keyword: Memory

gigabyte-radeon-rx-6700-xt-gaming-oc-review

Gigabyte Radeon RX 6700 XT Gaming OC Review

Introduction

The Gigabyte Radeon RX 6700 XT Gaming OC is the company’s factory-overclocked graphics card for those seeking a close-to-reference custom rendition of the RX 6700 XT from Gigabyte without the bells and whistles of the AORUS Gaming brand. The Radeon RX 6700 XT is AMD’s fourth RX 6000 series RDNA2 graphics card, and arguably its most important so far as it targets a sub-$500 (MSRP) price point, bringing the architecture to a wider audience. A successor to the RX 5700 XT, it brings full DirectX 12 Ultimate readiness, including real-time raytracing, and is suited for maxed out gaming at 1440p. AMD claims that the card is competitive not only against NVIDIA’s RTX 3060 Ti, but also the pricier RTX 3070 in certain games.

The RDNA2 graphics architecture powering the RX 6700 XT is spread far and wide in the current generation, spanning not just the Radeon RX 6000 series, but also the latest game consoles. This makes it easier for game developers to optimize for the architecture on the PC. AMD’s approach to real-time raytracing involves using special hardware called Ray Accelerators to handle the most compute-intensive task in raytracing, of ray intersections, while compute shaders are used for almost everything else, including de-noising. A side effect of this approach is that AMD has had to bolster its SIMD muscle significantly over the previous generation, which can work wonders for conventional raster 3D games.

The Radeon RX 6700 XT is based on and maxes out the new 7 nm Navi 22 silicon, which physically features 40 RDNA2 compute units. This works out to 2,560 stream processors, 40 Ray Accelerators, 160 TMUs, and 64 ROPs. The stream processor count is exactly the same as with the RX 5700 XT, but besides having a higher IPC, they run at much higher engine clocks in excess of 2.42 GHz, compared to the 1.77 GHz game clock of the RX 5700 XT.

AMD has increased the standard memory amount to 12 GB, but over a narrow 192-bit GDDR6 memory bus. AMD attempts to overcome the bandwidth deficit compared to the 256-bit GDDR6 interface of the RX 5700 series by increasing the memory clocks to 16 Gbps and deploying its new Infinity Cache technology—a fast 96 MB on-die level 3 cache that accelerates the memory subsystem.

The Gigabyte RX 6700 XT Gaming OC comes with the company’s latest-generation WindForce 3X cooling solution found on several other current-generation products. Three aluminium fin stacks are skewered by five 6 mm-thick copper heat-pipes that make direct contact with the GPU at the base. This heatsink is ventilated by three fans. The cooler is longer than the PCB, so some of the airflow from the third fan flows through a hole in the backplate. The RX 6700 XT comes with factory overclocked speeds of 2514 MHz (game clock) as opposed to the 2424 MHz reference. In this review, we take the card for a spin.

Radeon RX 6700 XT Market Segment Analysis
  Price Shader

Units		 ROPs Core

Clock		 Boost

Clock		 Memory

Clock		 GPU Transistors Memory
RX Vega 64 $400 4096 64 1247 MHz 1546 MHz 953 MHz Vega 10 12500M 8 GB, HBM2, 2048-bit
RX 5700 XT $370 2560 64 1605 MHz 1755 MHz 1750 MHz Navi 10 10300M 8 GB, GDDR6, 256-bit
RTX 2070 $340 2304 64 1410 MHz 1620 MHz 1750 MHz TU106 10800M 8 GB, GDDR6, 256-bit
RTX 3060 $600 3584 48 1320 MHz 1777 MHz 1875 MHz GA106 13250M 12 GB, GDDR6, 192-bit
RTX 2070 Super $450 2560 64 1605 MHz 1770 MHz 1750 MHz TU104 13600M 8 GB, GDDR6, 256-bit
Radeon VII $680 3840 64 1400 MHz 1800 MHz 1000 MHz Vega 20 13230M 16 GB, HBM2, 4096-bit
RTX 2080 $600 2944 64 1515 MHz 1710 MHz 1750 MHz TU104 13600M 8 GB, GDDR6, 256-bit
RTX 2080 Super $690 3072 64 1650 MHz 1815 MHz 1940 MHz TU104 13600M 8 GB, GDDR6, 256-bit
RTX 3060 Ti $700 4864 80 1410 MHz 1665 MHz 1750 MHz GA104 17400M 8 GB, GDDR6, 256-bit
RX 6700 XT $800

MSRP: $480		 2560 64 2424 MHz 2581 MHz 2000 MHz Navi 22 17200M 12 GB, GDDR6, 192-bit
Gigabyte RX 6700 XT

