Review AMD Radeon VII review and availability thread

[Hitman928]

Availability
------------------------------------------------------------------------------------------------------------------
As of the time of this post, In stock at newegg. (edit, and it's sold out).

Available straight from AMD.



Written Reviews
------------------------------------------------------------------------------------------------------------------
Techspot
See post #2 for more

Video Reviews
-------------------------------------------------------------------------------------------------------------------

 

[coercitiv]

VideoCardz has a review round-up - https://videocardz.com/79943/amd-radeon-vii-review-roundup

 

[Hitman928]

Thanks, I'll just add any I don't see on videocardz to the op.

 

[jpiniero]

Pretty much what was expected. 1080 Ti/2080 performance for 1080 Ti/2080's price.

 

[BFG10K]

I'm a bit disappointed overall but at least we have something that's better than Vega. That beefy cooler should theoretically be quiet but the actual tested load noise levels show its an epic fail.

Also while RTX/DLSS certainly suck balls, nVidia has far more robust drivers. Their parts consistently slot performance where they should be while VII is barely faster than Vega in many games. You can easily tell only games like BF5 and Tomb Raider have received proper driver optimizations.

Things will no doubt improve further down the line, but for $700 that's simply not good enough. Nobody drops down $700 with the expectation to wait months/years to get good performance.

 

[realibrad]

I'm a bit disappointed overall but at least we have something that's better than Vega. That beefy cooler should theoretically be quiet but the actual tested load noise levels show its an epic fail.

Also while RTX/DLSS certainly suck balls, nVidia has far more robust drivers. Their parts consistently slot performance where they should be while VII is barely faster than Vega in many games. You can easily tell only games like BF5 and Tomb Raider have received proper driver optimizations.

Things will no doubt improve further down the line, but for $700 that's simply not good enough. Nobody drops down $700 with the expectation to wait months/years to get good performance.

It feels like this card is to basically undercut Nvidia and get some revenue. I just does not seem like desktop graphics is important to AMD right now. Seeing Nvidia launch super expensive cards meant they could slot this card and make some money. If the Nvidia prices were not so high, this card would be horrible in margins as it would have to be much cheaper.

This appears to be a very expensive card to make.

 

[SKORPI0]

Wow, already OOS and reviews indicating that it sucks, not worth the $700 price.
Was ready to pull the trigger to replace my ancient R9 290X crossfire, looks like I'll have to wait.
Need a video card for 4K video re-encoding, not video games.

Now in back order at Newegg, limit 1 per customer.

 

[Rifter]

I'm a bit disappointed overall but at least we have something that's better than Vega. That beefy cooler should theoretically be quiet but the actual tested load noise levels show its an epic fail.

Also while RTX/DLSS certainly suck balls, nVidia has far more robust drivers. Their parts consistently slot performance where they should be while VII is barely faster than Vega in many games. You can easily tell only games like BF5 and Tomb Raider have received proper driver optimizations.

Things will no doubt improve further down the line, but for $700 that's simply not good enough. Nobody drops down $700 with the expectation to wait months/years to get good performance.

Agreed, the performance differential between games from the reviews so far looks like you are correct, only a few games are optimized for it so far. Nvidia has always had a lead in game optimization which is why whenever i buy AMD its almost never at launch.

Of course on the flip side of that AMD optimizes drivers for much longer than Nvidia. Nvidia is much more likely to move onto the new cards and abandon the old.

All in all though the new vega is priced where it needs to be and competes nicely with the 2080.

 

[AtenRa]

Also while RTX/DLSS certainly suck balls, nVidia has far more robust drivers. Their parts consistently slot performance where they should be while VII is barely faster than Vega in many games. You can easily tell only games like BF5 and Tomb Raider have received proper driver optimizations.

I believe it has to do with the games and not the drivers. It has to do whether the game is AMD or NV or if it is optimized for both. You can see that Radeon VII performance is around 20-35% faster than Vega 64 in every game at 4K.

https://www.anandtech.com/show/13923/the-amd-radeon-vii-review

You can see this at games that Vega 64 is bellow GTX 1080, then Radeon VII is slower than RTX 2080.
And in games that Vega 64 is faster/equal vs the GTX 1080 , Radeon VII is faster/equal against RTX 2080.

this review will help to illustrate my point.

https://www.guru3d.com/articles_pages/amd_radeon_vii_16_gb_review,11.html

 

[Arkaign]

Man I really wish Anandtech and others would retire the FE models from reviews, or at least deprecate them with a decent AIB in for more accurate representation of what people have. Outside of a handful of early adopters a couple of years ago it's so irrelevant to the reality of the AIBs that dominated the mass numbers of 10xx sold over the past 2+ years.

