AMD Radeon RX 5700 XT and RX 5700 Review Roundup

Being that Navi has no 64 cu limit, i wonder how it would sell and perform wth 64/80/more CUs.
 
Uvaman2 said:
Being that Navi has no 64 cu limit, i wonder how it would sell and perform wth 64/80/more CUs.
Click to expand...

I think a 64 CU NAVI would run with a 2080ti, trading blows or out right wins (or slight losses) depending on the gsme(s) in question.


The fact that AMD is allowing TSMC to maximize their yields, while releasing a mid-range smallish die chip that is as nearly on par with Nvidia's 2nd place SKU is pretty good despite the fan boy crap.


Once you see how little Navi does with 3rd party cooling and then water will give you a glimpse of what is to come.
 
  • Like
Reactions: N4CR
like this
Gamer X said:
I guess Gamers Nexus will be doing a review on RDNA's ability to OC. Seems like Navi will be a hit with the enthusiasts and Over Clockers, if this guy is seeing 12% increase in just volt moding.

I too, am wondering what the golden samples of the XT are like.
Click to expand...

Are you really surprised by this given polaris and vega's history with being overvolted from factory and pulling too much power from pci-e bus?
 
Ranulfo said:
Are you really surprised by this given polaris and vega's history with being overvolted from factory and pulling too much power from pci-e bus?
Click to expand...

I don't care about Polaris, never owned that chip.

I think AMD is smart under Dr Su's guidance and being focused on many of AMD's deficiencies and setting industry leading price/performance and in such a small package. As such, RDNA is much different than GCN and has different pros & cons, but what I find coy... is that AMD is still holding back. The announcement of the Radeon 7's death, before 5700 Series aftermarket solutions are released/benched, hints at some things to come. And soon it seems, given how Jensen is holding out on releasing the 2080 SUPER on us...

So, it seems from all of the reviews, that the 5700 series clearly sits above Vega in games. And the 5700xt was pushing past the R7 in many game specific benches. Illustrating that a little ole 251mm^2 40 CU chip ($399) is poised to set the state for a 335mm^2 chip with 64CUs...



I believe that will be Dr Su's counter to Jensen RTX 2080 SUPER.
AMD will just announce the 5800 & 5900 Series Radeons. Waiting for the big announcement because these cards will make 4k gaming viable to the masses. I expect AMD to drop 5700 Series (ie: little Navi) prices a little more just before 64CU big-Navi hits.

I think both are coming sooner, than later.
 
Last edited:
Gamer X said:
I don't care about Polaris, never owned that chip.
Click to expand...

You might want to consider it since it was AMD's mainstream/mid-range chip for the last 3 years. That last minute rename to 5700/xt from 690 might be an indicator of something.
 
Ranulfo said:
You might want to consider it since it was AMD's mainstream/mid-range chip for the last 3 years. That last minute rename to 5700/xt from 690 might be an indicator of something.
Click to expand...

Yes, Polaris & Vega are indeed AMD's past. And today, Navi is AMD future...

So again, why are you brining up AMD's past GPU architectures and talking about past overvolting..? Navi is all new, with all new process and all new architecture why would you believe it will act like Polaris..?
 
Speculative but highly logical that we will see 5600, 5800 and perhaps 5900 & 5500 SKU's over the next 6 months. You don't just launch 2 SKU's on a whole new architecture.
 
Gamer X said:
So again, why are you brining up AMD's past GPU architectures and talking about past overvolting..? Navi is all new, with all new process and all new architecture why would you believe it will act like Polaris..?
Click to expand...

We should expect Navi to act like an AMD design, because that's how AMD designs have universally behaved.
 
Gamer X said:
Yes, Polaris & Vega are indeed AMD's past. And today, Navi is AMD future...

So again, why are you brining up AMD's past GPU architectures and talking about past overvolting..? Navi is all new, with all new process and all new architecture why would you believe it will act like Polaris..?
Click to expand...

