AMD Radeon VII 16 GB

[Xzibit]

\
Look how bad AMD sucks in Ace Combat 7, which is a UE4 title, nvidia is basically miles ahead of AMD when it comes to less known or older titles.
Go bash AMD instead of the reviewer.

I would expect as much given their partnership

"Epic developed Unreal Engine 4 on NVIDIA hardware, and it looks and runs best on GeForce."

 

[efikkan]

A wide selection of games should be a requirement for any review, as it evens out some of the outliers. It also tends to cancel out some of the shader-tweaks both vendors are doing ("optimizations" as they call it, but is really cheating). So a good selection of scalable games can only be positive to illustrate the true and unbiased performance of a card, which is also the only metric that gives us an indication useful for predicting future games.

I wouldn't worry about game engines like Unreal being biased, if anything it's the opposite, it's one of the most bloated engines there is. Many of the games using it are not scalabe enough.

 

[M2B]

Many of the games using it are not scalabe enough.

What do you exactly mean by "scalable"? Explain please.

 

[efikkan]

What do you exactly mean by "scalable"? Explain please.

As you know, some games simply don't benefit from higher performance. I would consider benchmarking such games irrelevant, unless the point is to check if the game runs or not.

 

[purecain]

where is everyone thinking of putting the card straight under water? I wonder how the card reacts?

theres a lot of fun to be had experimenting with this card. I hope crossfire is working...

ps I also think the dodgyness of the cards drivers etc is all part of the amd experience. I used to get really excited over getting good frames and decent benchmarking results...

 

[M2B]

As you know, some games simply don't benefit from higher performance. I would consider benchmarking such games irrelevant, unless the point is to check if the game runs or not.

As for the Unreal Engine 4, all I see here is good scalability across all different segments/generations of cards, turing is exceptionally good at UE4 titles and the RVII provides significant performance jump over the Vega 64, too.
Recent Assassin's creed games are not that good at scalability, but honesly, it doesn't matter to the generall reader, all they want to see is the performance on the latest AAA titles.

 

[rtwjunkie]

I strongly criticize reviewer for his selection of games being heavily weighted against AMD. Almost every game is DX11, when everyone knows DX 12 games run better on AMD graphics.

So AMD users only play DX12 games? If they don’t, then sit down and be quiet. It is a selection that covers many of the games played by AMD and Nvidia users, because gamers are gamers.

Gamers play games based on fun factor, not whether it has been tuned to their species of GPU. Your complaint against W1zzard is silly, frivolous, and even ludicrous.

 

[notb]

Again, Wizzard's performance summary paints Radeon 7 in the worst possible light out of all the big sites. It's like an anomaly.

Mathematically speaking, because there is a finite set of review sites, one of them had to be the one that puts AMD in the worst light.
So actually it's not an anomaly, but a necessity.

This is partly because of the choice of old titles or severely unoptimized games for AMD (this is no-one but AMD's fault, but still, it skews the performance figures) included in the suite.

No, it doesn't.
You know which games were used and which settings - you have the data you need to make a conscious decision as an intelligent human being.

An issue would arise if we would not know which games were used or how well these cards performed in each of them.

That is a joke. Sure, if you play old unoptimized DX11 games, that's the case.

So you'll forbid us to play old games now? Seriously?

In games where the drivers are there, like Battlefield 5, it is faster than a 2080 and will only continue to get faster in the coming months.

So there's a clear recommendation from user @Shatun_Bear : if you need a card today, buy something else. If you can wait until Navi arrives, buy Radeon VII. Oops...

I’m curious why the mix of titles is a bad thing. Will a Radeon 7 user only play games in which the game was optimized for AMD or made in DX12? With my last AMD card I didn’t decide to not play certain games just because they work better with Nvidia cards.

