One Power Hungry GPU: AMD Radeon Vega 56 Revisited in 2020

[neeyik]

Nvidia is still using the older 14nm process node so they can gain a further 20-30% power usage reduction with their 7nm next gen parts.

Turing is 12nm FinFet

Nvidia currently use a specialised 16 nm process node, that TSMC calls 12FFN - it has tighter pitches than the 'regular' 16 nm node, aimed at improved transistor density. So the potential gains going to a tweaked 7 nm node are greater.

Edit: Sorry folks, missed the other comments on this point.

 

[Stainomo]

For all the people who are wondering what is going on with the power consumption.

The Asus Vega cards, both 56 and 64, come with a 260 watt bios and a 220 watt quiet bios.

However something seems off about these results, my Arez 56 hits 62fps at 1440p in the tomb raider benchmark. Latest drivers, stock settings but using a way slower i7 4790 and it is in a pci-e 3.0 8x slot.

Maybe the Vega did have some throttling issue, even with a decent vrm thermal pad I found cooling a problem, now I have added a 2x120mm aio to the core and heatsinks on the vrm components and it has made the card more stable!

In my opinion using a good custom Vega like a Red Devil or a stock Vega would have made this article a lot more useful.

 

[DanGer3]

I'd like to see a 5700xt vs Radeon7 comparison.

 

[HugsNotDrugs]

This article has limited usefulness without including results from a basic undervolt and basic memory overclock.

There really aren't studs or duds between Vega GPUs. An undervolt of 0.1v is a standard starting point and transforms the cards performance and power consumption. Memory overclock of 930mhz is the very low end of what can be expected.

Never in history has there been a GPU that benefits so much from basic tuning. It's like reviewing the Celeron 300a and deciding not to mention there is a 50% overclock potential simply because different people get different results.

 

[HikyBrugh]

What's the point of a Vega 56 when both the RX 5700 & RX 5600 XT are the Vega 56 done right? Vega 56 served its purpose as a "prototype" for the 5700 & 5600 gpus.

 

[CharmsD]

An undervolt of 0.1v is a standard starting point

The default is "as shipped". Graphics cards are drop in devices sold to the general public, they should be no more complex to operate properly than a space heater or a crockpot. Tweakers on a high-horse over modifications are the exceptions here, the Doe's are not.

 

[Puiu]

Then why debate it?



Turing is 12nm FinFet

"Jen-Hsun is still talking about working on node designs with TSMC, there is still the possibility that it’s adapting a 7nm process with the company’s traditional pureplay foundry partner. “In terms of process nodes we tend to design our own process with TSMC,” explains Huang. “Buying an off-the-shelf process is something that we can surely do, but we want to do much more than that.” "​

​

seriously, why the hell do ppl care so much about a naming gimmick? it's like I've hit some sort of nerve.

12nm is just the older "16nm" process with some small tweaks. it's the same as 14nm+ vs 14nm from Intel.

From all intents and purposes calling all of the 16/14/12nm processes as the 14nm process generation is normal and there is nothing to debate.

And that quote has absolutely nothing to do with the discussion, you just put it in because it contains the word "process". I can do that too :

"Our N5 technology development is well on track, with customer tape-out schedule for first half 2019 and volume production ramp in first half 2020. We are already in preparation for N5’s ramp. All applications that are using 7-nanometer today will adopt 5-nanometer. In addition, we are expecting the customer product portfolio at N5 and see expanding addressable market opportunities. We expect more applications in HPC to adopt N5. Thus we are confident that N5 will also be a large and long-lasting node for TSMC."

 

[Steve]

For all the people who are wondering what is going on with the power consumption.

The Asus Vega cards, both 56 and 64, come with a 260 watt bios and a 220 watt quiet bios.

However something seems off about these results, my Arez 56 hits 62fps at 1440p in the tomb raider benchmark. Latest drivers, stock settings but using a way slower i7 4790 and it is in a pci-e 3.0 8x slot.

Maybe the Vega did have some throttling issue, even with a decent vrm thermal pad I found cooling a problem, now I have added a 2x120mm aio to the core and heatsinks on the vrm components and it has made the card more stable!

