HIS RX 560 and its variant

[Archmage1809]

Hi guys, I just bought a RX 560 4GB from HIS and it pretty weird.
I have noticed on HIS homepage this is the GREEN version, which is 14 Compute Unit (just like the RX 460), been cut down to 896 Shader Unified, 16/56 ROP/TMU. Underclocked memory to 1500mhz.

I have overclocked memory to 1750mhz got no problem from doing it.
And I noticed a thing that Release Date (highlighted) is Aug 8th 2016, the release date of RX 460.
I have a feeling that my RX 560 was flashed from a RX 460 right out manufacturer. AMD Crimson still recognize it as a RX 560. And someone reported that if you using an older Crimson driver, it will shown as a RX 460

BIOS is BAFFIN instead of POLARIS21 make it more suspicious.


A friend reported his also GREEN version from HIS is a full RX 560.

I'm asking him for the BIOS and thinking about flash his bios to my RX 560 to unlock it to full power.

 

[Archmage1809]

My BIOS (the weaken RX 560/the re-flashed RX 460): https://www.techpowerup.com/vgabios/195328/195328
Friend BIOS (the full power RX560): https://www.techpowerup.com/vgabios/195343/195343

 

[eidairaman1]

You have what is known as the 560D Series.

Baffin and polaris are different.

I wouldn't attempt putting a different bios from a different card on yours as it will brick and can be unrecoverable.

Just don't do it.

You would be better off sending the card back for a 570 or 580.

 

[MrGenius]

It might work. But it probably won't. I don't think there's any major risk of bricking the card, since they're so similar. I can't say for sure. But I doubt it. The memory timings for Micron MT51J256M32HF are a little different for the 1625MHz and 1875MHz straps. Which shouldn't matter compatibility wise(exact same model memory ICs supported). But will have an impact on performance between 1625MHz to 1749MHz and 1875MHz to 1999MHz memory speeds. If you're running the memory @ 1750MHz to 1874MHz(or anywhere from 200MHz to 1624MHz or 2000MHz+) the timings are the same(so performance will be the same).

Memory timings (Micron MT51J256M32HF)

Your card BIOS 1625MHz: 17-17-22-22-60-21-164 (tighter)
Your friend's card BIOS 1625MHz: 19-19-24-24-65-22-178 (looser)
Advantage = your card BIOS between 1625MHz to 1749MHz memory speeds

Your card BIOS 1875MHz: 22-22-28-28-75-24-206 (looser)
Your friend's card BIOS 1875MHz: 21-21-26-26-70-23-192 (tighter)
Advantage = your friend's card BIOS between 1875 to 1999MHz memory speeds

Anyway...the difference between Baffin PRO/XT and Polaris 21 XT/XL is basically in name only. They wanted to pretend like they weren't doing a full on rebrand so they gave Baffin a new name to throw people off. Same thing with Ellesmere PRO(D)/XT and Polaris 20 XL/XTX. Same chips basically...different name. And people are flashing their Ellesmere cards to Polaris cards successfully(in most cases). So...flashing Baffin cards to Polaris cards shouldn't be a problem either(in most cases).

But...like I said...it probably won't work. Although it might. Or it might even appear to work...yet yield no performance increase. Since additional shaders and TMUs might show(in GPU-Z for example) as being unlocked when they really aren't. And the memory timings thing will definitely affect memory performance at the speeds specified above. And there's also a slim chance you might brick the card. Exceedingly slim though IMO. It's a risk I would be willing to take. But I'm also very confident I could unbrick it if I had to.

Final note: You will have to force flash it because the Device ID and Subsystem ID are different.

 

[PLAfiller]

Made me check my RX560. I have Gigabyte Gaming RX560, but its alright :

 

[eidairaman1]

It might work. But it probably won't. I don't think there's any major risk of bricking the card, since they're so similar. I can't say for sure. But I doubt it. The memory timings for Micron MT51J256M32HF are a little different for the 1625MHz and 1875MHz straps. Which shouldn't matter compatibility wise(exact same model memory ICs supported). But will have an impact on performance between 1625MHz to 1749MHz and 1875MHz to 1999MHz memory speeds. If you're running the memory @ 1750MHz to 1874MHz(or anywhere from 200MHz to 1624MHz or 2000MHz+) the timings are the same(so performance will be the same).

Memory timings (Micron MT51J256M32HF)

Your card BIOS 1625MHz: 17-17-22-22-60-21-164 (tighter)
Your friend's card BIOS 1625MHz: 19-19-24-24-65-22-178 (looser)
Advantage = your card BIOS between 1625MHz to 1749MHz memory speeds

Your card BIOS 1875MHz: 22-22-28-28-75-24-206 (looser)
Your friend's card BIOS 1875MHz: 21-21-26-26-70-23-192 (tighter)
Advantage = your friend's card BIOS between 1875 to 1999MHz memory speeds

Anyway...the difference between Baffin PRO/XT and Polaris 21 XT/XL is basically in name only. They wanted to pretend like they weren't doing a full on rebrand so they gave Baffin a new name to throw people off. Same thing with Ellesmere PRO(D)/XT and Polaris 20 XL/XTX. Same chips basically...different name. And people are flashing their Ellesmere cards to Polaris cards successfully(in most cases). So...flashing Baffin cards to Polaris cards shouldn't be a problem either(in most cases).

But...like I said...it probably won't work. Although it might. Or it might even appear to work...yet yield no performance increase. Since additional shaders and TMUs might show(in GPU-Z for example) as being unlocked when they really aren't. And the memory timings thing will definitely affect memory performance at the speeds specified above. And there's also a slim chance you might brick the card. Exceedingly slim though IMO. It's a risk I would be willing to take. But I'm also very confident I could unbrick it if I had to.