Actually, that quite possibly is correct.
Do we still remember Ashes of the Singularity? As a game: not so good. As a benchmark convincing you that you made a good choice buying AMD: perfect.
And this game wouldn't be remotely as popular as it used to, if not for strong interest from AMD fans.
In fact in one of discussions on this forum a known AMD supporter said he would buy and play this game only to get the feeling he is using the whole potential of his hardware.

I really don’t understand why then we shouldn’t see a performance summary rating, since it includes a wide variety of games played.

Because some people don't understand statistics and because calculating any aggregate always makes some people confused and enraged.
It's like with these people who constantly moan that an average salary is stupid because more than half of population earns less.
@W1zzard We totally appreciate the work you're doing processing the data and preparing the final presentation. But have you ever considered just supplementing the raw measurements?
Wouldn't that be an awesome new phenomenon on PC review sites?

I am suggesting him some changes in his gamelist that would make the results better balanced.

If you're suggesting games based on whether they hurt AMD or not, you're actually adding some bias AMD supporters are so worried about all the time.
For an ideally unbiased set of games, we would have to take all the titles that exist and draw few (random sample). I doubt that would make the review more sensible.
So maybe a popularity approach? For example: get the 20 most popular demanding games from Steam?
It could be biased, but would you agree it's at least representative? Well.. Civ VI would certainly be on that list...

And don't be silly - Nvidia have invested far more money and investment into their DX12 drivers than AMD and they're still behind in lots of titles.

If Nvidia was ahead in everything, I guess we wouldn't be having this discussion, right?
But yeah, it's quite possible that if you spend more money on drivers, software may work better on your hardware. AMD should try that.

omg.. its bad enough the fan boy bickering, but do we really need to drag flat earthers in to this ??

It's almost certain that flat-earthers are the dumbest representants of our species. So yeah, I think it's worth mentioning from time to time.

Imagine you don't know what a flat-earther is and one day your daughter tells you that she has a boyfriend who plays the piano and is a flat-earther. And you think: "hmm, a he plays the piano - he'll make a great son-in-law one day!"

 

[Shambles1980]

Imagine you don't know what a flat-earther is and one day your daughter tells you that she has a boyfriend who plays the piano and is a flat-earther. And you think: "hmm, a he plays the piano - he'll make a great son-in-law one day!"

you can change the word(s) "flat-earther" in to any religion you want and then its a harder stance to take. although probably equally as valid. Yet still has no place in a tech forum discussion

 

[RichF]

I think people are being way too....aggressive over this.

We knew it wont be that amazing. It isn't a bad card either. It is just loud, and over volted. Clearly it shows that the voltage can be reduced considerably. Noise is something I cant stand.

Which is it? Not bad or something you can't stand?

The undervolting solution is based on speculation. Just because some cards can be undervolted quite a bit doesn't mean they all can, unless you or anyone else has solid evidence to support that speculation.

And, if that speculation is actually correct then AMD should be pummelled for gross incompetence — by sending out review samples that were grossly overvolted and/or selling such cards to consumers.

Also, since there has been talk about Navi's release date, here is the latest from the rumour mill:

AMD Navi 7nm GPUs Based Radeon RX Graphics Cards Reportedly Delayed Till October 2019
https://wccftech.com/amd-navi-7nm-radeon-rx-gpus-delayed-q4-2019/

Another reason could be that the Navi GPUs are also allegedly going to be featured in next-gen consoles and AMD would want to dedicate a good chunk of supply for those before they ship the chips out to AIBs for production of desktop-based Radeon RX graphics cards.

Yeah, AMD is really working for PC gamers, if this is true.

I've told people many times that the console scam* hurts everyone. If this rumour is accurate then it's even more sharp-and-stinking evidence to back that.

*Selling low-end PCs with artificially redundant software walled gardens is a scam. They're not consoles — they are cheap PCs that aren't compatible, intentionally and for nothing but an anti-consumer purpose, with the PC gaming software standard. Something like the Switch, which uses novel form factors and usage modes, is a different story. Sony and MS "consoles" could very easily be eliminated from the market and replaced by a Vulkan + OpenGL on Linux universal PC gaming platform. It's not necessary to pay the Windows tax.