In my opinion using a good custom Vega like a Red Devil or a stock Vega would have made this article a lot more useful.

We don't use the canned benchmark.

 

[Stainomo]

We don't use the canned benchmark.

Ah did not know that! Thank you the explanation!

 

[Steve]

Ah did not know that! Thank you the explanation!

You can see the section of the game used for testing in the video:

6:30

 

[Stainomo]

You can see the section of the game used for testing in the video:

6:30

Not a huge fan of review videos for things which are mostly benchmarks. However when reading an article I might only look at the graphs and that is how this happend

 

[HugsNotDrugs]

The default is "as shipped". Graphics cards are drop in devices sold to the general public, they should be no more complex to operate properly than a space heater or a crockpot. Tweakers on a high-horse over modifications are the exceptions here, the Doe's are not.

Is your audience not primarily tech enthusiasts? Your writing is certainly aimed as such because it contains information and concepts the general public wouldn't understand.

But the testing in this case seems dumbed down. Vega has such a fascinating story behind it and interesting tuning characteristics that it seems like a missed opportunity.

 

[CharmsD]

Nvidia currently use a specialised 16 nm process node, that TSMC calls 12FFN - it has tighter pitches than the 'regular' 16 nm node, aimed at improved transistor density. So the potential gains going to a tweaked 7 nm node are greater.

Edit: Sorry folks, missed the other comments on this point.

No, thank you for weighing in it's appreciated! I'm just trying to induce others to be on the same page. Since AMD is marking it at 12nm shouldn't we follow since the pitch is greater than on 16nm? [edit, punctuation]

 

[CharmsD]

Is your audience not primarily tech enthusiasts? Your writing is certainly aimed as such because it contains information and concepts the general public wouldn't understand.

Techspot is regularly a top result if not the #1 result for some topic searches, AND in Tech news, making the audience potential world-wide. Readership for this site is ~9MM /month it's not all tech-heads! I try to write in a conversational tone, I am not part of the staff if you were addressing that to me, the little blue badge is an indicator for contributors.

But the testing in this case seems dumbed down. Vega has such a fascinating story behind it and interesting tuning characteristics that it seems like a missed opportunity.

Steve has done other videos on Vega, he's no slouch check out his video list here.

 

[Porkous]

These cards have tremendous value at this point, as gamers tend to mature and actually play less. I would be surprised if the average grown up gamer spends more then 2 hours of maximum game time per day. One of this gards can play you any game at maximum settings. Its is crazy.

The articles should have two bold bullet point on its cases, as right now it seems a bit out of ordinary in those charts, and with this, adds to the complexity of the articles that differ with each publication from different sites.

Take maximum, and average, and minimum. The spikes are through the roof, am rite? What are the averages vs minimum?

Also, please add the description of the benchmark, was it generic, was it in an area with just clear skies, or it was in a crowded place?

 

[tom111]

It works well for workstations but it’s terrible for gaming when energy efficiency is a concern.

Energy Efficiency of a Graphics-Card is irrelevant for a Gaming-System, because it does not run under full Load all the Time. Or are you Gaming 24/7 with 144 fps?

 

[QuantumPhysics]

Energy Efficiency of a Graphics-Card is irrelevant for a Gaming-System, because it does not run under full Load all the Time. Or are you Gaming 24/7 with 144 fps?

Energy efficiency matters to people who actually pay electric bills.

 

[Squid Surprise]

Energy efficiency matters to people who actually pay electric bills.

His point is if you are gaming, the difference is marginal - it only “wastes” energy while at full load - which is probably only a couple of hours a day....

A workstation, on the other hand, can be at full load for hours... and there are probably lots of them... so energy efficiency matters more.

 

[pcnthuziast]

Can't see myself ever getting an AMD card. Always watch for their releases when I'm in the market and in a decade not 1 of their products has ever been a compelling answer for whatever the green team had in the same price range.

 

[tom111]

Energy efficiency matters to people who actually pay electric bills.

And have no Money.
You don't have 20€ per year?

 

[ZackL04]

Hey Steve,

How about a Vega 64 revisit?

They can be found around $230 here in the US and id argue are the best price/performance you can currently buy