AMD Radeon RX 5700 XT and RX 5700 Review: Navi Graphics Tested

[NightAntilli]

Hey Steve.
I have a request. There's this rumor flying around that the 5700 series cards gain ~5% performance when paired with a 3900X compared to the 9900K, while the new Super cards lose some performance instead. Care to test that?

 

[Markoni35]

I just bought Radeon VII 3days ago....for 800 USD (I lieve in Japan) and 5700XT is selling for 500USD...
and I don't even play 4K!! I only use 1440p monitor, so basically I paid 300USD more for tip of a FPS and 3D designs that I will never do...

Why didn't I wait for couple of more days...I honestly though 5700XT wouldn't be this close. I mean good for AMD! yay...

Edit: Forgot to mention, good review, thank you for your hard work!

Don't eat yourself. If you waited for a few weeks, there would be some problem with 5700 and it would be delayed. Then you would buy your Radeon VII and immediately next day they would release 5700. That's how it works.

I know it from my own experience. If it's gonna raining on a particular day, it's gonna start raining the moment I exit the house. Doesn't matter if I hurry up, or I wait for 2 hours more, whenever I exit, THEN it will start raining. Same with stock or crypto trading. It's as if someone is waiting in ambush for me to buy or sell a particular item. Fun times...

 

[Khanonate]

I just bought Radeon VII 3days ago....for 800 USD (I lieve in Japan) and 5700XT is selling for 500USD...
and I don't even play 4K!! I only use 1440p monitor, so basically I paid 300USD more for tip of a FPS and 3D designs that I will never do...

Why didn't I wait for couple of more days...I honestly though 5700XT wouldn't be this close. I mean good for AMD! yay...

Edit: Forgot to mention, good review, thank you for your hard work!

Don't eat yourself. If you waited for a few weeks, there would be some problem with 5700 and it would be delayed. Then you would buy your Radeon VII and immediately next day they would release 5700. That's how it works.

I know it from my own experience. If it's gonna raining on a particular day, it's gonna start raining the moment I exit the house. Doesn't matter if I hurry up, or I wait for 2 hours more, whenever I exit, THEN it will start raining. Same with stock or crypto trading. It's as if someone is waiting in ambush for me to buy or sell a particular item. Fun times...

Hmmm, it looks like your Murphy's (Murphy's Law) favorite too.

 

[ATInsider]

I wonder if these cards perform any differently when matched with a x570 board, new 3000 CPU's, and faster RAM? I know Ryzen likes fast RAM so much.

ZEN and ZEN+ require faster Ram because IF is tied to the IMC. ZEN2 changes this a little, so that RAM can increase in speed significantly.

 

[hahahanoobs]

I know Ryzen likes fast RAM so much.

That is incorrect according to TPU:
https://www.techpowerup.com/review/amd-zen-2-memory-performance-scaling-benchmark/4.html

 

[Shadowboxer]

Must say I’m not entirely sold. Of all the new cards released right now the 2070 super looks like the Best buy. I don’t like the temperatures and the cooler on these new Navi cards and apparently undervolting is once again yielding some increases, why can’t this be delivered at stock? Im sure the aftermarket ones will be better. But for me, I’d rather pay a little more for the 2070 super as it’s a faster card and I think it will last you a lot longer going forward. Not to mention Gsync & freesync compatibility, oh and RTX features for the tiny handful of games that support it. Still, way better than the loud heap of garbage that was the Radeon VII.

I do think those on a tighter budget are best on a Navi 5700 than any of the other new cards from AMD/Nvidia - an aftermarket one that is. But if you can spend a bit more for the 2070 Super then I’d say get that.

As for us 4K gamers, well, there isn’t much here still. Getting a bit frustrated actually as I’d love to see 4K gaming become more mainstream.

 

[OptimumSlinky]

Must say I’m not entirely sold. Of all the new cards released right now the 2070 super looks like the Best buy.