The past is indicative of the future. Not always a guarantee but a nice place to start. I also came across a nice bug/issue with older 2011-2013 Gigabyte motherboards that clearly did not like AMD's gpu's from the R9 series to RX 480 and 570 likely due to power issues (that or bios vs uefi bios). So I pay attention to the power issue with their cards more closely now. Definately more interested to see the custom cards now.

Oh and some more data for you:
 
Ranulfo said:
Oh and some more data for you:
Click to expand...

Something worth bringing up on the subject of AMD overvolting is that we have to expect that AMD has set voltages at what appear to be higher than necessary across the board for a reason. The most obvious reason is that this is needed to cover for a relatively wide variation in ASIC performance- and if this is the case, then undervolting is something worth exploring but also must be approached just like overclocking, where no gains are guaranteed. And forum members have absolutely seen that be the case, with AMD GPUs that simply will not run at lower than stock voltage, while others of the same SKU can overclock respectably with sub-stock voltage.

It's really luck of the draw with AMD.
 
IdiotInCharge said:
Something worth bringing up on the subject of AMD overvolting is that we have to expect that AMD has set voltages at what appear to be higher than necessary across the board for a reason. The most obvious reason is that this is needed to cover for a relatively wide variation in ASIC performance- and if this is the case, then undervolting is something worth exploring but also must be approached just like overclocking, where no gains are guaranteed. And forum members have absolutely seen that be the case, with AMD GPUs that simply will not run at lower than stock voltage, while others of the same SKU can overclock respectably with sub-stock voltage.

It's really luck of the draw with AMD.
Click to expand...


While you are correct, I think given how large the mining community is that it is fair to say north of 75%+ of AMD SKUs can and have been successfully undervolted...I know I have done it on every card going back to the 6950X Lighting that MSI allowed you to flash to a 6970 under warranty...I had multiple 7950s, 3x 290s, 4x 580s, and 9X VEGA SKUs (covering V56,64, and VII) and now have a 5700XT....I have been able to undervolt EVERY card (haven't tried with NAVI yet)...Now I realize this is a small sample set, but if you know what my RL was life, you would know to say I am unlucky is a vast understatement to say the least.

I think the 5700/XT (and most likely the smaller Navi SKUs) are going to continue the trend...You can undervolt and drop power usage/temps/noise for little to no performance loss, or you can give it 15% more power and tap the SKU out while getting 10%+ performance increase without crazy power run away like with VEGA(s)....
 
ccityinstaller said:
I think the 5700/XT (and most likely the smaller Navi SKUs) are going to continue the trend...You can undervolt and drop power usage/temps/noise for little to no performance loss, or you can give it 15% more power and tap the SKU out while getting 10%+ performance increase without crazy power run away like with VEGA(s)....
Click to expand...

Nearly everything can be tweaked that isn't locked down. I have seen undervolting of NVidia GPUs as well, but there never was as much incentive since they were so far ahead on power usage.

Ultimately, I doubt these cards will sell on their ability to undervolt. What sells cards to most people is stock performance.
 
  • Like
Reactions: Auer
like this
Snowdog said:
Nearly everything can be tweaked that isn't locked down. I have seen undervolting of NVidia GPUs as well, but there never was as much incentive since they were so far ahead on power usage.

Ultimately, I doubt these cards will sell on their ability to undervolt. What sells cards to most people is stock performance.
Click to expand...


I do not remember saying that undervolting is going to be a selling feature. They are competitive cards, and with the 3rd party coolers/AIO hybrids they are going to be tied or besting 2070 Supers depending on how high the custom coolers are. It is quite clear that just this little Navi has a fair amount of headroom from the early OC'ing data out there. noko here is hitting 2.1Ghz on the stock blower, which is pretty nice. The fact that the 40CU part is as competitive as Nvidia's 3 fastest SKIU (let's ignore the RTX argument, as it has been beaten to death over and over) bodes well for even a 56CU part trading blows/besting 2080TI if the CU scaling does what I think it will do. It might take 60-64CU to take a clear lead, but I believe it could be done, and I suspect AMD is working on getting these out the door as we speak.
 
