwidth and additional PCI-E lanes so you can connect more graphics cards and other devices such as RAID controllers. These differences make Skylake X processors Intel’s true new high-end desktop CPUs.

Skylake X has a number of improvements and new features, most notably a complete reworking of the cache configuration. For instance, previous generation Broadwell E CPUs had a small 256KB Level 2 cache per core and a large Level 3 cache which all the cores shared. In Skylake X this design has been turned on its head, with each core having a much larger 1MB Level 2 cache at the expense of a smaller Level 3 shared cache. This change provides a significant performance boost as four times more data can be stored in the Level 2 cache than in Broadwell E. Skylake X also boasts a new mesh interconnect which links all the cores, memory controller and I/O together. Apparently this new link consumes less power and has lower latency than the old ring interconnect, and should also make it easier to produce CPUs with even more cores in the future.

Other improvements in Skylake X include support for AVX-512 instructions. Right now, you’d be hard pressed to find these instructions in any software outside a datacentre, but adding them to consumer CPUs such as Core X should encourage programmers to start making more use of them. AVX-512 is a new set of SIMD instructions, such as scatter-gather support, dedicated state and mark registers.

The initial line-up of Skylake X processors includes three chips, the Core i7 7800X, Core i7 7820X and Intel’s first Core i9 CPU, the 7900X, the table below highlights the key differences between the models.

With so many cores and so much cache the new Skylake X CPUs muscle through multi-threaded workloads incredibly fast. For instance, the flagship 10-core Core i9 7900X finished the Cinebench R15 rendering test with a record score of 2180. This is a significant 18% faster than this chips predecessor, the 10-core Core i7 6950X, and even better when you consider that the old chip was £350 more expensive.

With support for multiple graphics cards and SSDs the three Skylake X Core X CPUs are a great choice for a content creation workstation such as CAD, CAM, 3D modelling, rendering and animation and video editing. Plus, thanks to their higher clock speed and new cache architecture these CPUs are a much better option for gamers than their Broadwell E predecesors.

You can check out the full range of Core X processors, X299 motherboards and 3XS Systems using the new tech on our website.

2. Intel Xeon review

Not content with launching one new CPU architecture, Intel also has launched a new range of Xeon processors for servers. The new Xeons are a big change for Intel as for the first time rather than having multiple sub-brands such as E5 for dual-socket servers and E7 for quad-socket servers, all the new CPUs are branded Xeon.

Under the bonnet however there are substantial differences between the varous Xeon processors. For instance, there are still models that only support dual-CPU and some that scale up to four or eight sockets. These are differentiated by colour, scaling from Bronze, through Silver, Gold up to Platinum and a four digit number as you add in features such as socket support, cores and frequency. For instance, the most basic Xeon is a Bronze 3104 while the flagship model is a Platinum 8180.

Which ever model you choose all Xeon processors are based on the new Skylake SP architecture, where the SP stands for scalable performance. While Skylake SP shares the same 14nm manufacturer process as existing Broadwell EP and EX CPUs, there are a number of significant improvements in terms of raw performance and I/O.

For instance, Skylake SP adds support for AVX-512 instructions, which are an extension SIMD, which previously was only supported by Xeon Phi processors. AVX-512 should really help complex workloads such as HPC and data analytics.

In addition, just like Skylake X, in Skylake SP the various levels of cache have been reorganised, shifting away from the trend of a small Level 2 cache per core and a large shared Level 3 cache in favour of a much larger Level 2 cache and a smaller Level 3 cache. Similarly, Skylake SP also has the new mesh interconnect inside the CPU that consumes less power and has lower latency than the old ring interconnect. Skylake SP goes one step further by replacing the old QPI links between the CPU sockets with a new UPI link which is 45% faster. Skylake SP also marks Intel’s first CPU with more than quad-channel memory with a new six-channel memory controller, boosting memory bandwidth by over 50%. Finally, the PCI-E controller has been given an upgrade, going from 80 lanes in Broadwell EP to 96 lanes in Skylake SP.

All this new architecture, especially the additional memory channels and PCI-E lanes necessitated a much bigger CPU socket, the monstrous LGA3647, which is the same as used by socketed Xeon Phi processors. This is accompanied by a raft of new motherboard chipsets, the C62x series, with various features, however all the new chipsets will support at least four 10Gb Ethernet ports, the first time Intel has provided this natively in the chipset, plus a further 20 PCI-E lanes.

