AMD RX Vega Shown Against GTX 1080 at Budapest Event

Yes, because its basically the same thing.
If IHV is telling use that their product's TDP is 400W, that means that it can dissipate 400W, which means it can accommodate GPU with 400W of power consumption.

So unless IHV has purposely decided to shoot themselves in the foot by overestimating TDP(*), TDP=power consumption. <- my point
* this does happen in some cut down lower chip versions, but it's just a simplification. Overestimating TDP does not happen with major products, when the eyes of the press and everyone else is glued on TDP.
Or sometimes IHV can even cheat, and communicate lower TDP to the press, but that's dishonesty, and I am talking about the proper, the best attempt at TDP estimate.


Ingame power consumption(averaged across 6-7 reviews) vs TDP. See the trend?

 

Just wanted to add something positive to this thread... +1 on Romanian Babes!

Back on topic... did anyone notice the 2 and 3 fan versions of the Asus GPUs? I hope those were not RX570 or 580s.

Kinda sad that all the blokes were on their mobiles. At least snap more media friendly pictures.

 

I have to confirm it... those girls were very hot indeed... ah Vega? well AMD promise a 10000HP Dragster and we are getting a very fast Import, it is nice, it runs fast, it have a great design (AC version)and it is a little cheaper. But in the end it doesn't take the crown (or at least that's we have seen till now). I'm an AMD kind of guy, if i could i would buy one just because it is not Nvidia but we have to accept that they don't reach the gold of the single chip fastest GPU in the world (and if Vega is as hot as those babes well dual GPU looks hard also)

 

And you think they are Romanian because...? :3eyes:

 

Note taken, thanks. Our language uses "fuze" as fusion for generic nuclear reaction applying to both fusion/fission. That's unfortunate consequence of improperly taken word from other language.
Our Czech Terms do actually match their English counterparts.

@Noisiv: I kindly pointed out that he intentionally uses word "Heater" for AMD's GPU. With intent to degrade it in people's eyes.
It is same as if I wrote: "Why is this dummy always questioning simple things and taking them from original context?"
Substituting proper name/description is not very cool, right?

And please note that he wrote: "...healing our planet with 400W heater that perform the same as the competition 200W gpu"

My reply was pretty clear. 400W heater does its job better than 200W GPU.
Why? Because Job of heater is making heat. So, yes, 400W heater does easily produce double amount of heat than 200W GPU.

So, what was your long analysis of heat generated by 400W heater vs 400W GPU for?
Nothing. Because you created "counter argument" based on shifted premise. You know how that is called. And basically once shifted back to 400W vs. 200W, it confirms what I wrote.

 

@Foxx

Don't take everything personally.
All I did is wasted some bandwidth to bust the old myth of TDP .NE. Power Consumption.
Which gets too often repeated on these pages.

Yet our Hilbert says "TDP" when talking about power measurement, and rightfully so.
These two terms are often used interchangeably for a very good reason.

 

YAWWWWWWWWWWWWWN!!!! Waiting for a very long time, and still waiting, Benchmarks hopefully soon, I am going back to sleep, .

I believe the HBM2 is what holding the card's release date, if it were gddr5 or the x variant, It probably would been released for awhile now.

 

Unless you are from France or other European countries, most of electricity from anywhere else in the world is still generate from coal or natural gases, even nuclear reactors still generate nuclear waste. How in the world a less efficient product is saving the planet is beyond me. As an B787 pilot myself I'm pretty sure we are moving towards better efficiency machines, not backwards.

 

Since you wanted people to know what TDP is and that they may use it improperly... Why not link wiki? Wiki Quote 1st sentence:

I am pretty sure everyone can understand from this, that TDP is property of cooling solution.
Only connection this has to actual chip being cooled is T-Max of given chip.

 

Wow, people sure are getting pessimistic. This GPU is not likely to consume 400W, even when OC'd. That is too far beyond the specs of PCIe and PSUs, and it is within AMD's interest to stay below them.

Regardless, I said this before and I'll say it again - AMD hardware is weirdly power-hungry during stress tests (like FurMark), but under real-world applications, the gap in power consumption compared to a competing Nvidia GPU closes dramatically. The AMD card is still definitively more power hungry, but not the 100% increase that some people insist on.

 

What? 1st of all everyone knows (should know) that particular TDP definition.
That much I assume. If someome is clueless about meaning of TDP, my TDP myth busting effort is not aimed at him.

2nd... Stay with the wiki please. Instead of coming with partial remarks and non-sequiturs at the end of the discussion, like nothing has been said so far.

Is there anything in my "long analysis" that you disagree with?