IdiotInCharge said:
Something worth bringing up on the subject of AMD overvolting is that we have to expect that AMD has set voltages at what appear to be higher than necessary across the board for a reason. The most obvious reason is that this is needed to cover for a relatively wide variation in ASIC performance- and if this is the case, then undervolting is something worth exploring but also must be approached just like overclocking, where no gains are guaranteed. And forum members have absolutely seen that be the case, with AMD GPUs that simply will not run at lower than stock voltage, while others of the same SKU can overclock respectably with sub-stock voltage.

It's really luck of the draw with AMD.
Click to expand...


Theres an undervolt button in wattman. Luck of the draw, its basically the simple fundamentals of Silicon lottery which applies to every chip that is overclockable no matter what brand.
 
ccityinstaller said:
I do not remember saying that undervolting is going to be a selling feature. They are competitive cards, and with the 3rd party coolers/AIO hybrids they are going to be tied or besting 2070 Supers depending on how high the custom coolers are. It is quite clear that just this little Navi has a fair amount of headroom from the early OC'ing data out there. noko here is hitting 2.1Ghz on the stock blower, which is pretty nice. The fact that the 40CU part is as competitive as Nvidia's 3 fastest SKIU (let's ignore the RTX argument, as it has been beaten to death over and over) bodes well for even a 56CU part trading blows/besting 2080TI if the CU scaling does what I think it will do. It might take 60-64CU to take a clear lead, but I believe it could be done, and I suspect AMD is working on getting these out the door as we speak.
Click to expand...

What I find interesting is that both the 2070 Super and the 5700 XT each have 2560 processing units. It looks like NVidia has a slight edge at equal clock speeds, but overall it's very close. You are counting on a big overclock to move past it toward the 2080.

So I think you are bit optimistic on CUs needed to equal the 2080Ti. You would probably need 4352 processing units (2080Ti count) to match it, which equals 68 CUs. Though I think 64 CU part would be much more likely for Big Navi, and they might try to make it up with clock speed.

Big Navi may be in the works, but I think AMD will have smaller Navi first. AMD could really use a laptop GPU part. Another market they are nearly shut out of.

Especially given a slide from AMD, showing that even the current Navi die size is relatively expensive at 7nm, which might make them reticent to go big, or wait until 7nm becomes a cheaper more mature process.
 
Last edited:
The little navi was already being tested in AMD's labs in Oct 2018. So judging from the release date of the 5700XT, it took 9 months to go from lab testing to release. No news sites are saying anything about a "big navi" being tested yet, so if another larger version of navi is planned, it could be a year away.

A year away is about right actually, that would put it at ~ 2 years after the 2080Ti, which is the timeframe of the little navi equaling the 1080Ti just over 2 years after its release.

I'm hopeful, but not willing to place any bets just yet.
 
  • Like
Reactions: Auer
like this
Snowdog said:
AMD could really use a laptop GPU part. Another market they are nearly shut out of.
Click to expand...

Looking at the efficiency of their product, they are quite likely to remain shut out of that market outside of their APUs- which themselves aren't terribly promising.
 
IdiotInCharge said:
Looking at the efficiency of their product, they are quite likely to remain shut out of that market outside of their APUs- which themselves aren't terribly promising.
Click to expand...

It looks very similar to NVidia's parts, so they should be able to deliver a small Navi that would be good for notebooks.
 
Snowdog said:
It looks very similar to NVidia's parts, so they should be able to deliver a small Navi that would be good for notebooks.
Click to expand...

That really depends on the flexibility of the architecture. Nvidia appears to be able to produce a part that can both run up to >2GHz on desktops and then can maintain decent TDP-limited clockspeeds when outfit for mobile use.

And here's the challenge: AMD is barely keeping parity with a significant node advantage. That advantage is available to Nvidia as well, and by the time AMD bothers to release something suitable for mobile, Nvidia will likely have a shrunk part in the wings that can exceed the AMD part in every way and include updated ray tracing hardware.

AMD could have led with mobile here and perhaps made a dent, but they chose not to- and now it may be too late.
 
