How's your Navi (5700XT/5700) Treating You?

[xkm1948]

Some people buy AMD because they have open source drivers for Linux. To each their own.

I had really bad luck with AMD’s linux support when i was using FuryX. The OpenCL worked like shit and there no where to get help.

 

[Rahnak]

It's only 1-2% behind 5700XT + it has 4-5% additional OC headroom

Well, you're comparing an AIB card to a stock card on a blower, which is a little unfair. Would be nicer if you compared two similar AIB cards. TPU recently reviewed the MSI 5700 XT Gaming X and it improves relative performance at 4k by about 4% over stock (that would be a fairer comparison, imo).

 

[notb]

Some people buy AMD because they have open source drivers for Linux. To each their own.

You don't need to own an AMD card to install open source Radeon drivers. :-D

 

[bug]

I had really bad luck with AMD’s linux support when i was using FuryX. The OpenCL worked like shit and there no where to get help.

I didn't say they were working fine (though they're supposedly better these days), just that what's important for one person when choosing a video card may nit be as important to the next.

 

[Kanan]

Of course, from this very site even:

It's only 1-2% behind 5700XT + it has 4-5% additional OC headroom

Not in 1440p. Basically using irrelevant data to try to prove a point. The 5700 XT is clearly faster than the 2060 Super, AIB vs AIB. Here you also make another mistake, as in, comparing a AIB vs a ref model. A AIB model 5700 XT will have up to 5% extra performance.

Edit: I've also checked the prices. The models comparable to Red Devil, aren't cheaper, despite the lower performance.

 

[HenrySomeone]

Even the reference / founders edition 2060 super overclocks really nicely, while AIB 5700XTs are pretty much on the limit already - you do the math:

 

[Rahnak]

@HenrySomeone The Asus isn't that great of a card, so it's one of the worst examples. Granted, overclock isn't as good as nvidia, but it's more than 1 fps.

Also, it's clear by looking at the numbers that Unigine Heaven runs much better on nvidia than on AMD, hence also boosting the fps gained.

 

[Taraquin]

Also: Undervolting the Navi-cards yields better results than OC. I gained a few percent performance by just lowering stock voltage from 1200mv to 1100mv. Overclocking is blocked due to powerlimits on Navi, but undervolting fixes that.

 

[MDBumblebee]

It's bad. Every driver after 19.7.5 is a fountainhead of BSODs. Crashes in the middle of work. Something as normal as switching tabs in Firefox or loading YouTube can cause a nasty BSOD. Most monitoring tools that poll I2c cause memory or IRQ related BSODs. Also AMD did a nasty bait-and-switch between its initial drivers and 19.7.4, by raising minimum fan speeds on the reference card.

When it works, it's really fast. But it's not safe for PCs that put food on the table. Take it from someone with two such PCs.

19.9.2 fixed 90% of my BSODs. Very optimistic now.

View attachment 131559

You should really dust that case! Maybe install some filters too.

 

[cucker tarlson]

@HenrySomeone The Asus isn't that great of a card, so it's one of the worst examples. Granted, overclock isn't as good as nvidia, but it's more than 1 fps.
View attachment 132127
Also, it's clear by looking at the numbers that Unigine Heaven runs much better on nvidia than on AMD, hence also boosting the fps gained.

the asus card is also the fastest out of the box.
and the oc test is supposed to measure % gain on each card,do you mean that 10% on nvidia would be faster than 10% on amd ?

 

[Zubasa]

Even the reference / founders edition 2060 super overclocks really nicely, while AIB 5700XTs are pretty much on the limit already - you do the math:

@HenrySomeone The Asus isn't that great of a card, so it's one of the worst examples. Granted, overclock isn't as good as nvidia, but it's more than 1 fps.
View attachment 132127
Also, it's clear by looking at the numbers that Unigine Heaven runs much better on nvidia than on AMD, hence also boosting the fps gained.

MSI 5700XT vs Reference = about 7.5% OC
2060 Super FE OC vs stock = about 8.7% OC
So if you mean by the limit, then the AIB 2060 Super barely OC better. You are impling that AIB 2060 Super OC better, they do not.
In fact the good old 2060 FE clocked better than most of them thanks to Silicon lottery.

Also I am not sure why you keep thinking 2060S OC will make it better than 5700XT in performance.
Fact is with that 8.7% OC it will barely match the MSI 5700XT out of the box.

 

[Rahnak]

the asus card is also the fastest out of the box.

But he was talking about overclocking, which the Asus isn't the best at. And he was talking about a MSI Gaming card, so I showed the equivalent SKU.

and the oc test is supposed to measure % gain on each card,do you mean that 10% on nvidia would be faster than 10% on amd ?

That's a gross oversimplification of my point. What I meant was that the OC test is based on a synthetic benchmark that heavily favors nvidia. So just because a nvidia card gets a higher % value on that particular benchmark VS amd, that does not guarantee that same % in real world usage.

 

[notb]

But he was talking about overclocking, which the Asus isn't the best at. And he was talking about a MSI Gaming card, so I showed the equivalent SKU.

