AMD once again violating power specifications? (AMD RX-480)

[Majestic]

5 minutes ago, stconquest said:

I have seen Tom's torture test take a Windforce GTX 980 OC (not Ti) and cause it to draw a whopping 260W-280W (not system draw).  That is roughly 100W above most testing at load with that card.

That's because those PCB's allow more TDP to be drawn than the reference boards. Even at 100% power target.

 

Sure, wait for more reviewers to test this as extensively. Wouldn't know who though. PCPer's setup is pretty basic compared to tom's.

[Briggsy]

1 minute ago, Majestic said:

That's because those PCB's allow more TDP to be drawn than the reference boards. Even at 100% power target.

This is correct. If a card has 8+6 pin connectors, that test will pull 300 watts easy with a card.

[DevilishBooster]

1 minute ago, Majestic said:

But no test was deducted on that board, so why is it relevant. You have no idea how the powerlayout is set up...

In order for a product to meet a standard you have to do testing to prove it. That is how standards work. And it is relevant because MSI is powering each 980 (full 980 mind you, not a 980m) through not only a PCIe 16x Gen 3 slot, but also through an MXM slot (which is basically the same thing as a PCIe 16x slot). So therefore, if they are powering each card entirely through a PCIe 16x slot, it shows that you can crank a lot more power through the PCIe interface. And yes, you are correct, I do not know how the power layout is set up, but there are only so many pins in a PCIe 16x slot, and only a small number of them deliver power.

[stconquest]

4 minutes ago, Majestic said:

That's because those PCB's allow more TDP to be drawn than the reference boards. Even at 100% power target.

 

Sure, wait for more reviewers to test this as extensively. Wouldn't know who though. PCPer's setup is pretty basic compared to tom's.

ok.  I agree.

 

I still won't freak out about it.  Overloading the PCI-E slot is not an irrelevant issue.

 

Getting more testing from other sources is really beneficial.

 

 

[MageTank]

2 minutes ago, DevilishBooster said:

In order for a product to meet a standard you have to do testing to prove it. That is how standards work. And it is relevant because MSI is powering each 980 (full 980 mind you, not a 980m) through not only a PCIe 16x Gen 3 slot, but also through an MXM slot (which is basically the same thing as a PCIe 16x slot). So therefore, if they are powering each card entirely through a PCIe 16x slot, it shows that you can crank a lot more power through the PCIe interface. And yes, you are correct, I do not know how the power layout is set up, but there are only so many pins in a PCIe 16x slot, and only a small number of them deliver power.

Not a fair comparison. The GTX 980 Notebook (The real name of that GPU you mention) has a TDP that varies depending on how the vendor configures it. It is normally defined as 165w, but variations of it exist as low as 120w. Running two GPU's off a single PCIE slot is not new. We ITX enthusiasts have been doing this for years with powered risers, and some mobo manufacturers have even made it easier (ASrock has specific bios settings for this on their Z170 boards now). That specific MSI Vortex board is custom designed to supply additional power to those MXM modules. The same way enthusiast overclocking boards provide additional Molex power for PCIE slots. 

 

While I am not trying to disagree with anything anyone here is saying (As far as I am concerned, It is still a simple BIOS fix on these cards) I cannot let you make that comparison as if it is concrete evidence for your claim. 

[DevilishBooster]

6 minutes ago, MageTank said:

Not a fair comparison. The GTX 980 Notebook (The real name of that GPU you mention) has a TDP that varies depending on how the vendor configures it. It is normally defined as 165w, but variations of it exist as low as 120w. Running two GPU's off a single PCIE slot is not new. We ITX enthusiasts have been doing this for years with powered risers, and some mobo manufacturers have even made it easier (ASrock has specific bios settings for this on their Z170 boards now). That specific MSI Vortex board is custom designed to supply additional power to those MXM modules. The same way enthusiast overclocking boards provide additional Molex power for PCIE slots. 

 

While I am not trying to disagree with anything anyone here is saying (As far as I am concerned, It is still a simple BIOS fix on these cards) I cannot let you make that comparison as if it is concrete evidence for your claim. 

MSI did not use the notebook GTX 980 in the Vortex, they used the full size, full power, desktop GTX 980.

 

EDIT: by "full size" I mean the full size dye on a custom PCB, not the full length card.

