AMD Radeon RX 5700 XT

[Durvelle27]

It's excellent and works in all games. Have you read the review or the frontpage news?

My bad @W1zzard

I meant multi monitor

 

[roccale]

Really a bad advertisement for 7nm. They use 28nm voltages on a 7nm to recover clock.
A scandal.

 

[Raptor_i]

Good performance and good price but those temps are confusing considering other sites have 76c at load and 82c at load also and 84c at Furmark !!

@ W1zzard I have a question for reference did you test the card (especially thermals) after you replaced the Hitachi thermal pad ?

 

[efikkan]

Great performance across the board, seems like TPU got some bad samples that were probably damaged during shipping and are running absurdly hot. ALL other review sites reported much lower temperatures, in line with what we are used to and in line with Nvidia's offerings.

All in all the RX 5700 beats the RTX 2060 by 5%, while being the same price and has lower power consumption.

It all depends on which temperature measurement they rely on. GamersNexsus achieved similar high temperatures.

RX 5700 XT is much hotter than Nvidia's offerings, and has higher power consumption.

RX 5700XT beats the RTX 2060 super by about 5% at the same $400 price point

At the $400 price point RX 5700 XT marginally outperforms RTX 2060 Super, but would you seriously buy a hot and noisy card?

 

[W1zzard]

Good performance and good price but those temps are confusing considering other sites have 76c at load and 82c at load also and 84c at Furmark !!

@ W1zzard I have a question for reference did you test the card (especially thermals) after you replaced the Hitachi thermal pad ?

Of course not

 

[A.Stables]

yep as usual smth off with a radeon card here saw it with super review and numbers (radeon 7/ Vega numbers) and as 5700 hits ... as before (vega polaris) Noisy hot AMD from TPU .. is mention again. Steve @ hware ubxoed killed his CPU does that mean all 3900x are the same easy to kill .. ffs TPU sort it out.

*Edit a bit of reference

AMD Radeon RX 5700 XT and RX 5700

The world's first Navi GPUs analysed, benchmarked and rated.

www.hexus.net

AMD Radeon RX 5700 and 5700 XT review

In this review, we look at the two new graphics cards released by AMD, the Radeon 5700 and 5700 XT. Both cards are based on the new NAVI GPU, fabricated at a 7nm node and capable of battling with NVI... Graphics Card Acoustic Levels

www.guru3d.com

 

[64K]

That's what the speculation was. That the RX 5700 XT would come in somewhere around a RTX 2070. Looking at the price chart it seems it will cost $80 less than the 2070.

 

[natr0n]

AMD cards always need mature driver support otherwise nightmares are bound to happen.

 

[John Naylor]

I do wish the reviews would expand on the following conclusion.

"....beats RTX 2070 ...." yeah, it does outta the box, but not when both cards are overclocked.

From TPU graphs ....

@ 1080p ....

5700XT gets a 100% x 118.1 OCd / 111.8 Stock = 105.63
RTX 2700 gets a 99% x 138.7 / 128.3 OCd / 111.8 Stock = 107.02 ... 1.5 % faster the 5700XT

@ 1440p ....

5700XT gets a 100% x 118.1 / 111.8 = 105.63
RTX 2700 gets a 99% x 138.7 / 128.3 = 105.94 ... 0.2 % faster the 5700XT

@ 2160p ....

5700XT gets a 100% x 118.1 / 111.8 = 105.63
RTX 2700 gets a 100% x 138.7 / 128.3 = 108.11 ... 2.3 % faster the 5700XT

I'm not saying get rid of the stock numbers and use the OC instead. No review would be complete without the performance numbers out of the box. But, since the audience here overwhelmingly OCs their cards, the relativew performance with OC is actually what card selection will be based upon

As far as the 5% lead I saw several times in this thread, the TPU charts show 0%, 1% and 2%. , In addition , a) that's out of the box, b) cards OC'd where ?