Wut? The 5700XT delivers 1080 Ti performance for $400; the 2070 Super delivers on average 2% better performance for $500. How on earth does that make the 2070 Super the best buy to you?

I don’t like the temperatures and the cooler on these new Navi cards and apparently undervolting is once again yielding some increases, why can’t this be delivered at stock? Im sure the aftermarket ones will be better.

So wait for AIB coolers from MSI, XFX, and Sapphire. Not rocket science.

But for me, I’d rather pay a little more for the 2070 super as it’s a faster card and I think it will last you a lot longer going forward. Not to mention Gsync & freesync compatibility, oh and RTX features for the tiny handful of games that support it.

2% faster when averaged. 2% for $100 more. nVidia throwing in the towel on G-Sync should tell you everything you need to know about that feature, and by the time ray-tracing is actually a widely adopted thing in gaming, we'll be on Turing Gen-2 which will be likely significantly more efficient with RTX.

 

[m3tavision]

Well.… after the dust settled it looks like AMD's 5700 Series is the clear mid-range choice now...!

And reading nearly 15+ reviews it seems Nvidia's SUPER had no impact on Navi, and if anything showed how strong a $349 chip can be at 1440p. Little Navi kicked the shiz of big turing.!


Look above, or elsewhere on ALL of the reviews, where the $349 little Navi was close to a $499 Turing.... those cards have a $150 difference in price! As such, SUPER is overpriced and RTX on a whole is a flop. The rtx2080 is most likely my last Nvidia product, because I can already see which way Games are going.

RDNA is the future of gaming. In one years time, many of use will be using it on our smartphones. It's in the Xbox & PlayStation and going to dominate the cloud gaming. Navi's new architecture is much different from Vegas. It is more geared for Games, not scientific.

We play games...

 

[treetops]

Idk why this article doesn't compare cards of the same price towards the end. Still updating it maybe, here is a quick summary.
1440p:
2060 $350 < 5700 $350 AMD 11% Higher FPS
2060 Super $400 < 5700 XT $400 AMD 9.9% Higher FPS

"Across our benchmark suite, Radeon RX 5700 XT averages 9.9%-higher frame rates than the GeForce RTX 2060 Super at 2560 x 1440. Radeon RX 5700 averages 11%-higher frame rates than the GeForce RTX 2060 at the same resolution. The GeForce RTX 2070 Super does serve up average frame rates 6.9% higher than Radeon RX 5700 XT, but it costs 25% more." -Toms Hardware link below

https://www.tomshardware.com/reviews/amd-radeon-rx_5700-rx_5700_xt,6216-2.html

 

[treetops]

Now this is what I´m talking about. To me this launch is miles ahead of Ryzen 3000. These GPUs are really good for the price with the 5700XT even doing almost as good as 1080ti/Vega VII on a lot of games while using way less power. A 5700XT for 400€ is a good buy imo and completly makes the Nvidia competing offers irrelevant unless you want the Nvenc codec (I see few people metioning this, wich imo is one of the best selling points on RTX series).

I just think 5700 should be priced a bit lower, because it is too close to the 5700XT to justify the 50€/70€ saving.

Great GPUs, AMD finally nailed it on the power consumption too.

I cant believe how close this is to my Radeon VII..… if I wasn't a 4K gamer I would be pissed but 8GB of memory isn't enough for 4K and my card is running at 1000mV 1850Mhz Core 1100 Memory and is whisper quiet so I'm not that bothered but still I paid £680 for is in March, that RX 5700 XT is almost half price...….

Idk if you have Ebay were you're at but you can get 650$ for it used if you want to sell it. Which is about what they cost new here 670$?. Miners seem to be buying them up. It has compute unit value. Then spend 50$ more and grab a 2080 super ^^. 2080 Super comes out on July 23rd 700$. No time like the present bud. I'm all for these new cards but if I was in your situation at that price point... .

Right now you can get a 2080 for 550$+ or so on ebay used. Last month they were 670$+.

