Radeon 5700 Reviews Thread

[Head1985]

buildzoid comfirmed that 5700 is locked and cant be oc past 1800-1850mhz

 

[Paul98]

It's really nice to see that these have far higher performance per transistor and performance per watt. Should be even better with a quality cooler and a few tweaks to settings.

 

[amrnuke]

What are everyone's predictions about timing of potential 5600 and 5800 series? I have to imagine Navi 20 or Navi 21 (depends on the rumor) would be something they would pounce on after 2080 SUPER releases. But I think there's also a huge gap in the market for a 5600 in the $199-$249 price range, e.g. 5600 and 5600XT competitors for the 1660 and 1660 Ti.

 

[Udgnim]

buildzoid comfirmed that 5700 is locked and cant be oc past 1800-1850mhz

BIOS to go beyond AMD's cap on the 5700 & 5700 XT

https://www.tomshw.de/2019/07/11/un...playtables-fuer-die-rx-5700-und-rx-5700-xt/3/

considering that someone has been able to get the 5700 XT up to 2.2 Ghz under water

https://www.tomshw.de/2019/07/11/un...erplaytables-fuer-die-rx-5700-und-rx-5700-xt/

I can kind of see why AMD has the 5700 capped at around 1.8 to 1.85 Ghz.

The 5700 would be very good value with the available overclocking headroom under a decent third party HSF.

 

[tviceman]

Are we in 2021?

16nm isn't 5 years old.

https://www.tsmc.com/english/dedicatedFoundry/technology/16nm.htm

Yes, it is 5 years old this November. A quick and easy google search can easily prevent these kinds of interactions from happening.

 

[IntelUser2000]

Yes, it is 5 years old this November. A quick and easy google search can easily prevent these kinds of interactions from happening.

So what? Products weren't available at that time. The first GPU was much, much later.

 

[tviceman]

So what? Products weren't available at that time. The first GPU was much, much later.

the node's lifespan did not begin when the first GPU products became available. The node' lifespan began when it started manufacturing. How is this a fact I am still having to explain? I said 16nm was five years old which is 2014 which is 100% accurate. The node was up and running in 2014. Fact.

 

[crisium]

BIOS to go beyond AMD's cap on the 5700 & 5700 XT

https://www.tomshw.de/2019/07/11/un...playtables-fuer-die-rx-5700-und-rx-5700-xt/3/

considering that someone has been able to get the 5700 XT up to 2.2 Ghz under water

https://www.tomshw.de/2019/07/11/un...erplaytables-fuer-die-rx-5700-und-rx-5700-xt/

I can kind of see why AMD has the 5700 capped at around 1.8 to 1.85 Ghz.

The 5700 would be very good value with the available overclocking headroom under a decent third party HSF.

It's anti-consumer.

The cards are priced to goddam close already. Why artificially lock?

7950 was 87.5% of the shaders of the 7970 for $100 less than the 7970.
290 was 91% of the shaders of the 290x, for $150 less than the 290X.

5700 is 90% of the shaders of the 5700 XT for $50 less.

Yet the first 2 could be max OCed to be equal clocks to the more expensive GPU, and thus be around ~92-96% of the performance (fullchips have less performance-per-FLOP that cutdown because they have the same ROPs and bandwidth).

The same WOULD be true for the 5700, and it's even priced closer to the more expensive GPU so it has every damn right to continue the tradition. There's no need to artifically limit the OC. There never was, and there is even less reason now that they are within 15% of the price of each other, which is unheard of. If anything, the previous gen cutdown chips should have been OC limitted and the 5700 should have been unlocked because 1) it is so closely priced together and 2) AMD dGPU gaming market share is small so they need to throw that bone to enthusiasts.

Speaking of enthusiasts, it's a literal slap in the face to them. It's the casual users who just look at stock vs stock and agree to spend the $50 extra for the XT. It WOULD be the enthusiasts who realize it's ~92-96% as fast when both are max OC, and choose to save the $50. But now they are denied. Why is a market loser being anti-consumer? That's the market leader's job.