I find it particularly damning with the RTX reviews, as some of those models were ~10-15% and even 20% slower in some cases vs the 11GBPS AIBs.

 

[Mopetar]

Since they upped the double precision rate (1/4 vs 1/16 with Vega 64) this card might actually be good for people who want to run some compute workloads on the cheap.

I still don't think it's a good value as a gaming card though.

 

[ozzy702]

Extremely disappointing to see such poor performance in the majority of games vs the 2080, especially considering the power consumption and noise factors. AMD desperately needs a completely new architecture when 7nm cards are performing significantly worse per watt than 12nm and 16nm NIVIDIA counterparts.

The insane bandwidth obviously helped at 4k and in some scenarios at 1440p with higher "lows" than expected in some games which is awesome to see, but overall what a massive disappointment. This should be a $600 card.

 

[GodisanAtheist]

TPU has the card anywhere from 10-15% slower on average than a 2080 depending on the resolution (Larger difference at 1080p, smaller at 4K). That's actually a larger difference than the 5-10% I was initially expecting.

AMD really pulled off a feat here. They actually did something Nvidia couldn't do.

They made the RTX 2080 look like a great deal.

 

[Tup3x]

I don't see any reason to buy this over, say, GTX 2080. Horrible perf/w and incredibly noisy. GTX 2080 at least offers something new too.

 

[railven]

Going through some reviews, now that I've had a lot of coffee, and still woof.

I can't remember the last time a review like the TechSpot one in OP. The author didn't bite his tongue. He calls RTX a scam, and then STILL says that Radeon 7 is worse.

I'm not trying to dig up the past, but we're seeing again drivers just absolutely demolish an AMD card. If they want the meme to die, they have got to get this issue fixed consistently! Just about all the vlogs and few written reviews I've witnessed are talking about driver issues.

Also what is with this new method to report temperatures? Is there attempt to obfuscate actual temperatures? Curious why they'd go this route outside the obvious. Yes, it's more data, but you could have added this new data while still providing the older data.

The hole that Navi has to dig AMD out of is getting deeper and deeper.

 

[linkgoron]

I'm not sure why people are so surprised. The numbers are essentially what AMD said they'd be during their presentation. Quoting myself from a month ago...

Those numbers add up to an average of ~33% - which would place the Vega 7 at ~1080ti level and not 2080 level (which according to TPU is about 10% faster).

Essentially, nothing has changed from a month ago. It's around 10% slower than the 2080 and around 1080ti perf. It competes against a faster and more efficient 2080... The only thing going for it is "future proof" 16GB ram. I'm still waiting for the Vega magic drivers.

Nothing has really changed for RTG in the last 2/3 years. They need a node advantage to stay close to Nvidia's last generation's perf/watt. Polaris was around Maxwell perf/watt, and now Vega 20 is around Pascal perf/watt (somewhat weaker in that regard).

Pricing is ridiculous. This should've been at most $600. AMD are still paying the price for HBM/HBM2.

 

[moonbogg]

This thing is a piece of junk.

 

[railven]

I'm not sure why people are so surprised. The numbers are essentially what AMD said they'd be during their presentation. Quoting myself from a month ago...

I just finished watching the Gamer Nexus video. I was not expecting this thing to have a clear victory over RTX 2080. The biggest thing I'm taking away from the reviews/videos, is how horrendous AMD has handled the driver.

These things are crashing, left and right (EDIT: something I don't think anyone ever wants to read in a product review is "corrupted the system requiring a fresh install" That is beyond shooting yourself in the foot.) RTX got slammed for having features that were useless at launch. This thing is having issues just working! AMD needs to fire whomever thought this thing was ready for review/sell/public consumption.

This does not help AMD at all. Not as a "place holder" or anything. It's a black eye. Period. I don't even think the Pro AMD people can spin this around.