Why would anyone care, honestly?

All Navi cards are clocked near the limit and they all overclock to more or less the same level.
1) If one is going to OC, he can buy whichever card he prefers for the non-performance properties (noise, looks, size).
2) If one is not going to OC, then the ASUS will give him that extra 1-2% out of the box.

That said, this factory OC makes ASUS loud. With quiet BIOS (which should really be called "default BIOS") it becomes very similar to the MSI Gaming.

That's a gross oversimplification of my point. What I meant was that the OC test is based on a synthetic benchmark that heavily favors nvidia. So just because a nvidia card gets a higher % value on that particular benchmark VS amd, that does not guarantee that same % in real world usage.

The obvious, general observation is that every benchmark - synthetic or real-life - will prefer one architecture over another. No way around it.

The proper question is: why so many benchmarks prefer Nvidia? Why so many games used in reviews prefer Nvidia?
And answer is both simple and hard to digest for some: because Nvidia products are better optimized for popular existing loads. And the reason is simple as well: because Nvidia spends more on development.

 

[cucker tarlson]

That's a gross oversimplification of my point. What I meant was that the OC test is based on a synthetic benchmark that heavily favors nvidia. So just because a nvidia card gets a higher % value on that particular benchmark VS amd, that does not guarantee that same % in real world usage.

no because you're comparing amd stock to amd oc,not amd to nvidia

 

[ErnstInEarnest]

Why would anyone care, honestly?

All Navi cards are clocked near the limit and they all overclock to more or less the same level.

Does Nvidia OC better than AMD?

 

[notb]

Does Nvidia OC better than AMD?

It could seem like that, but it's not because Nvidia has a much better tech or production process.
It's because they tune their products very differently.

Nvidia aims at efficiency, low noise, stability and plug&play experience. GPUs are clocked very safely by default.
AMD aims at performance/price. They apply aggressive clocks.

The big picture is more or less like this:
1) If you're the kind of user who uses PC parts as they come from the factory - Nvidia cards will be easier to live with (noise, power consumption), while AMD will have better performance/price ratio.
So it's really down to how you, personally, define graphics card's "value".
2) If you're fine with OC, undervolting, modding etc - products from both brands can be tuned to very similar characteristics.
Which doesn't mean they can be made the same: there are many differences in features, driver quality and support.

 

[ErnstInEarnest]

1. I'm noob and learning.
2.I owned a 2080Ti hybrid on C7H/3800x, i guess it's overkill for my 1440p120Hz. gaming . Will give to my son.
3. Thinking of building one 'down to earth' 1440p gaming rig by myself (i'll keep the monitor X34/Seasonic Prime 650W Plat/Steelseries Artic 5).
4. Will go for 3600 , undecided on Red devil 5700XT or EVGA 2070s Black , 'value for money' is important criteria.
~~hence the question

 

[Rahnak]

Why would anyone care, honestly?

I personally don’t. I was just saying there are better overclockers than the Asus card.

no because you're comparing amd stock to amd oc,not amd to nvidia

What? My whole point was comparing the oc of both MSI Gaming versions of the 5700XT and 2060 Super. Either you didn’t get my post or I’m not getting what you’re trying to say.

 

[jormungand]

I think Navi is a great architecture and when it matures along with the drivers it will be extremely good

i think the same, i really was expecting to buy the Sapphire Nitro RX 5700 XT ( what a sexy beauty) but reading and reading more posts from 5700 xt owners including yourself
i decided to go nvidia again.
i remember how much love and praise i gave to these cards


But me personally i dont pay for problems.
really hope AMD mature that tech along with drivers.

 

[notb]

I personally don’t. I was just saying there are better overclockers than the Asus card.

For me it's really the same phenomenon. Why would you want to own a "better overclocker"? Why not "a better performer"?
Is is important to you that you get more OC gain?
Because, as I said earlier, these cards - after OC - will be almost indistinguishable in game fps.

 

[cucker tarlson]

What? My whole point was comparing the oc of both MSI Gaming versions of the 5700XT and 2060 Super. Either you didn’t get my post or I’m not getting what you’re trying to say.

your point was that somehow this particular benchmark shows greater gains from oc on nvidia cards than amd cards.
please provide a test of your own/source that would prove that in an amd favorable game navi oc headroom is bigger than in this benchmark,beyond margin of error of course.Cause in every damn review I've seen it's not.

I personally don’t. I was just saying there are better overclockers than the Asus card.

no,those are cards that clock lower out of the box.

i think the same, i really was expecting to buy the Sapphire Nitro RX 5700 XT ( what a sexy beauty) but reading and reading more posts from 5700 xt owners including yourself
i decided to go nvidia again.
i remember how much love and praise i gave to these cards
View attachment 132398View attachment 132399

But me personally i dont pay for problems.
really hope AMD mature that tech along with drivers.