[MageTank]

1 minute ago, DevilishBooster said:

MSI did not use the notebook GTX 980 in the Vortex, they used the full size, full power, desktop GTX 980.

No. The specific SKU is called the "GTX 980 Notebook Graphics". It's not the 980m. It's a desktop 980 in MXM form. 

 

http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-980

 

Confusing? Yes. Not exactly a fan of Nvidia's nomenclature. 

[DevilishBooster]

2 minutes ago, MageTank said:

No. The specific SKU is called the "GTX 980 Notebook Graphics". It's not the 980m. It's a desktop 980 in MXM form. 

 

http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-980

 

Confusing? Yes. Not exactly a fan of Nvidia's nomenclature. 

I already edited my last post. The PCB is just in the MXM form factor, it still has all the same power draw of a GTX 980 that is sitting in hundreds of thousands of desktops right now. It is not a cut down chip.

[Curufinwe_wins]

4 minutes ago, MageTank said:

No. The specific SKU is called the "GTX 980 Notebook Graphics". It's not the 980m. It's a desktop 980 in MXM form. 

 

http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-980

 

Confusing? Yes. Not exactly a fan of Nvidia's nomenclature. 

 On a different custom PCB with specifically validated power draw techniques...

 

I think either way that is much less of a big deal than a situation where the default configuration on a least common denominator situation draws much more than spec. All laptop pcb's have to be specifically made for the hardware anyways.

 

I doubt highly that it will ever cause issues with people using modern MOBO's, but esp being a mainstream/budget platform, you really need to be following spec (I know people who are upgrading on c2q systems and they may very well be SOL with this issue. We all know how much trouble older mobo's had with pcie power draw requirements.)

[MageTank]

4 minutes ago, DevilishBooster said:

I already edited my last post. The PCB is just in the MXM form factor, it still has all the same power draw of a GTX 980 that is sitting in hundreds of thousands of desktops right now. It is not a cut down chip.

They do make different versions with varying TDP's. 

 

 

Much like the memory configurations on MXM SKU's differ, so does the design of the entire module itself. Vendors are given a lot of leeway when it comes to these modules. I have seen them as low as 120w, and as high as 200w. While i do not expect you to take my word for it, there is plenty of resources on the internet about this specific GPU SKU to validate my claims. You are correct for the most part though. Most of them are indeed 165w, like the desktop 980. However, I would need to see the 980 MXM module in the Vortex to be certain it was not modified by the vendor.

[laminutederire]

by the way, I've just had a closer look at their workstation picture:

the backplate of the card says r9 390... isn't it a bit weird?

[GzeroD]

I think that is just a stock picture of their test setup.

[Curufinwe_wins]

5 minutes ago, MageTank said:

They do make different versions with varying TDP's. 

 

snip

 

This is a render, you'd have to go into the LTT teardown for an actual image, but it does look custom to me.

[DevilishBooster]

6 minutes ago, MageTank said:

They do make different versions with varying TDP's. 

-pic snip-

Much like the memory configurations on MXM SKU's differ, so does the design of the entire module itself. Vendors are given a lot of leeway when it comes to these modules. I have seen them as low as 120w, and as high as 200w. While i do not expect you to take my word for it, there is plenty of resources on the internet about this specific GPU SKU to validate my claims. You are correct for the most part though. Most of them are indeed 165w, like the desktop 980. However, I would need to see the 980 MXM module in the Vortex to be certain it was not modified by the vendor.

Well, I have never claimed to be a GPU expert, but everything that I have read on the Vortex says they are full power 980s. Here is a pic so you can let me know what you think, and I put the link to the article that image is from below. I'll try to find out more information. I'm not saying you're wrong, I'm just telling you what I have read/know. The point still stands that they are sending waaaaaaay more than 75W through the PCIe/MXM interface.

http://www.gizmodo.com.au/2016/06/msi-vortex-g65-sli-gaming-pc-australian-review/

[NumLock21]

Msi vortex uses a desktop type gtx 980 with a mxm interface. Not a mobile gpu. Nvidia says laptop is due to the mxm interface cause desktop do not have mxm.

