UserBenchmark

AMD’s Ryzen 5 3600 is one of five new processors based on the latest Zen 2 7nm microarchitecture. Although the 3600 is at the bottom of the 3000 line-up, it is no less innocuous to the competition. This 6-core, 12-threaded processor is expected to be priced at 199 USD and succeeds the Zen+ Ryzen 5 2600 improving upon it by 20% in terms of effective speed. In fact the 3600 base and boost clocks of 3.6 Ghz and 4.2 GHz respectively, match the previous generation hex-core Ryzen 5 2600X and therefore indicate an 11% clock for clock (IPC) improvement over the previous generation. At this price point, the Ryzen 5 3600 only has to compete with Intel’s 6-core i5-9400. Whilst single and quad core speeds are comparable, AMD continues to excel at multi-core performance, with early benchmarks showing a 63% multi core advantage over the 9400. If that was not enough to tempt Intel loyalists to defect, the Ryzen 3000 processors all include coolers and although they will be launched alongside the new mainstream X570 chipset (which is PCIe 4.0 enabled), they are backwards compatible with 400 and even the 300 series motherboards. The 3600 is likely to be the most popular gaming CPU for the rest of 2019. ^{[Jun '19 CPUPro]}

Early Q3 of 2019 welcomes the Ryzen 9 3900X, AMD’s current top of the range, third generation Ryzen flagship which is setting a new standard for high-end workstation processors. This is AMD’s first Ryzen 9 processor and is a 12-core, 24-thread CPU based on their latest 7nm Zen 2 microarchitecture. A second Ryzen 9 CPU in the form of the 16-core, 32-thread 3950X is expected around early Q4 of 2019. The 3900X has a boost/base clock of 3.8/4.6 GHz, 64 MB of L3 cache (double the L3 on the rest of 3000 series), support for 3200 MHz DDR4 system memory and a TDP of 105W. All this, plus a Wraith Prism cooler, at a relatively accessible launch price of $499 USD. By comparison, Intel’s coolerless 12-core, 24 thread i9-9920X is currently available for $1200 USD which is well over twice as expensive as the 3900X. Early benchmarks illustrate that the extra dollar spend does not translate to extra performance, with the i9-9920X having just a 4% greater peak overclocked benchmark than the 3900X across single, quad and multi-core workloads. Currently in terms of multi-core performance at this price point, there is no threat from Intel. The only real competitor is the upcoming 16-core, 32-thread 3950X ($749 USD) or possibly the slightly more affordable (but significantly worse value for money) 8-core, 16-thread Ryzen 7 3800X ($399 USD) which sits just below in the new Zen 2 line-up. Like the others in the 3000 series, the Ryzen 9 3900X can plug into the new PCIe 4.0 enabled X570 motherboard via an AM4 socket, as well as 400 and 300 series motherboards. ^{[Jul '19 CPUPro]}

The RX 5700 XT is AMD’s latest flagship 1440p gaming graphics card which has been released alongside their 7nm Zen 2 processors. The 5000 series of GPUs feature the first iteration of AMD’s new RDNA 7nm microarchitecture, also known as Navi, which is able to deliver better and more efficient performance than the previous generation (Vega). Specifically the 5700 XT has 40 compute units (compared to 36 in the RX 5700 - a cut down version) and a game frequency (comparable to NVIDIA’s boost frequency) of 1.76 GHz (versus 1.63 GHz in the RX 5700). AMD have limited the frequency to a max boost clock of 1.9 GHz. Like the 5700, the 5700 XT has 8 GB of GDDR6 high-speed memory, capable of delivering 448 GB/s of memory bandwidth and features PCIe 4.0 support. The 5700 XT has slightly more power consumption with a 225 Watt TDP compared to 180 Watts for the 5700. The first benchmarks suggest that these differences in specifications result in the 5700 XT having around a 15% faster effective speed than the 5700 for a 13% price premium. The $399 5700 XT is also around 10% (varies by game) faster than the NVIDIA’s $460 RTX 2070. To counter, NVIDIA have announced refresh versions of their recent RTX models in the form of the RTX 2060 Super, 2070 Super and 2080 Super, which are poised to offer better performance and value than the previous generations. The RTX Super cards are also due for release now (July 2019) so shop prices will determine where the best value for money lies. ^{[Jul '19 GPUPro]}