 

[sepheronx]

You should learn to differentiate between performance and noise.

I'm sensitive to noise. But I'm willing to accept variations in performance and power use.

Maybe for some I should have specified that?

 

[RichF]

You should learn to differentiate between performance and noise.

I'm sensitive to noise. But I'm willing to accept variations in performance and power use.

Maybe for some I should have specified that?

You can't have your cake and eat it, too.

Claim 1: "It's loud. Noise is something I can't stand."
Claim 2: "It's not a bad card."

This is obvious so I'm not going to spend more time on it.

 

[GoldenX]

2019, the year of the shill. RTX Shills, Radeon 7 Shills, Intel i9 Shills, Zen Shills.

 

[sepheronx]

You can't have your cake and eat it, too.

Claim 1: "It's loud. Noise is something I can't stand."
Claim 2: "It's not a bad card."

This is obvious so I'm not going to spend more time on it.

I wasn't aware you made the rules regarding how I may measure the card. So it is either good or bad based upon exactly every criteria? So noise being bad and mentioning it as such still means card is bad overall, even though that maybe one can work around the noise in turn to enjoy the performance of the card if it is good? It sounds more like you are being pretentious.

May be obvious to you. Maybe not to others. I will take your constructive criticism into consideration next time I decide to measure how good or bad a device is.

 

[os2wiz]

Nobody buys a GPU just to play latest AAA Games, games such as Dragon Quest Xi are a good indicator of the performance in lesser known titles:

View attachment 116129

Look how bad AMD sucks at Ace Combat 7, which is a UE4 title, nvidia is basically miles ahead of AMD when it comes to less known or older titles.
Go bash AMD instead of the reviewer.

No the developers should be trashed they don't have the deep pockets to pay every developer to optimize their products for AMD hardware like Nvidia. THE manuals and software tools are available for free to code the optimizations. But the lazy f--ks will not do it unless AMD slips them cash and sends a consultant to help them with the task.

 

[c2DDragon]

This card is a joke.
I'm so conforted I didn't wait for Volta (the nVidia's unicorn) and did buy a 1080 Ti at the start of 2018 when prices were back to launch ones.
Right now if you want to upgrade, what do you have ? Overpriced failures VS Overpriced fake innovations.
AMD shouldn't have bought ATI. I can't remember a good power controlled performing card from the red team since it's called AMD...
Intel should launch in the dedicated GPU market it could be a not-so-bad thing from a consumer point of view I think.

 

[HD64G]

If you're suggesting games based on whether they hurt AMD or not, you're actually adding some bias AMD supporters are so worried about all the time.
For an ideally unbiased set of games, we would have to take all the titles that exist and draw few (random sample). I doubt that would make the review more sensible.
So maybe a popularity approach? For example: get the 20 most popular demanding games from Steam?
It could be biased, but would you agree it's at least representative? Well.. Civ VI would certainly be on that list...

I am simply suggesting not having 3 games using the same engine (Unreal engine 4) with 2 of them showing much worse performance for a specific brand (Only Senua's Sacrifice is optimised well for both AMD and nVidia out of those 3). Outliers aren't good for objectivity. And I don't think Civ6 is neutral in how it performs for AMD. But the solution to this could be for @W1zzard to show the performance of all games and he can count only the not biased ones (<25% diff for the same tier GPUs) in the summary. A win-win scenario for all imho.

 

[R0H1T]

Could geometric mean be more useful, instead of arithmetic

 

[efikkan]

No the developers should be trashed they don't have the deep pockets to pay every developer to optimize their products for AMD hardware like Nvidia. THE manuals and software tools are available for free to code the optimizations. But the lazy f--ks will not do it unless AMD slips them cash and sends a consultant to help them with the task.

