Polaris not so bright [Hardocp]From ATI to AMD back to ATI? A Journey in Futility

[therealnickdanger]

Polaris slower than GP104 is a given. Polaris slower than GP10-anything at 1080p is likely. Given that Hawaii/Fiji seem to perform better at 4K and VR than 1080p relative to Maxwell, my assumption is that Polaris and Vega will be better tuned and suited for 4K and VR than Nvidia's parts. By that I mean that we'll see Nvidia dominate 1080p benchmarks, but that the gap will narrow at 1440p, 4K, and VR.

Of course, all of this is based on speculation of what the target MSRP is...

Why is waiting so hard?

 

[Bryf50]

My sources in the industry have told me that everything in this article is wrong.

 

[raghu78]

this is a pathetic piece of trash written by a guy with big ego and zero brains.


Insulting others is not allowed here. This is a technical forum
Markfw900

 

[DarkKnightDude]

How is Polaris supposed to compete with the 1080 if its in a difference price and performance segment? What is this guy's argument?

 

[b-mac]

What an awful website.

 

[crisium]

Well, he says it runs hotter and slower. Obviously the entire world except him processes that Polaris is designed to be slower.

It depends on what he means by hotter. Is Pascal still more efficient at performance-per-watt? This is not unexpected, although hopefully AMD has cut the gap significantly.

Does he literally mean that full Polaris 10 has more power consumption then 1070 and 1080, while also (obviously) being slower? I guess so, since hotter means hotter.

The literal interpretation of Kyle is that P10 has higher power consumption than GP104.

His sources already have both performance and wattage for Polaris 10, apparently.

 

[Face2Face]

I go to [H] to buy and sell used PC parts...that's it. I never was a fan of their testing methodology.

 

[S.H.O.D.A.N.]

His sources already have both performance and wattage for Polaris 10, apparently.

Which he chose to withheld as far as I can tell (not giving him a click, so feel free to correct me on that one).

Anyone who does the whole 'I know something and you don't lalala' thing is automatically a suspect source.

 

[frowertr]

Wasn't there an article about a year or so ago from Hardocp where they were pissed off at AMD for something else?

This thread will end up locked like the last thread was here at Anandtech, but Kyle sure seems like he is butt hurt about AMD in general. His site is nothing more than a tech blog. Any one who takes HardOCP seriously is just mis-informed.

 

[Rvenger]

Wasn't there an article about a year or so ago from Hardocp where they were pissed off at AMD for something else?

This thread will end up locked like the last thread was here at Anandtech, but Kyle sure seems like he is butt hurt about AMD in general. His site is nothing more than a tech blog. Any one who takes HardOCP seriously is just mis-informed.

These are all opinions being stated and you guys are keeping it civil so far. As long as no "fanboy" arguments come out of it, I will allow it.

 

[flopper]

Kyle isnt professional.
simple as that.

 

[Dribble]

It's mostly about internal AMD politics. Looks like AMD graphics lot are trying to jump ship from the AMD cpu lot, with an aim of getting bought by Intel. If Intel buy AMD gpu that's actually a very good thing for those that want radeon gpu's and a competitor to Nvidia.

Obviously you'd have to actually read the article to notice that - something most people here seem to have avoided doing instead trying to discredit the source as fast as possible...

 

[trane]

The literal interpretation of Kyle is that P10 has higher power consumption than GP104.

Is that even possible? If all they did was blindly ported Hawaii to 14nm, that would still not be the case. Also, we have seen P11 run with 2x perf/W of 950.

 

[Silverforce11]

It could run hotter if Polaris 10 upper SKU is 150W. All that power dissipated through a smaller surface area means the chip runs hotter naturally.

But that's assuming it's 150W. Who knows. Certainly not [H] who have been ignored by AMD.

 

[frowertr]

If the issues Kyle are reporting are true, this is great news for those of us that own stock in Nvidia. Of course this is bad news for consumers of GPU cards...

 

[96Firebird]

Well, he says it runs hotter and slower. Obviously the entire world except him processes that Polaris is designed to be slower.

It depends on what he means by hotter. Is Pascal still more efficient at performance-per-watt? This is not unexpected, although hopefully AMD has cut the gap significantly.

Does he literally mean that full Polaris 10 has more power consumption then 1070 and 1080, while also (obviously) being slower? I guess so, since hotter means hotter.

The literal interpretation of Kyle is that P10 has higher power consumption than GP104.