Well, maybe I should be less negative and hope it is just some reference design mistake and the AIB's will fix it. Hopefully no one defends AMD here. Even a stockholder would question why the market loser is choosing anti consumer when the market leader doesn't. IMO, if AMD wants more market separation, then cut down the chip more as NVDA does, but have them clock about the same. So the casual users see the big gap, and the enthusiasts know it cant catch up even with an OC, but at least they have fun OCing vs now where they just take an artificial OC limit.

 

[.vodka]

It's anti-consumer.

The cards are priced to goddam close already. Why artificially lock?

7950 was 87.5% of the shaders of the 7970 for $100 less than the 7970.
290 was 91% of the shaders of the 290x, for $150 less than the 290X.

5700 is 90% of the shaders of the 5700 XT for $50 less.

Yet the first 2 could be max OCed to be equal clocks to the more expensive GPU, and thus be around ~92-96% of the performance (fullchips have less performance-per-FLOP that cutdown because they have the same ROPs and bandwidth)..

I agree. Seeing them cap the cards is weird. I'd expected that from nV, not AMD.

I mean... Your CPUs are all unlocked (they'd even unlocked the little 200GE after one or two AGESA revisions) and... your brand spanking new GPUs have a hard cap, even for the flagship?

I guess they don't want the tradition to continue. As you mention, historically ATI's (and then AMD's) cut down chip was the better buy, unbeatable price/perf, and could be overclocked to match the full chip in the stock flagship.

I suppose they won't go full nV right down to voltage locking the cards in the future. At least they didn't take away the soft mods. Hawaii cards were the last you could BIOS mod... Polaris and Vega locked the BIOSes down, but you could do whatever you wanted.

 

[mopardude87]

In another thread someone told me costs are sky rocketing. Explains why AMD couldn't include the washers. Haha.

Joking aside, this cooler is a joke, and AMD deserves all the flak for trotting it out. If you watched his tear down video, he tried to remove the dent from the shroud (which is 100% cosmetic).

For sure, such a easy fix and so dang cheap. fair to assume if they dropped the pad for a non curing paste and washers that just MAYBE it would have about evened out cost wise? Or is it the actual man power of someone on a line somewhere who has to physically add the paste, then mount the cooler while now adding 4 washers then 4 screws? Of course if the heatsink was machined right or the backplate had premounted washers then that can save time?

Also kind of curious if this moves over to the 5700 which does not have that dent which from when i first saw a picture of it i thought someone had dropped it good. What is the dent doing?

 

[tviceman]

For example lets take the GTX980 vs the GTX1080,
GTX980 has 5.2B transistors with a die size of 398mm2 at 28nm and the GTX1080 came with 7.2B transistors with a die size of 314mm2 at 16nm.
So if they will increase the transistor count of the x06 Ampere they could be come close to or bigger than 300mm2 and that will compete with another NAVI and not NAVI 10.

GP106 had 1.2 billion less transistors and was 20% smaller than Polaris 10. I'm sure Nvidia will figure it out and still maintain a tech lead when on the same node.

 

[AtenRa]

GP106 had 1.2 billion less transistors and was 20% smaller than Polaris 10. I'm sure Nvidia will figure it out and still maintain a tech lead when on the same node.

GP106s (GTX1060) FP16 and FP64 performance was a joke against Polaris 10. Polaris 10 used a lot more transistors for FP16 and FP64 than GP106, that increased its die size.
When NVIDIA allocated more transistors for FP16 and FP64 the transistor count ballooned to 6.6B on the TU116 (GTX1660ti). And this is without RT cores.

Both AMD and NVIDIA are now allocating more transistors for FP16 and FP64 than before and NVIDIA also have RT cores that increase transistor count and die size even more. Dont expect 40-50% higher performance over TU106 (RTX2060 Super) without higher transistor count with Ampere.

