Overclock.net - An Overclocking Community - View Single Post - [Various] Radeon VII Reviews

View Single Post
post #13 of (permalink) Old 02-07-2019, 07:00 AM
Waiting for 7nm EUV
tpi2007's Avatar
Join Date: Nov 2010
Posts: 11,363
Rep: 894 (Unique: 503)

So against its primary competition, the RTX 2080, the Radeon VII is 5-6% behind, reference-to-reference. For our suite, the only games that it takes the lead are in Far Cry 5 and Battlefield 1, so perhaps less of ‘trading blows’ than it would appear. And if one considers the RTX 2080 Founders Edition as a reference design (albeit with non-reference TDP and boost clock), then this gap widens. Speculating on AIB designs is a little out-of-scope here, but given the open-air triple axial fan we’re seeing most, if not all, of the hand Vega 20 can play, as opposed to a traditional blower that by nature has some extra performance overhead.

So, depending on the review it's either on par overall or, like, AnandTech above, 5-6% slower against the non FE 2080 chip, with the gap widening against the 90 Mhz faster clocked FE card, or 4% slower overall in the case of TechSpot/Hardware Unboxed's review. GN, Hardware Unboxed and Jayz are saying that the drivers were incredibly buggy and Linus also mentioned it. Also, no Crossfire Support?

So, Lisa, you said that Jensen hadn't seen the card yet. Did you?

AMD kept adjusting parameters on the card even after having announced it. FP64 support is the obvious example here, which in this case is a good thing, but not for gamers, which ultimately means that they figured out that the card is not a competitive offer overall for gamers and thus decided to open FP64 even more to lure people interested in compute tasks:

The Radeon VII graphics card was created for gamers and creators, enthusiasts and early adopters. Given the broader market Radeon VII is targeting, we were considering different levels of FP64 performance. We previously communicated that Radeon VII provides 0.88 TFLOPS (DP=1/16 SP). However based on customer interest and feedback we wanted to let you know that we have decided to increase double precision compute performance to 3.52 TFLOPS (DP=1/4SP).

If you looked at FP64 performance in your testing, you may have seen this performance increase as the VBIOS and press drivers we shared with reviewers were pre-release test drivers that had these values already set. In addition, we have updated other numbers to reflect the achievable peak frequency in calculating Radeon VII performance as noted in the [charts].

There was an extra step in the conversation above: the first word was that it was 1:16, but then AnandTech got confirmation from AMD that it was 1:8 (Ryan Smith said so on Twitter). Now it's 1:4.

As inelegant as Jensen was, his comment of AMD having thought of the card that morning is starting to make a lot of sense.

Anyway, in summary, the current GPU gens from both makers are a pass. Both are transitional products while we wait for the real stuff on 7nm from Nvidia and Navi from AMD.

Last edited by tpi2007; 02-07-2019 at 07:17 AM.
tpi2007 is offline