AMD Navi Product Reviews and Previews: (5500, 5600 XT, 5700, 5700 XT)

[DavidGraham]

That has nothing to do with 5600XT and everything to do with PowerColor

I strongly disagree, PowerColor is one of the AIB powerhouses of AMD, this whole BIOS mess clearly pressured them to hastly release untested BIOS versions into the wild, this shows why AIBs like MSI and ASUS chose not to gamble on untested memory speeds and risk taking increased RMA from faulty products.

 

[Deleted member 2197]

It's all on AMD. As one 5600 XT review this week stated:

SAPPHIRE could not help the new BIOS or last-minute changes from AMD. I hold no ill feelings towards SAPPHIRE for that, as it was all AMD. I'm not going to punish SAPPHIRE or AMD in my final charts, but it will make it into the bottom line of this review. AMD -- learn from this, you're hurting AIB partners and gamers and haven't even really apologized for this gigantic cock up. You jebaited yourselves.

https://www.tweaktown.com/reviews/9...pulse-oc-review-please-update-bios/index.html

 

[Kaotik]

I strongly disagree, PowerColor is one of the AIB powerhouses of AMD, this whole BIOS mess clearly pressured them to hastly release untested BIOS versions into the wild, this shows why AIBs like MSI and ASUS chose not to gamble on untested memory speeds and risk taking increased RMA from faulty products.

It's all on AMD. As one 5600 XT review this week stated:

https://www.tweaktown.com/reviews/9...pulse-oc-review-please-update-bios/index.html

AMD didn't force any AIB to release new higher clocked BIOS. AMD allowed AIBs to release BIOS with higher factory overclocks, some took everything they could, others less, some nothing.
Not all cards will receive any sort of updated BIOS, it's up to the AIBs, and not all models that did get BIOS update are be clocked the same, it's again up to the AIB.

Only thing AMD did wrong was allowing the higher clocks so close to launch, but it was the best way to counter NVIDIAs $299 RTX 2060. Blaming AMD for AIBs BIOSes being unfinished or some such is just BS, that's AIBs fault, not AMDs - if you're not confident in your product at higher clocks, don't clock it that high, no-one is forcing you to.

 

[DavidGraham]

AMD didn't force any AIB to release new higher clocked BIOS. AMD allowed AIBs to release BIOS with higher factory overclocks, some took everything they could, others less, some nothing.

It's basically the same thing, if those AIB didn't release new BIOS their cards will be DOA, AMD instructed them to release the BIOS quickly so they can send fast cards to reviewers to obtain favorable reviews against the 2060, and in their haste it backfired. It's not the AIB's fault at all, it rests solely on AMD.

 

[Kaotik]

It's basically the same thing, if those AIB didn't release new BIOS their cards will be DOA, AMD instructed them to release the BIOS quickly so they can send fast cards to reviewers to obtain favorable reviews against the 2060, and in their haste it backfired. It's not the AIB's fault at all, it rests solely on AMD.

And if they weren't confident with them they shouldn't have sent them, or do as MSI and Asus did and release separate new models for the new BIOS where they have confidence in them. Sapphire doesn't seem to have any issues with their BIOS. Come to think of it, I don't think even the usual suspects which didn't do so well in 5700 have had such issues.
In fact, I have hard time believing what Steve has is by any means normal for the Red Dragon either, since all other reviews of said model are showing quite a bit different results
Guru3D: Load temperature 73C https://www.guru3d.com/articles_pages/powercolor_radeon_rx_5600_xt_red_dragon_review,9.html
Coolaler: Load temperature (Furmark) 72C https://www.coolaler.com/threads/rx5600xt-nvidia-amd-5600xt-rtx2060-1660ti-super-vega64-56.358202/
Igor's Lab: Load temperature 70-71C (old bios 68-69C) https://www.igorslab.de/en/amd-rade...ios-boost-to-an-efficient-rtx-2060-killer/12/
(Found 2 more reviews, JayZ2Cents and BitWit, but they didn't have any temperature results (at least I didn't spot such), so didn't include them)
Haven't heard any reports on our forums about Red Dragons overheating either, haven't scoured other forums.

Also PowerColor has actually gone and confirmed on Twitter that they consider running the 5600 XT memory at 14 Gbps, as rated by memory manufacturer, to be running within spec, regardless if the exact modules have or haven't been certified by AMD to do so.

https://twitter.com/x/status/1223016604390182915

 

[DavidGraham]

fact, I have hard time believing what Steve has is by any means normal for the Red Dragon either,

Yes it is, PowerColor provided another BIOS to fix the fan issue, they wouldn't do so if it was a random problem.

 

[Kaotik]

Yes it is, PowerColor provided another BIOS to fix the fan issue, they wouldn't do so if it was a random problem.