EDIT these are the BUY IT NOW prices I've been observing. Not bidding.

 

[parowOOz]

Well now, this is more like it. First we get the SUPER update that corrects one of the biggest flaws of the 2xxx series - huge jump in price/perf - and makes RVII, and OG 2060/2070/2080 irrelevant. Now we have these little gems to make things more interesting.

From a pure value perspective there's no denying that Navi is a top contender. And holly smokes what is going on in Forza 4 ? Navi is flying there. Can anyone more knowledgeable on the subject explain is it a case where Navi is flying or is this a case of Turing underperforming ? Perhaps a mixture of both ? What's so special about Forza 4 that makes AMD cards shine oh so bright ?

Every rose has thorns though and these cards are no exception. Say what you will about the look of NVIDIAs' stock/FE cards but those are some decent coolers. Same cannot be said about their AMD counterparts. 84C on such a small chip, made on a brand new process node doesn't look right. These cards clearly need AIB treatment.

Power consumption is also a bit worrying. Again - 7nm, 250mm2 and yet it pulls nearly as much as RTX 2070 SUPER ? Is this a remnant of their GCN heritage or does this point to TSMCs' 7nm process not being well refined ? Is this one of the reasons NVIDIA went with Samsung ? Can someone shed some light on this subject ?

Last but not least - performance. On average, it's awesome, especially given the price. There is a caveat however. Averages can be misleading and 5700XT vs 2070SUPER highlights it perfectly. 2% difference on average seems like nothing, and it is. But when we look at games individualy we get a different picture. Performance swings between both offerings are quite substantial. It will boil down to which games you're playing/planning on buying. I'd love to see a more comprehensive (30+ titles) test.

All in all, I consider this a very solid launch for AMD. Performance is good, price is good and AIB cards should deal with high temepratures and give extra fps here and there. Could we please get some high-end Navis, AMD ? That 2080Ti price tag is a travesty and it needs to go.

Let us also comiserate RVII buyers I guess they now know how early adopters of Titans usualy feel when a xx80Ti is introduced into the market Not dissing anyone just to be clear. I happen to think that, while situations like these may seem rough to some, progress is something to be positive about and hey, RVIIs are still solid performers.

 

[jgraham11]

Testing methodology should change.

Stop trying achieve maximum frames per second when over 60FPS doesn't matter.

These chips were built for PS5/Xbox where there is a 60FPS cap. Show the cards caped at 60FPS and the new metric is consistency and efficiency.

My bet is that the AMD cards perform way more efficiently than Nvidia's at lower frequencies.

A way to test this:

Method 1:
As described above with FPS being limited

Method 2:
Lock the GPU at lower frequencies, do benchmarks, measure performance and power use, plot on a graph frequency versus performance. It would be a pretty thorough look at each architectures frequency scaling.

 

[neeyik]

Okay, imagine a scenario where 4 graphics cards (A, B, C, D) are being tested. Let's say your method 1 is used, and here are the results:

Card A - price $1000
Max FPS - 60
Min 1% - 60

Card B - price $800
Max FPS - 60
Min 1% - 60

Card C - price $600
Max FPS - 60
Min 1% - 55

Card D - price $400
Max FPS - 58
Min 1% - 42

What possible conclusion could one draw from this? That A is a pointless purchase, as it is seemingly no better than B but costs $400 more? That C is the best value? D? How could one determine from such information how the same cards will perform in games 3 years into the future? One must assume that A is in some way more powerful than B, but there is seemingly no way to show it, given that one is artificially capping the output.

The solution is obvious: remove the limitation and see how the products compare in a level playing ground, where they can display the maximum potential in that particular test. In other words, much like how nearly all benchmarking - be it in academic examination or sporting events - is done.

Method 2, on the other hand, without a 60 fps cap, would be a useful examination; however, time constraints always limit what can be achieved in any review. There is only so much time a tester has with a piece of software before the review must go out live. Naturally, frequency scaling could always be examined in a follow up article though.