MSI Radeon RX 5700 Gaming XOctober, 2019
AMD's Radeon RX 5700 series is finally getting the much-awaited custom variants and MSI is out with an entirely new lineup designed just for RDNA based graphics cards. Launched last month, the AMD Radeon RX 5700 lineup introduced very competitive prices for mainstream tier graphics cards which would go against the NVIDIA GeForce RTX lineup, now AIBs are further expanding the lineup with their own non-reference variants that offer better cooling performance and higher out of box clock speeds.
The Radeon RX 5700 series uplifted AMD by bringing a modern architecture design and moving away from their GCN design. This allows AMD to bring more streamlined graphics performance in modern workloads and gaming titles. AMD was already ahead of the curve in utilizing new techs such as HBM and smaller process nodes and Navi is no exception. Aside from the new graphics architecture, AMD has also introduced GDDR6 memory and a smaller 7nm process node for their mainstream lineup which is a big update from the 14nm process on Polaris and Vega series cards.
Compared to NVIDIA's RTX 20 SUPER lineup, the AMD Radeon RX 5700 is much cheaper. The AMD Radeon RX 5700 XT is $100 cheaper than the GeForce RTX 2070 SUPER while the Radeon RX 5700 is $50 US cheaper than the GeForce RTX 2060 SUPER and costs the same as the GeForce RTX 2060 (non-SUPER). The AMD Radeon RX 5700 XT costs closer to the RTX 2070 but that card has been replaced by the new SUPER option which means that the RX 5700 XT, while positioned against the GeForce RTX 2070 SUPER is priced at RTX 2060 SUPER level.
Well, in terms of performance the AMD Radeon RX 5700 XT is supposed to be 10% faster than the RTX 2070 on average and the Radeon RX 5700 is supposed to be 10% faster than the RTX 2060 on average. The SUPER cards are almost 15% faster than their predecessors on average and since the Radeon RX 5700 series is much lower-priced, the should offer slightly better value. The biggest take away is that Radeon RX 5700 series doesn't support extra RTX features such as Ray-Tracing, DLSS that do make the RTX series a more compelling option and future-proof for next-gen titles that are going to support these features.
So we can say that the AMD Radeon RX 5700 series is great for users who are purely eyeing raw performance in gaming at better prices. The Radeon RX 5700 series is definitely a much-needed lineup and an upgrade from the older Polaris cards but we will find out if they hold up in our tests.
So for this review, I will be taking a look at the MSI Radeon RX 5700 Gaming X. This is MSI's new and flagship custom design for the Navi 10 Pro GPU that features dual TORX 3.0 fans along with the renowned MSI features such as Zero Frozr and Smooth heat pipe design. The card will retail at $379.99 US which is a $30 US premium over the reference model.
The AMD Radeon RX 5700 Series Family
The Radeon RX 5700 series includes three graphics cards, the Radeon RX 5700 XT, Radeon RX 5700, and the Radeon RX 5700 XT Anniversary Edition. The Navi based Radeon RX 5700 series is also the first graphics lineup to feature PCIe 4.0 support which offers twice the bandwidth when compared to PCIe 3.0.
AMD Radeon RX 5700 XT Official Specifications ($399 US)
Starting with the specifications, the Radeon RX 5700 XT comes with 40 compute units in total and since AMD has already confirmed that the Compute unit design still features 64 stream processors, we will be getting 2560 stream processors in total. The AMD Navi GPU featured on the Radeon RX 5700 series comes with 160 Texture Mapping Units and 64 Raster Operation units.
The chip itself is clocked at 1605 MHz base clock but includes two additional clock speeds, a boost clock, and a game clock. The boost clock is rated at 1905 MHz while the game clock is rated at 1755 MHz. The difference between the three clock speeds is that the base clock is the target under full load (power virus), the game clock would be the traditional boost target under gaming while the boost clock is the maximum target that the card could achieve (based per chip).
With the said boost clock, AMD expects a maximum of 9.75 TFLOPs of single-precision Compute from the Radeon RX 5700 XT. The card also features 8 GB of GDDR6 memory which runs across a 256-bit wide bus interface. AMD will be using the latest 14 Gbps memory dies which put them on par with the Turing TU104 cards that offer bandwidth of up to 448 GB/s. The card also features two 8 pin connectors and has a total board power or TBP of 225W. The graphics card costs $399 US in reference flavors and a slight premium for the non-reference variants such as the MSI EVOKE OC which I am testing today.
AMD Radeon RX 5700 Official Specifications ($349 US)
The second card is the Radeon RX 5700 based on the Navi Pro GPU. The reason we are not getting a Radeon RX 5700 Pro naming scheme is that it would be harder to differentiate that with AMD's own pro series cards which are aimed at content creators and workstation PCs.
This card has 2304 stream processors, 144 TMUs, 64 ROPs. The clocks are maintained at 1465 MHz base, 1725 MHz boost clock and 1625 MHz game clock. At peak boost clocks, the card will be able to deliver 7.95 TFLOPs of Compute performance. The card features an 8+6 pin connector config & has a rated TBP of 180W.
Now based on the TBP numbers, this card should be put against the RTX 2070 which is a 175W TBP graphics card. It will be interesting to compare both cards in terms of efficiency since the NVIDIA Turing cards are based on 12nm FinFET while AMD is using the latest 7nm process node. The card costs $349 US for reference flavors.