1080p = 100 / 99 - h ttps://tpucdn.com/review/amd-radeon-rx-5700-xt/images/relative-performance_1920-1080.png
1440p = 100/ 98 - h ttps://tpucdn.com/review/amd-radeon-rx-5700-xt/images/relative-performance_2560-1440.png
2160p = 100 / 100 h ttps://tpucdn.com/review/amd-radeon-rx-5700-xt/images/relative-performance_3840-2160.png

... and c) the 5700 XT loses at all resolutions with both cards OC'd

Still, just reading this for curiosity's sake ... have to wait for the AIB cards on both sides in August to have their reviews out. It's not like non AIB cards have are something I'd ever recommend. But the temps and sound are untenable options at this point.... If AIB versions of the 5700XT can significantly improve on this and keep the performance deficit when compared to nvidia small when all cards are over clocked, they could cut into nvidias dominance in this segment if they they will remain $50 - $75 cheaper.

A for the sound thing... folks are confusing things ... while +3dB is effectively a doubling of sound pressure level + 10 dbA is a twofold in creases in perceived sound

+ 3 dbA = doubling of power level
+6 dbA = doubling of amplitude
+10 dbA = doubling of loudness

As for pricing ... speculation at this point is just that.

a) Is anyone actually planning to buy a "reference card ?
b) The AIB cards will have better numbers so why talk proce of cards you won't buy ?
c) The "I need to be the 1st one on my block to have the new shiny thing " crowd will keep prices high fpr at least 6 weeks and a lot of this "I only buy (red of green) team stuff".

Here, we won't be recommending any new cards till mid to late August at best.

 

[Markosz]

Seems like most of the negatives comes from this horrible reference card design.
The chip itself seems pretty good, but we really need some third party card designs to see what is possible.

 

[efikkan]

Partner cards can deal with the noise and temperature problems, but everything else is going to stay pretty much the same.

 

[Xaled]

Partner cards can deal with the noise and temperature problems, but everything else is going to stay pretty much the same.

5700s performance may improve with new drivers because it's new gpu, while Nvidias cards are not.

 

[efikkan]

5700s performance may improve with new drivers because it's new gpu, while Nvidias cards are not.

Ahh, there it is, I was waiting for someone to mention the mythical missing AMD driver.
We hear this excuse with every new generation from AMD; don't judge it yet, it will improve over time.
The truth is that AMD's drivers improve no more than Nvidia's over time, and the expectations shouldn't be too great since Navi isn't a huge architectural change.

 

[EarthDog]

We hear this excuse with every new generation

FTFY. People on both sides of the fence say this... though one side of the fence screams louder because their marketing diety likes 'fine wine'.

From an article covering it...

Think of it this way, at 60 FPS a 5% advantage is only maybe 3 FPS? Free performance is great, we’ll take all we can get, but anything under 10% is impossible to notice in the real-world while gaming. This goes to show a couple of important facts. Drivers are not going to be the miracle answer to a video card’s performance over time.

This is good because on the one hand you are getting most of the potential out of the video card from the start. On the other hand, if you aren’t happy with that performance, then there isn’t much hope (at least as far as our testing has proven) that the performance profile of the video card will vastly improve over time.

Is this Fine Wine? That of course is very subjective and up to your interpretation.

Don't hold your breath for anything noticable outside of a title that needs to be 'fixed' which is not a result of maturation.

 

[efikkan]

FTFY. People on both sides of the fence say this... though one sode.of the fence screams louder because their marketing diety likes 'fine wine'.

I was actually serious.
On the driver side, Navi uses the same old GCN driver with some minor tweaks, so the core driver should be very "mature". In fact, probably the most "mature" they've ever released. There isn't anything in the driver to optimize for Navi as it uses the same ISA. There might be improvements in the driver over time, but this will be general improvements unrelated to Navi.

 

[Animalpak]

Thanks wizzard and TPU staff for the review...

AMD nothing new mate... You still behind the green team.

 

[EarthDog]

I was actually serious.

And so was I.

 

[tajoh111]

Giving an overpriced card like 2060 Super gold award while again asking for a lower price for a card that kicks it's ass.

İmo AMD should once again stop sending review samples to TPU for while.

Ridiculous accusation there. It's a changed market place since the release of super which is why it is better to release earlier than the competitor(it's also why AMD dropped the price of these cards).

Additionally being a pain in the ass to benchmark because of driver issues and dropping a new driver just 30 hours prior to review embargo isn't doing reviewers any favors. It is likely to effect your score.