His sources already have both performance and wattage for Polaris 10, apparently.

Even though nearly all energy consumed by a video card is converted to heat, it can "run hotter" while consuming less power if the cooler cannot remove the heat from the chip as well as it should. In other words, GPU temp (which is usually what is referred to if someone says it is running hot) is different from power consumption, depending on the cooler.

 

[crisium]

Even though nearly all energy consumed by a video card is converted to heat, it can "run hotter" while consuming less power if the cooler cannot remove the heat from the chip as well as it should. In other words, GPU temp (which is usually what is referred to if someone says it is running hot) is different from power consumption, depending on the cooler.

Of course. This is why I have to assume he's using wattage as heat, because the 295X2 runs cooler than the GTX 1080 FE. But that doesn't mean a single GP104 is hotter than 2x Hawaii, not unless you want to be misleading "technically correct". If Kyle doesn't mean wattage than he's lost it more than I thought.

 

[tviceman]

The only significants things I gathered from that article is that AMD is in trouble (nothing new) and Intel might buy RTG. If the latter is true, then that could be a huge boon for PC gaming with Radeon GPU's getting made at Intel's fabs. Then again, by the time that happens it'd be 4+ years and Intel's manufacturing lead will likely have eroded more so than right now.

 

[MagnusTheBrewer]

It's another Kyle being Kyle article. He relishes being the curmudgeon of the tech 'elite' school of journalism. In Kyle's world, performance is the only thing that matters and cost has zero importance. He reports the facts as he knows them but, can't resist injecting large amounts of ego.

 

[tviceman]

Also, perhaps Kyle is also referring to the leaked benches from the other day show Polaris 10 being 33% slower (GP104 being 50% faster), then it will end up as a larger gap than GK104 had over Pitcairn when GK104 came out.

 

[AtenRa]

Well, he says it runs hotter and slower. Obviously the entire world except him processes that Polaris is designed to be slower.

It depends on what he means by hotter. Is Pascal still more efficient at performance-per-watt? This is not unexpected, although hopefully AMD has cut the gap significantly.

Does he literally mean that full Polaris 10 has more power consumption then 1070 and 1080, while also (obviously) being slower? I guess so, since hotter means hotter.

The literal interpretation of Kyle is that P10 has higher power consumption than GP104.

His sources already have both performance and wattage for Polaris 10, apparently.

Temperature and Power consumption is not the same.
You can have low Power consumption but if you use a small heat-sink the Chip will operate at higher temperatures than using a higher power consuming Chip on a water cooler.

Have a look at Fury X vs GTX 1080.

Fury X operates at lower temperatures but consumes more power than GTX 1080.

 

[MrTeal]

Of course. This is why I have to assume he's using wattage as heat, because the 295X2 runs cooler than the GTX 1080 FE. But that doesn't mean a single GP104 is hotter than 2x Hawaii, not unless you want to be misleading "technically correct". If Kyle doesn't mean wattage than he's lost it more than I thought.

That he meant power instead of die temp is the only logical thing to conclude, but that doesn't make it so. I have a 28nm Oland GPU that runs hotter than my old unlocked 40nm 6950 while also being considerably slower, but that's a useless comparison. The 6950 obviously uses a lot more power.

It is possible that the big P10 uses more than 180W, but that would be pretty disastrous for AMD.

 

[Arachnotronic]

Amazing how Kyle gets ripped apart for bringing us news. I bet if it were gushingly positive, he wouldn't be getting the hate that he's getting.

FYI, to all of those saying that his opinion isn't important, apparently AMD tried to pay him off to remove that Roy Taylor story:

AMD's Chris Hook called me and offered an all expense paid long weekend trip out to San Francisco for my wife and I, and then the first exclusive interview with Raja Koduri and he would hold all press so HardOCP was first. I declined the offer.

https://hardforum.com/threads/from-...-in-futility-h.1900681/page-2#post-1042317688

 

[tential]

I go to [H] to buy and sell used PC parts...that's it. I never was a fan of their testing methodology.

I'm a huge fan of their testing methodology but they need to do more 4k max playable settings this generation of gpus or I will not need their site.

 

[crisium]

AtenRa, Mr Teal,

So you believe that's what Kyle is saying? He has inside information that the Polaris 10 fan curve lets P10 get hotter than the 82 degree limit in the 1080 FE? And that the fan curve is the reason he is disappointed - not power consumption?