System news

Computex 2021 revealed much for the PC market this year. Mark Williams covers some of the most exciting highlights.

Computex is always a big item on the calendar each year for new PC tech announcements and products and the 2021 show did not disappoint. Held virtually for the second year running due to COVID, the big three in Intel, Nvidia and AMD had plenty to reveal and show off, leaving us excited for the year to come.

Intel

Keen to keep the momentum going with its 10nm manufacturing process ramp-up, Intel unveiled the first of its new Tiger Lake-U refresh processors, bound for 15W-28W thin and light laptops in the coming months. The two processors revealed were the i7-1195G7 and i5-1155G7.

The first will sit at the top of the 11th Gen U-class processor stack, still with four cores and eight threads, but will be the first this generation to boost up to 5GHz. The new i5 also improves peak boost clocks, offering up to 4.5GHz speeds.

Also revealed was the newest NUC. Called the NUC 11 Extreme, the system takes on a larger toaster-sized form-factor of many SFF cases, allowing this NUC for the first time to offer full length GPU support. Enabling the tight 8-litre form factor is a return of Intel’s Compute Element platform where the motherboard, CPU, RAM, cooling and other core components are all housed in a compact PCIe card, not dissimilar looking to a short and stocky graphics card. While exact specifications weren’t revealed it’s known that it’ll house Tiger Lake-H processors from i3 through i9s and support SODIMM 3200MHz memory.

Nvidia

In a move surprising no one, Nvidia revealed the new RTX 3080 Ti and RTX 3070 Ti. Apart from now supporting Nvidia’s light hash rate (LHR) cryptomining limiters, these are largely like previous offerings.

The RTX 3080 Ti is more like the RTX 3090 than the RTX 3080

AMD’s 3D V-Cache technology stacks relatively huge cache directly on the die, allowing for speed increases of up to 25 percent. (non-Ti) with just 2.5% less shaders and slightly lower clocked core and VRAM, meaning it’s basically a RTX 3090 but at a much more palatable price. And with 12GB of VRAM it is much more suited to 4K gaming than the 10GB of the RTX 3080 ever was.

The RTX 3070 Ti gets a fully enabled GA104 GPU die giving it four percent more shaders than the non-Ti version and most importantly gets VRAM upgraded to GDDR6X, albeit at the same 8GB capacity. All of which means an average performance uplift of 7 percent at 4K resolutions. Both are available now with the RTX 3070 Ti selling from $1,700 and the RTX 3080 Ti from $2,900.

AMD

Announcing what many have been begging for, AMD unveiled the new Ryzen 5 5600G and Ryzen 7 5700G APUs for the DIY market. Releasing on August 5th these APU’s will be great for those not needing discreet graphics performance. With similar Vega 8 IGPs to that of the Ryzen 4000G series, the main talking point is the upgrade to the Zen 3 CPU architecture giving the two close to 5600X and 5800X performance respectively.

Mobile GPUs also made a splash with the announcement of the RX 6000M series. With RTX 3080 mobile performance levels, the RDNA2 touting RX 6800M will sport up to 12GB of VRAM and has already shown in demo units to stand up well to Nvidia’s best. RX 6700M and RX 6600M GPUs were also announced and will cater to the mid and entry level gaming segments.

AMD demonstrated its upcoming FidelityFX Super Resolution (FSR) technology, being AMD’s DLSS competitor. Showing up to two-fold performance increases and beyond, it’s looking like a very exciting technology that was hilariously shown to also work on Nvidia GPUs that don’t support DLSS, like the GTX 1060. What the image quality will be like is another story, but for now this is exciting news.

Saving the best until last, an engineering sample processor was shown with a new stacked 3D V-Cache technology. Essentially AMD is using TSMC’s SoIC Chip-on-Wafer technology, enabling AMD to place a 64MB SRAM chip directly on top of the cache portion of a Ryzen 5000 CCD, effectively tripling the cache size of the CCD to 96MB. For dual CCD parts like the 5900X that means 192MB of L3 cache!

Demonstrating the benefits of this, AMD showed a 5900X vs 5900X with 3d V-Cache technology with game benchmarks showing between four and 25 percent higher performance depending on the title. Quite amazing.

Products with 3D V-Cache were announced to be coming by year end, likely in the form of a Ryzen 5000XT line-up.