I was gonna go with a brand new ryzen 3000 build but decided that money is much better put aside for ps5 rdna 2 next year.rdna 1 is already very good in terms of performance,though unusable for a considerable number of users.rdna 2 in ps5 is gonna be fantastic,I hope they pair it with ryzen 7nm not 12nm under a misleading 3000 name like the apus.

 

[Rahnak]

For me it's really the same phenomenon. Why would you want to own a "better overclocker"? Why not "a better performer"?
Is is important to you that you get more OC gain?
Because, as I said earlier, these cards - after OC - will be almost indistinguishable in game fps.

I guess I wasn't clear the first time? I. Don't. Care. I was merely contradicting someone else's point. And again, personally, I don't overclock my cards. I suspect not many people do these days, with all the auto overclocks and factory overclocks. So yeah, man, I agree with you. Buy the best you can afford right out of the box.

your point was that somehow this particular benchmark shows greater gains from oc on nvidia cards than amd cards.
please provide a test of your own/source that would prove that in an amd favorable game navi oc headroom is bigger than in this benchmark,beyond margin of error of course.Cause in every damn review I've seen it's not.

Alrighty. Quoting from TPU's reviews (overclocking section, using Unigine Heaven):
2060 Super FE: "Actual 3D performance gained from overclocking is 8.7%." (The one used in Henry's post)
MSI 2060 Gaming X: "Actual 3D performance gained from overclocking is 7.4%."
Asus 5700XT Strix OC: "Actual 3D performance gained from overclocking is 0.7%." (The one used in Henry's post)
MSI 5700XT Gaming X: "Actual 3D performance gained from overclocking is 3.3%."
My point was that these percentages are from a benchmark that favors nvidia hardware and thus not indicative or real world results.
Guru3d has an overclocking section on their reviews were they retest 4 games (Witcher 3, Strange Brigade, Deux Ex: Mankind and Shadow of the Tomb Raider) using overclocked cards. The average performance increase for the MSI 2060 Super Gaming X over the reference card was 108% (review link here).
The average performance increase for the MSI 5700XT Gaming X over the reference card was 107% (review link here).
So as you can see, their overclocking gains are much closer than what Unigine Heaven would have you believe.

no,those are cards that clock lower out of the box.

How is that not a better overclock on the MSI?

 

[Kanan]

Is this not the wrong thread for having a Nvidia vs AMD debate? The 5700 XT is a great card, new drivers also fix a lot of issues. The performance will also get up as they mature more, the 5700 XT, from a purely technical standpoint, has more compute units and should therefore be better than the 2600S in the long run. There's only one possibility to get a better card, by spending those 50-80 bucks extra to get a 2700 Super, but then you will only have the cheapest ones, aka Galax, but if you dont care about how the card looks it'll be good enough. But again, depending on the model compared to, up to 80 bucks more for a mesely 8% more performance. Not worth it. Honestly, the worthiest cards right now, are probably the 5700 XT Pulse aib model, which performs well and doesn't cost a lot more than the ref, and the same version as the non XT model. Also the Gigabyte gaming seems to be nicely priced, at least in Europe.

 

[cucker tarlson]

Is this not the wrong thread for having a Nvidia vs AMD debate? The 5700 XT is a great card, new drivers also fix a lot of issues. The performance will also get up as they mature more, the 5700 XT, from a purely technical standpoint, has more compute units and should therefore be better than the 2600S in the long run. There's only one possibility to get a better card, by spending those 50-80 bucks extra to get a 2700 Super, but then you will only have the cheapest ones, aka Galax, but if you dont care about how the card looks it'll be good enough. But again, depending on the model compared to, up to 80 bucks more for a mesely 8% more performance. Not worth it. Honestly, the worthiest cards right now, are probably the 5700 XT Pulse aib model, which performs well and doesn't cost a lot more than the ref, and the same version as the non XT model. Also the Gigabyte gaming seems to be nicely priced, at least in Europe.

neither is it about 5700xt pricing or performance.

 

[Cheeseball]

• Setting a monitor to 75 Hz or 144 Hz causes the memory to hit max clocks (although the GPU is fine) during idle. 60 Hz and 120 Hz works fine. I experienced the same thing with a HD 7870 XT back in 2013 when I still had a 72 Hz monitor:

• Sometimes attempting to open the RADEON SETTINGS will result in a blurry box, but the buttons and toggles are there, you just can't see them. Killing the Radeon Settings: Overlay and Host Service tasks and accessing the menu again will fix it:

• Looks like Enhanced Sync is broken with the 5700 XT. A lot of my games (particularly PUBG) crash with it turned on. Just having FreeSync is on is fine though.

All these issues are experienced with the latest Adrenalin drivers: 19.9.2 WHQL (sept23)

They have been reported to the Radeon Community in AMD.

This MSI EVOKE OC Radeon RX 5700 XT is a damn beast, and I'm extremely satisfied with it's performance, especially compared to my older ZOTAC GTX 1070 non-AMP. I have fully accepted that the AMD drivers for the RX 5700 series are currently trash, but I know they will continue to improve upon them as long as we keep reporting and keeping these issues in view.