 

My bios moded Titan X maxwell can draw up to 400W and still stable (pcie slot configured to draw up to 66W only), and that is on a 6 pin and 8 pin power connectors, it's not that out of spec for VEGA with 2x8 pins pcie connector to draw 400W

 

Well assuming RX Vega is similar to FE, it will use roughly ~280w out of the box in actual games. PC Per upped the power limit by 20% and peak clock by 7% which resulted in a 350w load in Uniengine's Heaven benchmark. There is an optimal efficiency point that AMD seems to be running the card on the brink of.

I've wrote this before, but the other issue is that Vega is going to be soon up against Volta - which has a 50% perf/w increase according to Nvidia. So theoretically Nvidia could build a card with 1080 performance @ 120-140w. In fact if they prioritized it, they could have it out now - they'd just rather reap the lucrative margins of $10,000+ GV100 chips. So unless AMD finds a way to radically reduce the power consumption, or increase performance, or both, it's going to have a problem competing. Die size is also an issue. Vega is ~480mm2 - 1080 is only ~320mm2. Scaling this architecture is going to be more difficult for AMD come next round.

 

Your comparison of 400W heater vs 400W GPU is correct.

But I disagree with your twisting 400W Heater vs. 200W GPU into 400W heater vs. 400W GPU. (Guess who were you quoting )
Because they have very different conclusion. And note that my post with those 400/200 was reaction to someone too.

And I can't even begin with:
"Q: What is the difference between a GPU which consumes 400W and a 400W TDP GPU?"
Since your post about "cluelessness" about meaning of TDP, I guess you just worded it differently than you meant it.

Rest of your 400W TDP this and that discussion with others really missed me, it is meaningless from all sides.
I look at GPU paired with 300W TDP cooling solution running sub 55°C, then I look at GPU paired with 250W TDP cooling solution running 75°C.
What do I make of it? mentioned chip being cooled down to 55°C has much higher thermal (Power consumption) headroom, before it saturates its cooling solution.

 

Maybe I should have called VEGA a 200W heater in the begining since it does have the same perf as competing 200W GPU and chew an extra 200W, my bad.

 

What if I called you by byproduct of your elaborate inner workings too? No, I am not into fecal matters.

But impression it leaves is same. You say that it is best at making heat. And rendering is just byproduct of this function.

 

@Fox2232

400 vs 200 turned into 400 vs 400 <--- I wasn't twisting anything. I merely used your post as jumpboard for my personal agenda But apparently you're the guy who perceives being attacked, just from being quoted
I didn't say you were wrong, or that your Fury X sucks bajillion Watts WTF...


You get that when AMD/Nvidia says Volta/Navi is 250W TDP they are referring to cooling solution, yes? OK
What I am saying that if estimated honestly, that number of Watts (TDP) is not so accidentally(!!) equal to maximum sustained power consumption.

What are the alternatives?

• They have retarded marketing department which purposely overestimates needed TDP solution <--RIGHT. THAT WILL BE THE DAY!
• They are falsely advertising by underestimating TDP needs. YUP. HAPPENS NOW AND THEN. Nothing dramatic though. And being dishonest or erroneous does not change the actual TDP needed.

OK?

 

Using his logic my R9 290 is 180W TDP, because that's just about how much its reference cooler can dissipate.

Look ma, I got 180 Watts Hawaii
EDIT: Actually I did. Because Wattman undervolting finally works for me :infinity:

 

That wattage is more what I was expecting. And sure, that's still high, with or without comparing to a 1080, but it doesn't warrant the amount of whining we're getting.

That I find to be a much more glaring comparison. AMD is a bit overdue for an architecture refresh; perhaps that would be what it takes to push them to do so. This of course is assuming Nvidia isn't cherry-picking results, which they could be.

Why exactly does die size matter? If anything a bigger die is better due to more surface area and therefore better heat dissipation. But also, a good chunk of Vega's die is the HBM, which the 1080 doesn't have, so they're not really an apples to apples comparison.

 

Yeah. I considered your quote as sort of counter argument. Then it is OK, since you meant it as disconnected.

As for TDP vs GPU power consumption. In Case of HD7970 reference (blower), it was rated as 250W TDP cooling. GPU itself ate around 185~205W depending on leakage.
All of GPUs reached around 80°C.
Compared to that Fury X with 275W TDP and GPU which by default is limited to ~275W keeps at lowly 50~55°C.

TDP does not have strong correlation with actual power consumption or final temperature. Other than it should be higher than actual Power draw of device being cooled.
And if you ask me, Fury X has cooling with TDP which is at least 375W.
That incoming RX Vega with liquid cooling may have pretty much same cooling performance.
If GPU itself is limited to same 275W or 350W, I can't tell.