Gaming OC		 $820 2560 64 2424 MHz 2614 MHz 2000 MHz Navi 22 17200M 12 GB, GDDR6, 192-bit
RTX 2080 Ti $1000 4352 88 1350 MHz 1545 MHz 1750 MHz TU102 18600M 11 GB, GDDR6, 352-bit
RTX 3070 $800 5888 96 1500 MHz 1725 MHz 1750 MHz GA104 17400M 8 GB, GDDR6, 256-bit
RX 6800 $1000 3840 96 1815 MHz 2105 MHz 2000 MHz Navi 21 26800M 16 GB, GDDR6, 256-bit
RX 6800 XT $1300 4608 128 2015 MHz 2250 MHz 2000 MHz Navi 21 26800M 16 GB, GDDR6, 256-bit
RTX 3080 $1300 8704 96 1440 MHz 1710 MHz 1188 MHz GA102 28000M 10 GB, GDDR6X, 320-bit
RX 6900 XT $1500 5120 128 2015 MHz 2250 MHz 2000 MHz Navi 21 26800M 16 GB, GDDR6, 256-bit
RTX 3090 $2000 10496 112 1395 MHz 1695 MHz 1219 MHz GA102 28000M 24 GB, GDDR6X, 384-bit
nvidia-reportedly-transforming-a100-into-a-mining-gpu,-with-a-210-mh/s-hash-rate

Nvidia Reportedly Transforming A100 Into a Mining GPU, With a 210 MH/s Hash Rate

Nvidia A100
(Image credit: Nvidia)

Nvidia’s A100 accelerator, which is based on the GA100 silicon, might not be hitting the sales numbers the company hoped for–or perhaps the company just thinks there’s room for bigger sales in a different market. Hardware leaker kopite7kimi, who has a solid track record with Ampere leaks, broke the news that Nvidia is allegedly turning the A100 into a cryptocurrency powerhouse.

Obviously, the Ampere graphics card will launch under Nvidia’s Cryptocurrency Mining Processor (CMP) series that’s purposely built for professional mining, so you won’t have to worry about Nvidia’s anti-mining hash rate tomfoolery. Although kopite7kimi didn’t share performance numbers, he claimed that it’ll be “a mining monster,” and we don’t doubt his assertion, considering the A100’s impressive specifications.

I_Leak_VN, another hardware leaker, is pretty confident that the A100 will debut as the CMP 220HX, with a whopping hash rate up to 210 MH/s. Since CMP is built for Ethereum, we assume that figure corresponds to the device’s performance in mining Ethereum. If the number is accurate, the 220HX would deliver up to two times higher Ethereum performance than a GeForce RTX 3090, which is one of the fastest mining graphics cards on the market.

The A100 PCIe ticks all the right boxes for a cryptocurrency mining part, which may explain why Nvidia is reportedly making the 220HX in its image. For starters, the accelerator features a dual-slot design and passive cooling, so miners won’t have to worry about heat. More importantly, the A100 PCIe is rated for 250W, so it should also help keep electricity bills under control.

Nvidia CMP 220HX Specifications

Model 30HX 40HX 50HX 90HX 220HX*
Ethereum Hash Rate 26 MH/s 36 MH/s 45 MH/s 86 MH/s 210 MH/s
Rated Power 125W 185W 250W 320W ?
Power Connectors 1 x 8-pin 1 x 8-pin 2 x 8-pin 2 x 8-pin ?
Memory Size 6GB 8GB 10GB 10GB ?
Starting Availability Q1 Q1 Q2 Q2 ?

*Specifications are unconfirmed.

Ethereum is addicted to memory bandwidth, and that’s where the A100 arguably excels the most. We’re uncertain just how close the 220HX will be to the A100, but even a fraction of the A100 should turn the 220HX into an Ethereum-crunching beast.

The A100 PCIe is equipped with 40GB of HBM2e memory, which operates at 2.4 Gbps across a 5,120-bit memory interface. If we do the math, the maximum theoretical memory bandwidth works out to 1,555 GBps. However, it would be naive to think the 220HX would come with that much memory since Ethereum has no use for that much. Nvidia will likely chop off some of the memory to make the graphics card more attractive in regards to pricing.

The retail price for the A100 in its PCIe format is well over $10,000. According to the I_Leak_VN, the 220HX could land around the $3,000 range. For comparison, a quartette of GeForce RTX 3060 with the help of some $6 HDMI dummy plugs can collectively put up a 192 MH/s on Ethereum. We know you basically have to visit fantasy to find a GeForce RTX 3060 at Nvidia’s $329 MSRP, but in a perfect world, a four-way GeForce RTX 3060 setup will only set you back $1,316, and with some tinkering, could probably catch up to the 220HX’s rumored Ethereum performance. But the 220HX makes a lot of sense for professional mining farms, where you can get the same performance with a single unit instead of four.

The pricing for Nvidia’s CMP HX offerings remains a mystery, although we’ve seen a CMP 30HX land overseas for $723. It’ll be interesting to see if the $3,000 price tag sticks with the 220HX. Obviously, mining farms won’t be buying just one, so Nvidia may also be tempted to offer a juicy discount for bulk orders. Let’s not forget that between $100 to $300 million of the chipmaker’s Q4 2021 revenue was thanks to sales to cryptominers, so it’s obviously a very lucrative business.