[Curufinwe_wins]

5 minutes ago, NumLock21 said:

Msi vortex uses a desktop type gtx 980 with a mxm interface. Not a mobile gpu. Nvidia says laptop is due to the mxm interface cause desktop do not have mxm.

MXM 980 is technically a desktop-type 980 as well... But yes even more than normal the Vortex uses much more of a custom desktop style pcb layout and clock speeds.

[Doobeedoo]

Now I'd like to see custom cards, how much will they differ.

[Dionyz]

"

1) The RX 480 meets the bar for PCIe compliance testing with PCI-SIG. //edit: and interop with PCI Express. This is not just our internal testing. I think that should be made very clear. Obviously there are a few GPUs exhibiting anomalous behavior, and we've been in touch with these reviewers for a few days to better understand their test configurations to see how this could be possible.

2) Update #2 made by the OP is confused. There is a difference between ASIC power, which is what ONLY THE GPU CONSUMES (110W), and total graphics power (TGP), which is what the entire graphics card uses (150W). There has been no change in the spec, so I would ask that incorrect information stop being disseminated as "fact."

We will have more on this topic soon as we investigate, but it's worth reminding people that only a very small number of hundreds of RX 480 reviews worldwide encountered this issue. Clearly that makes it aberrant, rather than the rule, and we're working to get that number down to zero.

/edit for absolute factual clarity. 

https://www.reddit.com/r/Amd/comments/4qfwd4/rx480_fails_pcie_specification/d4sy0c3

"

 

Then the counter argument is listed https://www.reddit.com/r/Amd/comments/4qfwd4/rx480_fails_pcie_specification/

[laminutederire]

32 minutes ago, GzeroD said:

I think that is just a stock picture of their test setup.

Some of these power draw actually are in the range of a 390 though...

I've re read 390 480 and 390X reviews to help with my GPU choice, and that 300W is actually what I've seen in some 390 reviews.

That'd be sooooooooooooo ridiculous, but who knows what happens there?

Anyhow, we'll see what AMD has to respond with enough time to work on that themselves, they seem genuinely surprised of those results.

[NumLock21]

25 minutes ago, Curufinwe_wins said:

MXM 980 is technically a desktop-type 980 as well... But yes even more than normal the Vortex uses much more of a custom desktop style pcb layout and clock speeds

I'm saying mxm is a laptop interface No desktop has mxm. Not the gpu. You can have 2 mxm cards one using a desktop gtx 980 and the other a mobile gtx 980.

They mention laptop is from the mxm, don't want those buying the gtx 980 mxm card for their desktop and then finding out it won't work, so they just wrote "laptop".

Like ssd upgrade kits for laptops and desktops. Both have the same ssd, to avoid confusion from the consumer, they have different accessories depending on the kit you buy.

[Curufinwe_wins]

3 minutes ago, NumLock21 said:

I'm saying mxm is a laptop interface No desktop has mxm. Not the gpu. You can have 2 mxm cards one using a desktop gtx 980 and the other a mobile gtx 980.

They mention laptop is from the mxm, don't want those buying the gtx 980 mxm card for their desktop and then finding out it won't work, so they just wrote "laptop". 

Sure, but full gm204 is available in other laptop mxm versions. 

[NumLock21]

10 minutes ago, Curufinwe_wins said:

Sure, but full gm204 is available in other laptop mxm versions. 

Even if there is, it's still the same gm204. They can have mxm or something crazy like m.2.

Speaking of mxm there are 3 types. Not sure with this msi one, uses a new kind. But the newest was mxm type III.

[VagabondWraith]

I think, at the end of the day, right or wrong, everyone can agree this is a "botched" (can't think of the right term) launch and a black eye on AMD... especially in a moment where they can least afford it. Such a shame really...

[Curufinwe_wins]

8 minutes ago, VagabondWraith said:

I think, at the end of the day, right or wrong, everyone can agree this is a "botched" (can't think of the right term) launch and a black eye on AMD... especially in a moment where they can least afford it. Such a shame really...

I mean, honestly this doesn't mean a damn thing for people running modern motherboards, but yea it's pretty shitty for a decent bit of the market this card is for.

[RagnarokDel]

2 hours ago, Majestic said:

guess ill have to post it again

 

 

That 155W WAS the mainboard powerspike.

There's no way in hell that's accurate. Just on the temperatures alone, this would make the card meltdown.