The Xeon SP CPUs themselves scale up to 28-cores, a noticable increase on the 22-core limit of the last generation of Xeons. At the time of writing we had only managed to test the Xeon Gold 6154, which runs at 3GHz and has 18 cores and is intended to replace the existing Xeon E5 2698 v4, which runs at 2.2GHz with 20 cores. The new Xeon SP CPUs proved incredibly quick at 3D rendering, with a pair of 6154s completing the Cinebench R15 test with a score of 5805, which is a magnificent 36% faster than the older 2698 v4s.

Skylake SP marks the biggest advance in performance, bandwdith, features and I/O for an Intel server platform in several generations and is hugely significant. The server CPU market is certainly going to be much more interesting in 2017 than it has for several years.

Scan supplies the full range of Intel Xeon SP processors, plus custom-built servers built by 3XS Systems which are trusted and used by companies of all sizes plus government and education.

3. AMD Epyc CPU preview

AMD has been notably absent from the server market for quite some time, with its last server processor, the Opteron 6300-series, launching way back in late 2012. After the successful launch of the new Zen architecture in Ryzen desktop CPUs earlier this year it should come as no surprise that AMD also has a Zen-based server CPU in the works.

Rather than a play on Zen or resurrecting the Opteron brand, the new CPUs have a new name, Epyc. Like Ryzen, they are based on the Zen architecture although there are some very substantial differences. For instance, as you might expect Epyc CPUs support SMP, so you can install two together in one server. Inside the CPU itself, everything is built out of clusters of 8-cores, so you can expect 8, 6, 24 and 32-core models, with each core being able to process two threads in parallel curtesy of SMT technology that works in a similar way to Intel’s HyperThreading.

Intel has upped the ante with its latest Xeon processors, upgrading the memory controller from four to six channels, but Eypc goes one further and has a eight-channel memory controller which on paper gives AMD’s server CPUs a massive advantage in memory bandwdith. AMD has also gone to town with the I/O, giving even the cheapest Epyc CPUs 128 lanes of PCI-E 3.0, noticeably more than the 96 of Xeon. This could potentially lead to a big cost saving for GPU servers, which historically have required two Xeon E5 processors just to get 80 lanes; I’m hoping that at least one motherboard manufacturer is bold enough to design a single socket Epyc board with eight dual-spaced 16x PCI-E slots. Curiously, dual-socket Epyc systems will have the same number of usable PCI-E lanes as single-socket servers, as half of the 128 lines are used for CPU to CPU communication.

Servers using the new AMD Epyc processors are expected to start shipping later this summer, keep an eye out on the Scan 3XS Enterprise website for the latest.

4. The best PC games of E3 2017

The E3 show in Las Vegas every June is where the hottest new games get revealed. This year was no exception, with a barrage of sequels from the big publishers, some really interesting looking indie games plus a slew of new VR titles.

I’ve yet to see the killer must-have VR game that I’ve been waiting decades for, but Bethesda’s VR version of Doom for the HTC Vive looks promising. While it doesn’t look quite as frenetic and intense as the last year’s awesome reboot, Doom VFR still looks very promising. Check out the gruesome trailer below in anticipation of the release in Q4 this year.

Another long running franchise getting a new game this year is Wolfensten, with The New Colossus scheduled for release in October. Brilliantly rebooted by Machine Games in 2014, the next chapter in the pre-eminent Nazi murder sim is set in the USA and should be both dark and ridiculous at the same time. Feast your eyes on the tongue-in-cheek trailer below.

Destiny 2 also got a lot of air time at E3 as Bungie’s game will get released on PC and consoles this autumn. You can get a feel for the upcoming team-based online shooter in the trailer below.

Finally, even though it still doesn’t have a release date, Talewords was at E3 showing Mount & Blade 2: Bannerlord again, this time showing off some epic medieval battles. Unfortunately there weren’t any demos of the strategy and roleplaying aspects of the game that made the original Mount & Blade so compelling, but the battle gameplay does look much improved. Check out the gameplay in the video below.

5. Vote in the PC Pro Awards

Our friends over at PC Pro magazine are asking for customers to vote on the products and services they have recieved in the last months. If you complete the online survey you not only help to share your experience with other shoppers, you’ll be in with a chance to win some great prizes. Please visit the PC Pro awards website for more details and to share your opinion.