 

[darkswordsman17]

I agree. Seeing them cap the cards is weird. I'd expected that from nV, not AMD.

I mean... Your CPUs are all unlocked (they'd even unlocked the little 200GE after one or two AGESA revisions) and... your brand spanking new GPUs have a hard cap, even for the flagship?

I guess they don't want the tradition to continue. As you mention, historically ATI's (and then AMD's) cut down chip was the better buy, unbeatable price/perf, and could be overclocked to match the full chip in the stock flagship.

I suppose they won't go full nV right down to voltage locking the cards in the future. At least they didn't take away the soft mods. Hawaii cards were the last you could BIOS mod... Polaris and Vega locked the BIOSes down, but you could do whatever you wanted.

I'd guess its because of them doing more with the boost clocks, where they'd prefer to let the chips automatically determine clocks and this makes it so they can't boost to stupid levels when doing that (so that the cards don't do something like boost way high, then throttle, then boost and throttle, causing really inconsistent performance).

 

[Bouowmx]

Don't think NVIDIA will have an advantage in transistor count with Ampere 7 nm, but area depends on how NVIDIA takes advantage of 7 nm density, as AMD did not at all.

TU102 * 1.33 = ~24800 M transistors, 604 mm^2 with same density as Navi 10 (41.0 M/mm^2) or 322 mm^2 with 80% of theoretical TSMC 7 nm (77.0 M/mm^2).

 

[tviceman]

Don't think NVIDIA will have an advantage in transistor count with Ampere 7 nm, but area depends on how NVIDIA takes advantage of 7 nm density, as AMD did not at all.

TU102 * 1.33 = ~24800 M transistors, 604 mm^2 with same density as Navi 10 (41.0 M/mm^2) or 322 mm^2 with 80% of theoretical TSMC 7 nm (77.0 M/mm^2).

The smaller the chip, the harder it is to cool at the same TDP. Navi was probably designed less dense to hit higher clock speeds. With many reviews seeing it hit 90 degrees, imagine if it was producing the same heat but with 10-15% less surface area to cool.

 

[.vodka]

The smaller the chip, the harder it is to cool at the same TDP. Navi was probably designed less sense to hit higher clock speeds. With many reviews seeing it hit 90 degrees, imagine if it was producing the same heat but with 10-15% less surface area to cool.

It's more of an issue with the stock blower, which blows, as expected. The RX5700 cards have the same mounting dimensions as Hawaii, so strapping an open air heatsink + fans as seen in the old Hawaii cards bring the temperature down to a chilly ~65-70°C while remaining silent, and running at the maximum boost clocks all the time (~1850MHz for the 5700, ~2050-2100 for the 5700XT)

https://twitter.com/x/status/1148113823137533952

https://twitter.com/watchtimmyjoe/status/1148026906991833088

Heat output relative to die area for GPUs is still manageable on TSMC 7nm. Big Navi shouldn't be too hard to cool.

What is hard to cool are the Zen2 chiplets, the cores are tiny (<4mm²) and clockspeeds above 4.2-4.3GHz become difficult to manage on ambient temperature at ~1.3v. Cold lets the cores scale to >5GHz at the same voltage as tested by der8auer...

-----------------------------------------------------------

Also, transistor for transistor Navi 10 and TU106 as seen in the RX5700XT and the 2070 are pretty similar overall (10.3 billion, 10.8 billion). Someone at B3D said that RT support for these cards accounts for ~3% of the die size, it would seem most of the extra space is consumed by the tensor cores. TU106 on 7nm would still be bigger than Navi 10, would probably clock a bit higher, too.

Power table modded 5700XT on water with a +90% power limit @ 2300MHz is at ~2080 levels of performance while consuming ~250w. Not too bad for it being pushed so far out its comfort zone. This again bodes well for big Navi.