AMD Radeon RX 5700 XT 50th Anniversary Edition Official Specifications ($449 US)
In addition to the AMD Radeon RX 5700 XT and Radeon RX 5700, AMD also introduced a 50th Anniversary Edition variant of their Radeon RX 5700 XT featuring a black and gold shroud with frequencies of 1680 MHz base clock, 1830 MHz game clock and boost clocks of up to 1980 MHz. This variant would deliver a total Compute power of 10.14 TFLOPs and should be around 5-10% faster than the Radeon RX 5700 XT. The card will be rated at a 235W TBP.
The reference variant of the Radeon RX 5700 XT cards would feature an all-aluminum alloy shroud and backplate. Inside the card is an enhanced vapor chamber which is cooled off by a blower fan. The base of the vapor chamber makes use of graphite thermal interface material which is similar to the pads used on the Radeon VII graphics card. The PCB of the card offers a 7 phase digital VRM which AMD says is designed for overclocking. The Anniversary Edtion costs $449 US and comes in reference only flavors.
AMD Radeon RX 5000 '7nm Navi RDNA' GPU Lineup Specs:
|Graphics Card||Radeon RX 5700 XT 50th Anniversary||Radeon RX 5700 XT||Radeon RX 5700||Radeon RX 5600 XT||Radeon RX 5500 XT||Radeon RX 5300|
|GPU Architecture||7nm Navi (RDNA 1st Gen)||7nm Navi (RDNA 1st Gen)||7nm Navi (RDNA 1st Gen)||7nm Navi (RDNA 1st Gen)||7nm Navi (RDNA 1st Gen)||7nm Navi (RDNA 1st Gen)|
|Stream Processors||2560 SPs||2560 SPs||2304 SPs||2304 SPs||1408 SPs||1408 SPs|
|TMUs / ROPs||160 / 64||160 / 64||144 / 64||144 / 64||88 / 32||88 / 32|
|Base Clock||1680 MHz||1605 MHz||1465 MHz||1130 MHz||1670 MHz||TBD|
|Boost Clock||1980 MHz||1905 MHz||1725 MHz||1560 MHz||1845 MHz||1645 MHz|
|Game Clock||1830 MHz||1755 MHz||1625 MHz||1375 MHz||1717 MHz||1448 MHz|
|Compute Power||10.14 TFLOPs||9.75 TFLOPs||7.95 TFLOPs||7.19 TFLOPs||5.19 TFLOPs||4.63 TFLOPs|
|VRAM||8 GB GDDR6||8 GB GDDR6||8 GB GDDR6||6 GB GDDR6||8 GB GDDR6|
4 GB GDDR6
|3 GB GDDR6|
|Bandwidth||448 GB/s||448 GB/s||448 GB/s||288 GB/s||224 GB/s||168 GB/s|
|Price||$449 US||$399 US||$349 US||$279 US||$169 US (4 GB)|
$199 US (8 GB)
|Launch||7th July 2019||7th July 2019||7th July 2019||21st January, 2020||7th October 2019||28th August, 2020|
Radeon RX 5700 "7nm Navi RDNA GPU" Feature Set and A Word on HW-Enabled Ray Tracing
While we would share a few tidbits of the RDNA architecture itself below, there are also some highlights we should mention for the Navi GPU. According to AMD themselves, the Navi 10 GPU will be 14% faster at the same power and should consume 23% lower power at the same clock speeds as Vega 64 GPU. The AMD Navi GPU has a die size of 251mm2 and delivers 2.3x perf per area over Vega 64. The chip packs 10.3 Billion transistors while the Vega 10 GPU packed 12.5 Billion transistors on almost twice the die space.
Also, when it comes to ray tracing, AMD is indeed developing their own suite around it. According to their vision, current GCN and RDNA architecture will be able to perform ray tracing on shaders which will be used through ProRender for creators and Radeon Rays for developers. In next-gen RDNA which is supposed to launch in 2020 on 7nm+ node, AMD will be bringing hardware-enabled ray tracing with select lighting effects for real-time gaming. AMD will also enable full-scene ray tracing which would be leveraged through cloud computing.
Radeon Multimedia Engine – Seamless Streaming
- Improved Encoding (New HDR/WCG Encode HEVC)
- 8K Encode (HEVC & VP)
- 40% encoder speedups
- 40 RDNA Compute Units
- 80 Scalar Processors
- 2560 Stream Processors
- 160 64b bilinear filter units
- Multilevel Cache
- 4MB L2, 512Kb L1
- 2x V$L0 Load Bandwidth
- DCC Everywhere
- Streamlined Graphics Engine
- Geometry Engine (4 Prisms Shader Out, 8 Prim Shader In)
- 64 Pixel Units
- 4 Asynchronous Compute Enginers
- Balanced Work Distribution & Redistribution
- Designed for higher frequencies at lower power
New Compute Unit Design
Great Compute Efficiency For Diverse Workloads
- 2x Instruction Rate (enabled by 2x Scalar Units and 2x Schedulers)
- Single Cycle Issue (enabled by Executing Wwave32 on SIMD32)
- Dual Mode Execution (Wave 32 and Wave 64 Modes Adapt for Workloads)
- Resource Pooling (2 CUs Coordinate as a Work Group Processor)
As you can tell, AMD is changing a lot in terms of architecture with RDNA (Radeon DNA) compared to GCN. There's a new Compute unity design, a more streamlined Graphics pipeline & a multi-level cache hierarchy. Aside from the GPU architecture, support for GDDR6 memory is another major change that brings AMD's graphics cards on par with NVIDIA in utilizing modern memory designs for higher bandwidth.