 

[linkgoron]

Computerbase.de cards undervolted really well.

Their undervolted card's volume is competitive with Nvidia:


Their undervolted card's Perf/watt is also better than the 2080, 1080ti and 1080:

I'm not sure how many cards would be stable with computerbase's settings, but I'll never get why AMD puts so much voltage in their cards.

 

[railven]

Computerbase.de cards undervolted really well.

Their undervolted card's volume is competitive with Nvidia:
View attachment 3105

Their undervolted card's Perf/watt is also better than the 2080, 1080ti and 1080:
View attachment 3108
I'm not sure how many cards would be stable with computerbase's settings, but I'll never get why AMD puts so much voltage in their cards.

Really curious on their results. Lots of reviews are stating overclocking software doesn't work, and Wattman barely works. Der8aeur put the thing on dry ice and got improvements with weird results.

With Gamer Nexus and Der8aeur reporting issues with Wattman, I'm going to assume just about any result regarding overclocking/undervolting is questionable.

 

[Hitman928]

Really curious on their results. Lots of reviews are stating overclocking software doesn't work, and Wattman barely works. Der8aeur put the thing on dry ice and got improvements with weird results.

With Gamer Nexus and Der8aeur reporting issues with Wattman, I'm going to assume just about any result regarding overclocking/undervolting is questionable.

Computerbase undervolt showed no performance loss but had significantly reduced measured power. I'd say it was working. Not saying the situation isn't buggy and unstable, just that unless computerbase is flat out making up benchmark and power numbers, it seems like their Wattman settings were working, at least for an undervolt.

 

[Elfear]

Really curious on their results. Lots of reviews are stating overclocking software doesn't work, and Wattman barely works. Der8aeur put the thing on dry ice and got improvements with weird results.

With Gamer Nexus and Der8aeur reporting issues with Wattman, I'm going to assume just about any result regarding overclocking/undervolting is questionable.

I'd like to see more results as well but you can't fool the watt meter or FPS (unless something is being done in the background to alter picture quality). If most of the Vega II cards can run at the same volts as Compterbase's card, it does make you wonder why AMD consistently pushes cards out the door with so much voltage overhead.

Interesting as well that Anandtech doesn't mention driver issues in their review. They have the Vega II 5-6% behind the 2080 FE overall. At $700 it's still a tough sell over the 2080 though.

 

[railven]

Computerbase undervolt showed no performance loss but had significantly reduced measured power. I'd say it was working. Not saying the situation isn't buggy and unstable, just that unless computerbase is flat out making up benchmark and power numbers, it seems like their Wattman settings were working, at least for an undervolt.

You should watch that Der8uaer video. Second video I watched today (first being Gamer Nexus) that states changes in Wattman either A) fail to safe B) give outlandish results or C) barely change anything.

Wattman is reporting temps hitting 65K Celcius. Gamer Nexus got wattman to read 7K Ghz overclock. Wattman is bugged. Jayz2Cents/Gamer Nexus/Der8auer are all reporting issues with Wattman, and reporting AMD is aware of issues.

You can see in the Der8auer video raising power load 20% didn't affect power draw. The software is broken, AMD is aware of it.

 

[railven]

I'd like to see more results as well but you can't fool the watt meter or FPS (unless something is being done in the background to alter picture quality). If most of the Vega II cards can run at the same volts as Compterbase's card, it does make you wonder why AMD consistently pushes cards out the door with so much voltage overhead.

Interesting as well that Anandtech doesn't mention driver issues in their review. They have the Vega II 5-6% behind the 2080 FE overall. At $700 it's still a tough sell over the 2080 though.

Yeah, you can't, and there are now 3 sources with two having similar results and one having much better results. It's crazy. I'd love for some results. As it is now, this is one hell of a roller coaster.

AMD is juicing their cards to the moon for a reason. I wonder if some people just got golden samples, or different versions of the software.

 

[rainy]

I'm not sure how many cards would be stable with computerbase's settings, but I'll never get why AMD puts so much voltage in their cards.

Quite obviously, RTG could not learn from their mistakes - it was the same story with Polaris and then Vega 64/56.

AMD is juicing their cards to the moon for a reason.

If you raise voltage to high enough, you can avoid proper testing which is money/time consuming and this is probably the main reason.