Final note: You will have to force flash it because the Device ID and Subsystem ID are different.

What works for you great but you can lead a horse to water but you can't make them drink.

Most who come here seem to not listen thoroughly.

 

[Athlonite]

If it's not the GPU you expected you were buying then take it back to the place of purchase and tell them you thought it was a proper RX560 not the bodged one that this appears to be

 

[MrGenius]

If it's not the GPU you expected you were buying then take it back to the place of purchase and tell them you thought it was a proper RX560 not the bodged one that this appears to be

He's got what's advertised on the HIS site. Nothing "bodged" about it.

The question is when did they stop selling the exact same card with a full CU/shader/TMU count core? Because they obviously did sell them like that at some point.

https://www.techpowerup.com/gpudb/b4739/his-rx-560-green-icooler-oc-4-gb
http://www.guru3d.com/news-story/his-launches-radeon-rx-560-green-icooler-oc-4gb.html

 

[Athlonite]

He's got what's advertised on the HIS site. Nothing "bodged" about it.

What I got from the OP's post was that he'd bought a GPU but it wasn't the one he expected to buy ie: RX560D instead of the RX560 (polaris21) as there's nothing on the HIS site or the box to indicate one from the other as there's no indication of it being either Baffin core or Polaris 21

 

[Kanan]

Now Asus is doing the same, it's new in the news, releasing a new RX 560 with only 896 shaders enabled compared to their other RX 560 which has all 1024 enabled. Seems stock for fully working GPUs is low and they are trying to sell those 896-GPUs again like a generation before with the RX 460. I think the same reason why the RX 460 didn't had all shaders enabled was the same as it is now. This is basically shady marketing tactics, even worse compared to GTX 1060 "3 GB" versions which also have less shaders - but those are at least marked with the "3 GB" whereas these RX 560 aren't marked at all: if you don't look for the shader amounts specifically, you're gonna get screwed.

 

[eidairaman1]

What really needs to happen is that the crippled 560 needs to be just called the 560 and the full-fledged 560 needs to be called to be 565 or 562

 

[Kanan]

The funny thing is, that the RX 560 is supposed to be the full chip, it was released with 1024 shaders enabled, and now they're simply reducing it, in some GPUs, to 896 shaders enabled. They screwed the whole thing.

 

[eidairaman1]

The funny thing is, that the RX 560 is supposed to be the full chip, it was released with 1024 shaders enabled, and now they're simply reducing it, in some GPUs, to 896 shaders enabled. They screwed the whole thing.

The mistake that was made is not calling them 560Ds because those are actually 460s from last year.

 

[Archmage1809]

Yeah I found the different in HIS SKU:
Full Power RX 560: HS-560R4SCNR
Weaken RX 560: HS-560J4SCNR

Well, since I paid a cheaper price (equal amount to the perf loss) for the weaker one, I don't really blame the retailer, they know nothing, almost customers return it to retailer when they found it weakened. They are grey market imported units anyway.

The problem is WHY HIS DO THIS? They don't even marked it as RX 560D.

 

[Frick]

What really needs to happen is that the crippled 560 needs to be just called the 560 and the full-fledged 560 needs to be called to be 565 or 562

I say bring back the LE models!

 

[Kanan]

The mistake that was made is not calling them 560Ds because those are actually 460s from last year.

That's usually the name for mobile products not desktop.

 

[MrGenius]

Actually AMD and NVIDIA typically throw an M(for Mobile) in the name. Either before the subseries numbers(R9 M390X), or tacked on the end(HD 8970M, GTX 980M). I've never seen a D used for such.

In this case D denotes the difference between the 1024SP and 896SP versions of the desktop RX 560 card. The mobile RX 560 is only available with 896SP.

https://www.notebookcheck.net/AMD-Radeon-RX-560-Laptop.194834.0.html

 

[eidairaman1]

https://www.techpowerup.com/gpudb/2990/radeon-rx-560d

 

[Athlonite]

So really all the D stands for in the name is the "Dicked" with edition

 

[eidairaman1]

So really all the D stands for in the name is the "Dicked" with edition

Diminished/disabled

 

[Kanan]

Actually AMD and NVIDIA typically throw an M(for Mobile) in the name. Either before the subseries numbers(R9 M390X), or tacked on the end(HD 8970M, GTX 980M). I've never seen a D used for such.

That's correct, I was actually speaking about the GPUs integrated in the APU, those are named with a D in its name, at least sometimes. This also means it will not be used in desktop GPUs, wouldn't make any sense anyway because it makes not much sense in a PR perspective. AMD's moniker for "super" versions is usually "X", "XT" in the past or even "XTX", but the latter two aren't used anymore and I'm not even sure they will use the X again. They didn't for RX 400/500 series until now, I guess it is a thing of the past, wasn't really good for marketing anyway.

It is also not advisable to name a card with a D to note that it is worse - from a PR perspective, naming the shittier version "1050" and the super version "1050 Ti" is a much better sell. AMD is obviously doing the same. At least in the past. The RX 560 deviates from that. Basically what I already said before.

In this case D denotes the difference between the 1024SP and 896SP versions of the desktop RX 560 card. The mobile RX 560 is only available with 896SP.

That is wrong information as these cards are simply named the same - RX 560 - which was already debated in the thread.

That said, RX 560 was the perfect name for the full GPU this time around, and it got initially released with full shader amount of 1024 too, but they chose to change at least some models later down to 896 shaders, which completely confuses everything. This is basically what I already said at the beginning, strange that I have to repeat myself.