 

[tviceman]

It's more of an issue with the stock blower, which blows, as expected. The RX5700 cards have the same mounting dimensions as Hawaii, so strapping an open air heatsink + fans as seen in the old Hawaii cards bring the temperature down to a chilly ~65-70°C while remaining silent, and running at the maximum boost clocks all the time (~1850MHz for the 5700, ~2050-2100 for the 5700XT)

https://twitter.com/x/status/1148113823137533952

https://twitter.com/watchtimmyjoe/status/1148026906991833088

Heat output relative to die area for GPUs is still manageable on TSMC 7nm. Big Navi shouldn't be too hard to cool.

What is hard to cool are the Zen2 chiplets, the cores are tiny (<4mm²) and clockspeeds above 4.2-4.3GHz become difficult to manage on ambient temperature at ~1.3v. Cold lets the cores scale to >5GHz at the same voltage as tested by der8auer...

Not disagreeing with you on any of this, but my point still remains. A smaller chip with an equal TDP to a larger chip will be harder to cool. AMD said they engineered Navi to hit higher clock speeds, and being less dense may have been in purpose to achieve those results.

 

[mopardude87]

I am randomly running into this" conspiracy theory" about how AMD deliberately made sure the Navi reference blower was so bad, your just about forced into buying a partner card. As much as i think its crazy there might actually be some merit to this? I leave this video.

 

[Zstream]

I am randomly running into this" conspiracy theory" about how AMD deliberately made sure the Navi reference blower was so bad, your just about forced into buying a partner card. As much as i think its crazy there might actually be some merit to this? I leave this video.

Possibly but AMD always screws up their launches. It’s as if someone is directly sabotaging them on purpose, or afraid Nvidia will buy them if they offer too much performance.

 

[railven]

I am randomly running into this" conspiracy theory" about how AMD deliberately made sure the Navi reference blower was so bad, your just about forced into buying a partner card. As much as i think its crazy there might actually be some merit to this? I leave this video.

I feel like the only reason people would say that is because they'd rather not believe a company can be this careless. What's this, the fourth time? Fool me once, yada yada.

One thing I will state though, NV feels like their marketing and design teams are chasing - say Apple. Their box design is nice (much nicer than any of their AIB's I've ever purchased), their cooler while not amazing is aesthetically pleasing - it has "style", their designs are more plug and play - set it and forget it!

AMD seems more like a relic of the past (and I say this in a good way). You tinker with it, dropping volts, fine tuning your fan curve, mix and matching washers (this is a joke!), it's a shroud is just a box (now with a dent ), for that perfect balance of performance. A tinker's cards. Steve at GN said it best regarding the Vega 56.

 

[exquisitechar]

Navi seems to have some nice OC headroom. Bodes well for the upcoming pricier and higher quality versions, the reference version is really nowhere near good enough for these chips, especially the XT. I want to see a (reasonable) max OC 5700/XT vs max OC 2060S/2070S comparison at that point. Thinking of getting a 5700 this year...

 

[railven]

For sure, such a easy fix and so dang cheap. fair to assume if they dropped the pad for a non curing paste and washers that just MAYBE it would have about evened out cost wise? Or is it the actual man power of someone on a line somewhere who has to physically add the paste, then mount the cooler while now adding 4 washers then 4 screws? Of course if the heatsink was machined right or the backplate had premounted washers then that can save time?

I recall a while back a few people arguing about cost of manufacturing. I personally don't know, but from my understanding of how they apply paste, it's still a pad. They set it place (if not machine pre-applied to the cooler base). No way a human is getting a perfect square and thickness I see on custom coolers (assuming the coolers are all probably made in the same factory(ies)).

If you watch Steve's break down of the card, I get the feeling AMD just cut corners. Plain and simple. There are no thermal pads on the back plate (of which Steve seemed to have gotten pissed, dismantled two cards to make sure it wasn't just his dud sample). Not sure what AMD's game was with this blower, but it didn't help them any where and I feel the cost saving they might have had was not worth the backlash.