Like many, you have misconceptions of how games/game engines work.
It is exceedingly rare that games have specific optimizations for hardware, and even making such optimizations would be nearly pointless, since they would be tied to specific iterations of GPU architectures, not vendors.

The kind of bias that exists in some games are usually not intentional, but simply a consequence of the game being developed for and tested on one set of hardware. Such bias is in most cases favoring AMD, as many AAA titles are developed exclusively for AMD based consoles, while "no" games are developed exclusively for Nvidia.

And I want to emphasize; just because card A scales better than card B in a game, doesn't mean there is bias in the game engine, it can simply be a consequence of better balance of resources on that card. So a card performing better in some games than other is not proof of bias, but often a mismatch between people's expectations and the reality.

 

[HD64G]

Some new benchmarks using the latest public driver from AMD, show that R7 isn't as bad as it seems vs RTX2080 (-7% @1440P on average of 33 games)

Spoiler: 1440P 33 games graph @8m30s

 

[notb]

I am simply suggesting not having 3 games using the same engine (Unreal engine 4) with 2 of them showing much worse performance for a specific brand (Only Senua's Sacrifice is optimised well for both AMD and nVidia out of those 3). Outliers aren't good for objectivity. And I don't think Civ6 is neutral in how it performs for AMD. But the solution to this could be for @W1zzard to show the performance of all games and he can count only the not biased ones (<25% diff for the same tier GPUs) in the summary. A win-win scenario for all imho.

The test includes 21 games out of dozens of popular titles. How do you know they're representative for all games? And how do you know they're representing for gamers' needs?
If you don't know that, how can you say what's an outlier?
Because with such a small sample, how do you know that few games you've mentioned are outliers in the population?

Don't use the word "outlier". It contains a huge statistical burden.
If you don't like some results, don't look at them. TPU provides figures for each game.

The choice of games currently used by TPU is not representing my taste. I find only 6 out of 21 interesting, played 4 of them. I don't look at the average score when I choose a GPU. You shouldn't as well.

 

[HD64G]

The test includes 21 games out of dozens of popular titles. How do you know they're representative for all games? And how do you know they're representing for gamers' needs?
If you don't know that, how can you say what's an outlier?
Because with such a small sample, how do you know that few games you've mentioned are outliers in the population?

Don't use the word "outlier". It contains a huge statistical burden.
If you don't like some results, don't look at them. TPU provides figures for each game.

The choice of games currently used by TPU is not representing my taste. I find only 6 out of 21 interesting, played 4 of them. I don't look at the average score when I choose a GPU. You shouldn't as well.

Outlier is meant as a statistical paradox greatly increacing the variance. And I know of those things from my time in university, so I think I put this word in good practice here.

I agree with what else you wrote (not looking at the average is a good practice indeed) but most people know that @W1zzard is a great professional and blindly look at the average results that, imho, at the moment have some big outliers that heavily skew the results.

So, my suggestions are towards the best objectivity possible. Statistics is a great tool if practised correctly in order to get conclusions out of test containing many results.

 

[Ruru]

where is everyone thinking of putting the card straight under water? I wonder how the card reacts?

That was the first thing which came to my mind when I knew that the cooler isn't good. Well, that comes always to my mind as I run a CPU/GPU custom loop.

 

[Athlonite]

That was the first thing which came to my mind when I knew that the cooler isn't good. Well, that comes always to my mind as I run a CPU/GPU custom loop.

It'd probably be a great card in a custom WC loop mainly because the stock cooler on it is shite admittedly better than the single squirrel cage fan coolers but still shite at the end of the day none the less it will be interesting to see what the likes of EK do for full coverage blocks

 

[Voluman]

Thanks for the review again, great as always

Are there any chance to test with more modern apis too? In many title you use the dx 11 renderpath, which makes sense seeing the market (and your previous datapoints). Just makes me wonder how dx12 and vulkan are standing nowadays.
Can you check with Forza Horizon 4 at some point too? It has a reworked order processor/buffer which is supported by the VII, if i know correctly.