IdiotInCharge said:
That really depends on the flexibility of the architecture. Nvidia appears to be able to produce a part that can both run up to >2GHz on desktops and then can maintain decent TDP-limited clockspeeds when outfit for mobile use.
Click to expand...

If anything, I would expect 7nm to have an edge when clocking down a bit more for mobile. Whatever the minor differences, they are in the ballpark now, and a LOT more viable for a mobile part than they were with Vega/Polaris.
 
IdiotInCharge said:
That really depends on the flexibility of the architecture. Nvidia appears to be able to produce a part that can both run up to >2GHz on desktops and then can maintain decent TDP-limited clockspeeds when outfit for mobile use.

And here's the challenge: AMD is barely keeping parity with a significant node advantage. That advantage is available to Nvidia as well, and by the time AMD bothers to release something suitable for mobile, Nvidia will likely have a shrunk part in the wings that can exceed the AMD part in every way and include updated ray tracing hardware.

AMD could have led with mobile here and perhaps made a dent, but they chose not to- and now it may be too late.
Click to expand...
I have the laptop version of the 1660Ti and it's a very powerful mobile GPU but it's still at the mercy of "laptop gaming" memes like the fact that you have to use headphones or the fans cranking up inside overwhelm the built-in speakers of the Dell G5. I don't see how a RX 5700 further cut down for mobile is going to be any different in that regards. It's going to be expensive, add some ounces of mass for the HSF, and hit 70-90% of the performance of the desktop part. For those people that really feel the need for a laptop with this capability it's going to happen if not with this generation by the time Big Navi arrives and 7nm is more efficient.
 
Snowdog said:
If anything, I would expect 7nm to have an edge when clocking down a bit more for mobile.
Click to expand...

I do too, they just have to beat Nvidia for mobile efficiency to matter, and they have to do it before Nvidia releases their own shrunk architecture. Remember that AMD will be competing with Nvidia's 10-month old architecture whenever they release.

Snowdog said:
Whatever the minor differences, they are in the ballpark now, and a LOT more viable for a mobile part than they were with Vega/Polaris.
Click to expand...

AMD has definitely made progress- but they can't be compared only to themselves.

DedEmbryonicCe11 said:
but it's still at the mercy of "laptop gaming" memes like the fact that you have to use headphones or the fans cranking up inside overwhelm the built-in speakers of the Dell G5.
Click to expand...

This is more an artifact of the OEM's choice of cooling than anything else. AMD would have to show better performance per watt at the same power and thermal limits that apply to current performance-oriented mobile GPUs for there to be any draw for their products.

And it highlights one of the unfortunate truths of mobile gaming: you have to spend both for a competent mobile GPU and for it to be well-implemented. Same goes for mobile CPUs. We've seen instances of both where a lower-end but better cooled part can turn out better performance and battery life than a higher-end part, most especially given how well CPUs and GPUs track performance to thermal constraints.

DedEmbryonicCe11 said:
if not with this generation by the time Big Navi arrives and 7nm is more efficient
Click to expand...

AMD really doesn't have the history of improving efficiency on the same node that Intel and Nvidia do. I've no doubt that AMD could improve efficiency if they chose to focus on that goal, however, doing so would be the exception more than the rule.
 
They could just undervolt it 50%, it would run around 70% ish desktop performance, and hopefully drop the watts/heat by half. Might work like that if the laptop had a really good fan/heatsink on the gpu.
Or would that be a significant enough drop in power/heat? I don't know any specs on current mobile GPU's in regards to power needs and heat dissipation.
 
Randall Stephens said:
My thoughts exactly. They're running up against TSMCs production capactiy would be my guess.
Click to expand...

Given that Nvidia is looking to use more of Samsung's capacity, TSMC might be largely tapped out. If that's the case, then there may be a limit to AMD's ability to expand into various markets- which is what one would expect when using a third-party fab.
 
I am surprise that AMD has not tap into Samsung 7nm process. Both CPUs and GPUs consuming TSMC 7nm process.

Something I am wondering about is RNDA game optimizations and how much improvement can be had. Could be a lot.
 