NVIDIA's GTX 1660 follows hot on the heels of last month's release of the GTX 1660 Ti. As the name would suggest, the 1660 is a slightly scaled back version of the 1660 Ti. Both feature NVIDIAs's new TU116 Turing based die, have 6GB of VRAM, are without RTX cores and have a power draw (TDP) of 120W. The main differences arise from the number of CUDA cores: the 1660 has 1408 whilst the 1660 Ti has 1536, and memory bandwidth: the 1660 can deliver 8 Gpbs using ubiquitous GDDR5 (as featured in the GTX 1060 3GB and 6GB) versus the 1660 Ti which can deliver 12 Gpbs using newer, faster and dearer GDDR6. Early benchmarks show that the GTX 1660 has about a 20% lower effective speed than the 1660 Ti, but with an entry price of $219 USD, the 1660 is also about 20% cheaper. Further, the 1660 has a 12% effective speed advantage over the ~$230 USD 6GB 1060 and a similar real world effective speed to AMD's $265 USD RX 590. NVIDIAs strategy of offering great value Turing products at all price tiers can only be good for competition and consumers. ^{[Mar '19 GPUPro]}

The MX500 is Crucial’s current flagship consumer SATA SSD featuring their latest second generation 64-layer 3D TLC NAND. It’s available in 250 GB, 500 GB, 1 TB and 2 TB capacities in a 2.5-inch form factor. All but the 2 TB version will also be available in M.2 (2280) form in the future. The MX500 features a Silicon Motion SM2258 controller which is a change from the Marvell 88SS1074 controller featured in the MX300 (it’s nearly two year old predecessor). Performance is around 30% better than on the MX300 which currently retails at the same price. The MX500 has an SLC cache which increases with drive capacity. Consequently, the larger capacities are better able to sustain high sequential write speeds. The 250 GB version has 250MB of SLC cache, the 500 GB has 512 MB, the 1 TB has 1 GB and the 2 TB has 2 GB. Extended sustained write performance tests show that even though speed does drop off after the cache has been exhausted, it is still maintained at very respectable levels. The MX500’s five year warranty is in line with Samsung’s 850 Evo and exceeds it in terms of endurance (just 75 TBW for the 250 GB 850 Evo versus 100 TBW for the 250 GB MX500). It doesn’t quite match up to the Samsung’s 850 Evo’s performance (effective speed 8% slower), however at current prices it is about 20% cheaper, and on balance offers better value for money. ^{[Mar '18 SSDrivePro]}

Adata’s XPG SX8200 offers NVMe PCIe SSD performance at SATA SSD prices, thereby offering outstanding value for money to casual and power PC users alike. Adata have combined two high performance commodity components: a Silicon Motion controller (SM2262) and Micron’s second generation 3D TLC 64 layer flash memory. The XPG SX8200 also includes a generous SLC cache and a DRAM cache buffer. Adata have not specified the exact size of the SLC cache in each model but it’s clear from our 60 second sustained write tests that the buffer is sufficient for more than 60 seconds of continuous writes which equates to over 60GB (60s x 1 GB/s) for the 240GB model and over 90GB (60s x 1.5 GB/s) for the 480GB and 960GB models. Unlike many other drives with SLC caching, the SX8200 has a large enough cache to ensure that consumers will almost always operate the drive within the cache and therefore experience no write degradation at all. Adata offer a 5 year warranty and a limited TBW warranty (160 TBW for the 240 GB version, 320 TBW for the 480GB and 640 TBW for the 960GB) on these SSDs, which is in line with other premium manufacturers. The NVMe PCIe SSD consumer market has been dominated by Samsung in recent years but the 240GB SX8200 beats Samsung’s 250GB 970 Evo hands down on both price and performance (the 250GB 970 Evo is let down by a relatively small SLC cache which allows for less than 10 seconds of writes before saturation after which the write speed on the Evo drops to mere sub SATA 300 MB/s). The SX8200 is the new value leader and heralds a new era of competition for the mainstream segment of the NVMe SSD market. ^{[Jul '18 SSDrivePro]}

The new 3TB Seagate Barracuda 2016 (ST3000DM008) replaces its hugely successful predecessor, the 3TB Barracuda 7200.14 2011 (ST3000DM001). Comparing performance between the two models shows that the newer drive has 12% faster sequential speeds, comparable 4K speeds, improved mixed sequential speed and reduced mixed 4K speed. Overall, the effective speed is 12% faster on the 2016 model. Since there is normally little price difference between the two models the 2016 version is the clear winner especially for use as a backup drive with its impressive sequential read and write speeds of nearly 200 MBps. See the current value leaders here. ^{[Feb '17 HDrivePro]}

The 1TB Seagate Barracuda 2016 (ST1000DM010) has an impressive performance profile. With Sequential read/writes averaging 173 and 159 MBps respectively, the Barracuda can make short work of even moderately large backups. The small file (4K) performance profile is less impressive but still adequate with average read/writes coming in at 0.87 and 1.53 MBps respectively. For use as OS drives, rotational disks are quickly loosing market share to SSDs which offer orders of magnitude faster 4k read/write speeds. On the other hand cheap TLC based SSDs often have slower sustained write speeds than their rotational counterparts. Reasonably good overall performance can be achieved by using a TLC SSD to host the OS and a larger rotational drive such as the Barracuda for backups and media files. Larger capacity variants of this drive offer both better performance and better value for money. ^{[Feb '17 HDrivePro]}