Then how come no-one else is reporting such? Did Steve somehow get a card with differente BIOS than the rest?

 

[Davros]

This 14gpbs speed is it a standard running speed or is it the gddr equivalent of a xmp profile?

 

[Kaotik]

This 14gpbs speed is it a standard running speed or is it the gddr equivalent of a xmp profile?

It's standard running speed for the memory chips they're using.

 

[3dilettante]

It seems odd to me that AMD and the board vendors can't trust that modules sold by the DRAM manufacturers cannot meet the gbps specs they're sold at.
Perhaps it's a question of whether they can meet those speeds while keeping other latency settings consistent with the BIOS tables that exist?
If AMD didn't try to keep its supply segregated by those timings, perhaps it would take additional sorting to find the ones that work vs overly conservative BIOS settings.

That's the scenario I can think of besides other elements like boards and components not having the margins for an unexpected upclock.

 

[Deleted member 2197]

I found this curious because on PowerColor's website, its stock 5600 XT model (AXRX 5600XT 6GBD6-3DH/OC) lists 12Gbps memory in the specifications section. I reached out to PowerColor to ask if that meant the memory on that card can run at 14Gbps, but is clocked slower because it's a reference model. The answer basically amounts to 'sort of.'

"Our Stock model 5600 XT follows AMD's reference specs but we share [the same] memory modules as Red Devil and Red Dragon. You would ask why we don't run at 14Gbps then—the reason is Red Devil and Red Dragon have a higher PCB layer count, 10 versus 8 found on the stock 5600 XT, which means the memory signal is much cleaner and stable with our premium models. Having a 10 layer PCB is important to the higher clocks on the GDDR6 14Gbps memory," PowerColor said.

In other words, PowerColor's stock 5600 XT is in fact using memory chips validated to run at 14Gbps, the same as found on its Red Devil and Red Dragon variants, but still might not be able to run at that speed because of the lower cost and lower quality printed circuit board (PCB) design.

https://www.pcgamer.com/powercolor-sheds-light-on-why-memory-speeds-differ-on-the-radeon-rx-5600-xt/

 

[Gubbi]

It seems odd to me that AMD and the board vendors can't trust that modules sold by the DRAM manufacturers cannot meet the gbps specs they're sold at.

Seeing a few of Gamers Nexus' teardowns of some recently released GPUs, I would expect it has more to do with poor thermal solutions for the DRAM chip and/or VRMs.

Cheers

 

[Kaotik]

Seeing a few of Gamers Nexus' teardowns of some recently released GPUs, I would expect it has more to do with poor thermal solutions for the DRAM chip and/or VRMs.

Cheers

I would rather put my money on the earlier suggested thinner PCBs on some 5600 XT's, since some of the ones which kept 12 Gbps memory use exact same cooler and PCB layout (with memchip & some VRM stripped) as 5700-series, all of which have 14 Gbps mem

 

[Deleted member 2197]

February 20, 2020

 

[Deleted member 2197]

AMD Radeon Pro W5500 Review
April 10, 2020
https://hothardware.com/reviews/amd-radeon-pro-w5500-gpu-review

 

[Davros]

"the Radeon RX 5500 XT with its 4 GB VRAM and PCIe x8 interface" ( from : https://forum.beyond3d.com/threads/no-dx12-software-is-suitable-for-benchmarking-spawn.58013/page-55 )
Why 8x even the lowest of the low cards have a x16 interface (unless I'm wrong on that)

 

[CarstenS]

GT 1030 only has an x4 interface. For one, it saves power and it saves space.

 

[Pressure]

"the Radeon RX 5500 XT with its 4 GB VRAM and PCIe x8 interface" ( from : https://forum.beyond3d.com/threads/no-dx12-software-is-suitable-for-benchmarking-spawn.58013/page-55 )
Why 8x even the lowest of the low cards have a x16 interface (unless I'm wrong on that)

It's PCIe 4.0 x8.

 

[Kaotik]

"the Radeon RX 5500 XT with its 4 GB VRAM and PCIe x8 interface" ( from : https://forum.beyond3d.com/threads/no-dx12-software-is-suitable-for-benchmarking-spawn.58013/page-55 )
Why 8x even the lowest of the low cards have a x16 interface (unless I'm wrong on that)

It's indeed PCIe 4.0 x8 as pointed out above, but no, not all cards have x16, not even by a long shot.
On AMDs side at least Oland, Polaris 11 (Baffin) and Polaris 12 (Lexa) are all PCIe 3.0 x8 and on NVIDIAs side GP108 is PCIe 3.0 x4, GM108 PCIe 3.0 x8 and GK208 PCIe 2.0 x8