IdiotInCharge said:
Given that Nvidia is looking to use more of Samsung's capacity, TSMC might be largely tapped out. If that's the case, then there may be a limit to AMD's ability to expand into various markets- which is what one would expect when using a third-party fab.
Click to expand...

I doubt it is capacity. NVidia said in a statement elsewhere that they have always qualified both Samsung and TSMC, which means they have some idea of the limit of both processes and likely prefer Samsung for some economic or performance reason.


noko said:
I am surprise that AMD has not tap into Samsung 7nm process. Both CPUs and GPUs consuming TSMC 7nm process.
Click to expand...

No real sign that they are constrained. Plus it costs time/staff and a shitload of money to move a design to another Fab. It could have happened elsewhere, but the only design I am aware of that was every split between Fabs was Apple iPhone chip and their pockets are the deepest and their volumes are massive.
 
This card has other good uses as well other than gaming. It's good at rendering and streaming video too.
 
Snowdog said:
I doubt it is capacity. NVidia said in a statement elsewhere that they have always qualified both Samsung and TSMC, which means they have some idea of the limit of both processes and likely prefer Samsung for some economic or performance reason.




No real sign that they are constrained. Plus it costs time/staff and a shitload of money to move a design to another Fab. It could have happened elsewhere, but the only design I am aware of that was every split between Fabs was Apple iPhone chip and their pockets are the deepest and their volumes are massive.
Click to expand...
That is true. Now AMD could split up the foundries for something like the 5600 series (if they have one), APU? RyZen 3 has to be some rather large orders is my thought, more so then GPU's.
 
noko said:
That is true. Now AMD could split up the foundries for something like the 5600 series (if they have one), APU? RyZen 3 has to be some rather large orders is my thought, more so then GPU's.
Click to expand...

They could easily produce their chipset dies, uncore dies, and core dies at different fabs if need be. If they can efficiently split up GPU work they could probably do the same there too.
 
IdiotInCharge said:
They could easily produce their chipset dies, uncore dies, and core dies at different fabs if need be. If they can efficiently split up GPU work they could probably do the same there too.
Click to expand...

They Already do this...the I/O die is made on GF 12nm, and the chiplets are TSMC 7nm...I can't remember where they are fabbing x570, but I believe that everything prior to x570 was made on ab older 55nm process.
 
I have a Sapphire 5700XT right now in Wattman i have it set to 1078mv @ max clock of 2048 running heaven benchmark the card boosted to 2021mhz ram ran at 875mhz total power draw peak was 197w temp was 83c had fan profile stock except i boost 70degrees to 30% fan is not that noticable. I have it in a meshify-c mini with 2 140mm noctua case fans and 1 120mm noctua rear fan. I have a D15S on the cpu dual fan setup. Extremely high ambient temps in the room due to heat wave.
 
123Lanoix said:
I have a Sapphire 5700XT right now in Wattman i have it set to 1078mv @ max clock of 2048 running heaven benchmark the card boosted to 2021mhz ram ran at 875mhz total power draw peak was 197w temp was 83c had fan profile stock except i boost 70degrees to 30% fan is not that noticable. I have it in a meshify-c mini with 2 140mm noctua case fans and 1 120mm noctua rear fan. I have a D15S on the cpu dual fan setup. Extremely high ambient temps in the room due to heat wave.
Click to expand...


Nice results. I'm seeing 2100+ , so I am hoping to see 2200-maybe even 2300 once I get it under water!
 
ccityinstaller said:
Nice results. I'm seeing 2100+ , so I am hoping to see 2200-maybe even 2300 once I get it under water!
Click to expand...
I'm trying to keep my Power at 200watts or below for the card. I'm going to try the washer mod tomorrow and put thermal pads on the back plate. I should be able to repaste it with thermal grizzly. I got vulcanized rubber washers used usually for motherboards. That should lower temps at least 10 degrees?
 
I really like the stock look of the card with the red lit radeon logo. I don't want to ugly it up by putting on a Arctic cooler
 
You must log in or register to reply here.
Back
Top