Also kind of curious if this moves over to the 5700 which does not have that dent which from when i first saw a picture of it i thought someone had dropped it good. What is the dent doing?

It does absolutely nothing. The fan housing it self has a shroud which has a curve to control the air flow. Some genius over at AMD figured adding a dent, sorry a contour, to the shroud (which lines up with the fan shroud's curve) would be aesthetically pleasing. First time I saw the card, I too thought it was dropped, or just some weirdness with the image. Then I saw more, then I read AMD's explanation, then I saw Steve's break down video, and then I realized AMD has no marketing chops whatsoever.

Steve comically tried to remove the dent, he couldn't.

 

[DAPUNISHER]

Spoiler: quoted text is here

I feel like the only reason people would say that is because they'd rather not believe a company can be this careless. What's this, the fourth time? Fool me once, yada yada.

One thing I will state though, NV feels like their marketing and design teams are chasing - say Apple. Their box design is nice (much nicer than any of their AIB's I've ever purchased), their cooler while not amazing is aesthetically pleasing - it has "style", their designs are more plug and play - set it and forget it!

AMD seems more like a relic of the past (and I say this in a good way). You tinker with it, dropping volts, fine tuning your fan curve, mix and matching washers (this is a joke!), it's a shroud is just a box (now with a dent ), for that perfect balance of performance. A tinker's cards. Steve at GN said it best regarding the Vega 56.

Preach it brudda.

I will preface this by saying that the Thor V.2 case I use is far from silent, and ugly, very ugly. It is setup as a go case, not a show case, or slow case. Lots more tweaking to do, but so far my best mix for the 5700xt is 1000mv 1850MHz boost 900MHz memory mostly 2000rpm fan speed or less. I cannot hear it over the 230mm side fan which is dialed to full to blow big air over the card. It makes a soft hum white noise sound. I heard it for a second here in there as it ramped as high as 2500rpm but it settled down fast. 63c temp 74c junction temp 154W (wattman numbers). 21317 v. 22042 fire strike, with vacuum cleaner noise and 2100MHz boost 224W. 700 points for cool and much quieter is acceptable, but I will find a better blend yet.

 

[railven]

Preach it brudda.

I will preface this by saying that the Thor V.2 case I use is far from silent, and ugly, very ugly. It is setup as a go case, not a show case, or slow case. Lots more tweaking to do, but so far my best mix for the 5700xt is 1000mv 1850MHz boost 900MHz memory mostly 2000rpm fan speed or less. I cannot hear it over the 230mm side fan which is dialed to full to blow big air over the card. It makes a soft hum white noise sound. I heard it for a second here in there as it ramped as high as 2500rpm but it settled down fast. 63c temp 74c junction temp 154W (wattman numbers). 21317 v. 22042 fire strike, with vacuum cleaner noise and 2100MHz boost 224W. 700 points for cool and much quieter is acceptable, but I will find a better blend yet.

My wife pointed out to me that I do far less tinkering with my setup as I use to. I don't know if I'm just getting older (and thus losing as much interest in the hobby) or because I switched to NV and tinkering at best gives mild improvements. When I had my 7970s, I was making custom fan curves and loving it. Tinkering is a hard to quantify metric. For some people, a card like the 5700 XT is a project, for other's it's loud and hot. Why I always say - do you!

We (general 'we') won't always agree on things! But we're all nerds at heart!

 

[maddie]

My wife pointed out to me that I do far less tinkering with my setup as I use to. I don't know if I'm just getting older (and thus losing as much interest in the hobby) or because I switched to NV and tinkering at best gives mild improvements. When I had my 7970s, I was making custom fan curves and loving it. Tinkering is a hard to quantify metric. For some people, a card like the 5700 XT is a project, for other's it's loud and hot. Why I always say - do you!

We (general 'we') won't always agree on things! But we're all nerds at heart!

Pssst, a secret. You are getting older.