Thursday, January 10th 2019
AMD Radeon VII Hands On at CES 2019
While many have watched or at the very least seen our coverage of AMD's live stream at CES 2019, it just can't compare to seeing the latest graphics card from the company up close and personal. Therefore as soon as we had the opportunity, we took a closer look at the AMD Radeon VII and let us just say the reference card is indeed a bit fancy. The shroud itself is made of metal and has a very similar look and feel to the one used on the Radeon RX Vega 64 liquid cooled reference cards. However, instead of using an AIO for this release AMD instead opted for three uniform fans and a massive heatsink. Not only does this make the card more compatible with small form factor systems, it is also less of a hassle to install. Display outputs consist of 3x DisplayPort and 1x HDMI. Sadly AMD did not include a VirtualLink port (USB Type-C) like NVIDIA for VR headsets, which is rather odd considering AMD is also part of the VirtualLink consortium.
Power delivery is handled by two 8-pin PCIe power connectors giving the card access to a theoretical limit of 375-watts which is 75-watts more than its 300-watt TDP. Considering the Radeon VII has the same power level as the Vega 64 it offers 25% more performance at the same power level. Compute unit count falls between the Vega 56 and Vega 64 at precisely 60 CUs. That said, a few missing CUs are of no consequence when you consider how close the Vega 56 performed to the Vega 64 once tweaked. As for clock speeds AMD has stated the Radeon VII will have a 1.8 GHz core clock, while the 16 GB of HBM2 will deliver 1 TB/s of memory bandwidth over the 4096-bit memory interface.
Overall gaming performance is 29% higher according to AMD with the Radeon VII having been tested in 25 titles in order to reach that conclusion. Eight of them were DX12 and two of them Vulkan meaning they used a decent spread of games across multiple APIs. In regards to the games tested they used; Assassin's Creed Odyssey, Battlefield 1, Battlefield V, Destiny 2, Doom, F1 2018, Fallout 76, Far Cry 5, Forza Horizon 4, Grand Theft Auto V, Strange Brigade, The Witcher 3, and Monster Hunter World just to name a few. Add that to the information shown in AMD's graphs and it appears it really can beat or at least trade blows with NVIDIA's GeForce RTX 2080. However, before jumping to any conclusions we will verify that soon enough once we have a sample in for review. In regards to pricing and availability it was already revealed earlier that AMD's Radeon VII will release at $699 on February 7th and will come bundled with a few games including, Devil May Cry 5, Resident Evil 2 and The Division 2 for a limited time.
Power delivery is handled by two 8-pin PCIe power connectors giving the card access to a theoretical limit of 375-watts which is 75-watts more than its 300-watt TDP. Considering the Radeon VII has the same power level as the Vega 64 it offers 25% more performance at the same power level. Compute unit count falls between the Vega 56 and Vega 64 at precisely 60 CUs. That said, a few missing CUs are of no consequence when you consider how close the Vega 56 performed to the Vega 64 once tweaked. As for clock speeds AMD has stated the Radeon VII will have a 1.8 GHz core clock, while the 16 GB of HBM2 will deliver 1 TB/s of memory bandwidth over the 4096-bit memory interface.
109 Comments on AMD Radeon VII Hands On at CES 2019
- consumer friendly (supposedly)
- not as greedy as Nvidia/Intel
- price disruption
Few want to buy AMD at the very top end unless it is actually BETTER performance for a lower price, considering the fewer features and underdog status.
And basically 699 is top mainstream, ignore the 1200 dreams of Titan-equivalent prices from Nvidia. People hope for something akin to how the 4850 was back in the day, forcing Nvidia to drop prices and making them pay for their greed.
But just like in the past, when AMD had a good performing Athlon, they positioned it to 1000USD immediately, showing everyone what they should have always known:
- AMD is just as greedy as Nvidia
- neither of the big 3 is consumer friendly
And since these corporations don't look out for US, the consumer, we shouldn't EVER say their pricing schemes are OK and just fine, and try to find justifications, like "16GB HBM2" and "still Vega". For the consumer, these things don't matter, price is what matters first, and consumers should look for THEIR interest, NOT the corporation's.
TL;DR: people should each of the big 3 when they come with greedy disappointing prices, and also reward them for products such as the GTX970 (yes, it worked fine even with the fake 0.5VRAM, even on full 4GB VRAM used), HD4850 or even the 8700K, which in my country sold for the same price as the 7700K and BELOW the 1700x and 1800x.
personally i am still sticking with gamefly subscription 2 games out at a time, thats all i have time for anyway, and i never have to buy games that way, can play a single story game, enjoy it, go at my own pace, then swap it in when done or beat. works great at a great price.
Assassin's Creed Odyssey, Battlefield 1, Battlefield 5, Call of Duty: Black Ops 4,Destiny 2,Deus x: Mankind Divided, Doom (2016), F1 2018, Fallout 76, Far Cry 5, Forza Horizon 4, Grand Theft Auto V, Hitman 2, Just Cause 4, Middle-Earth: Shadow Of War, Monster Hunter World, Rise of the Tomb Raider, Shadow of the Tomb Raider, Sid Meier's Civilization VI, Star Control: Origins, Strange Brigade, The Witcher 3, Tom Clancy's Ghost Recon Wildlands, Total War: Warhammer 2, Wolfenstein II: The New Colossus
I do that with a r9 290 with 1280x1024.I call BS on Top mainstream, no highend card for consumers is worth more than 500.
Only when you get into workstations which are typically corporation bought count.
source: hardforum.com/threads/here-are-amds-radeon-vii-benchmarks.1975236/Quite a large number of FuryX pump died within or just outside of 2yrs warranty when you search around. The default AIO quality is meh at best.
Source: me as previous FuryX owner had to pay out of pocket to get it fixed. Would not touch another factory AIO GPU.
The main issue is the fact that Vega 20 keeps the same 4 Geometry Engine limit all the way back from Hawaii.
The whole Geometry Fast Path / Next Gen Geometry feature was suppose to get around that, and it was suppose be done hardware/driver side, and that never happened on Vega 10.AMD certainly did not, it is all speculation until AMD release detail spec / the card is released.