Gamers nexus also encountered issues when it came to drivers, noise and heat.

If AMD stopped sending reviews samples to techpowerup because they don't have the royal carpet thrown out for them it is there loss. TPU is one of the most viewed review websites easily.

Reviewers are already accommodating them enough by not being given permanent review samples they have to share, being given last minute drivers(Radeon VII launch and this launch) and putting up with BS like that the Vega 56 officially dropped to 250 when the GTX 1660 ti launch.

On top of this, I see unreasonable requests, like testing an AMD platform on top of a Intel platform for videocard launches, removing highly popular game titles that AMD cards don't do well in. Don't test too many cards using unreal engine. Undervolting AMD cards to find their optimum power efficiency, rereview cards cards 2 months after launch so AMD cards are given more time to mature. Literally the royal carpet treatment when they are treating reviewers like dirt. The last minute drivers before launch embargoes are a spit in the face of reviewers because it forces reviewers to do twice as much work.

So if AMD stops sending samples to techpowerup because the royal carpet is not thrown around them, it is their loss because reviewers are accommodating enough to them and a loss of review exposure on one of the biggest tech websites out there. And when a big website gets excluded because of AMD feels they are not get a fair shake(Fury nano), that says alot more about them if one of those websites is techpowerup.

 

[Naito]

I'm wondering how much of this is caused be the driver? If the driver has much to do with the power states of the card and it is not working correctly, it may cause a 'snowball' effect. Inefficient management could mean higher voltages, high voltages means more power and heat, which in turn, generally causes further increases in voltage as the card tries to stabilize. The overall power draw of the looks decent to me, aside from multi-monitor is tad high, but I again wonder if drivers could temper that more towards the Vega levels as opposed to Polaris levels? Worse case is that the quality of the first batch of 7nm dies are good, but not great. It's probably more likely that AMDs cooler, along with the early drivers, are just not up to scratch.

As I said earlier, I think these cards are very competitive and will only improve further when the AIB cards hit the market and the drivers mature (and that's coming from a long-term Nvidia fan - haven't had a red team card since my X1950 Pro AGP).

 

[64K]

Undervolting AMD cards to find their optimum power efficiency, rereview cards cards 2 months after launch so AMD cards are given more time to mature.

I've seen this before but the thing I wonder about is why AMD doesn't do this already if it has a good impact on performance. Maybe they don't it's a good idea to undervolt their cards for reasons we don't know about also it's very time consuming to review cards. Maybe there are not enough hours to do this and keep up with new card reviews.

Don't test too many cards using unreal engine.

Unreal is the most popular game engine with Developers.

List of Unreal Engine games - Wikipedia

en.wikipedia.org

 

[Minus Infinity]

Given how poor the drivers currently are I'd say the 5700XT is a viable option as long as the third party cards with better cooling solutions don't end up at 2070 super prices. If the drivers can improve things and you can get the 5700XT custom board for $449 with only a few 5% deficit to 2070 Super it looks decent value. I'm in no rush to upgrade my 1070 so hopefully over the next month or so we'll have a better picture of the 5700 twins. If performance doesn't improve much I'll definitely pay the 2070 super premium.

 

[-1nf1n1ty-]

even with everything in the review, I want this over my 2060. I wish the super cards weren't released as it felt like a big slap in the face to me, will probably sell 2060 for this once I see a sapphire cooler come into play or xfx

 

[A.Stables]

Navi is looking really good imo, New arch, lets see how software adapts and of course OS/driver/API optimisations.
also

really cool.

This release reminds me of Zen1 launch a bit rough around edges but you can see the perf is there, and I suppose the 3870 in some ways.

 

[wolf]

AMD has come a seriously long way with this card. It's impressive on a few levels but unfortunately a disappointment on others. Here's hoping in a few months AIB partner cards with excellent coolers and slightly more mature drivers bring it up to it's full potential.

Yet to find anything compelling enough to draw me into the market to replace my undervolted and overclocked GTX1080. I think the next 12-18 months the GPU market is going to be a ripper though.

 

[Eric3988]

Finally an impressive showing by AMD. If it weren't for the heat and sound issues I'd say they hit a home run. Hopefully the aib cards will take care of that for a reasonable price.