AMD: Speculation, Rumors, and Discussion (Archive)

[silent_guy]

I'd really have to look into the specs for the memory controllers more, but in theory it would be advantageous to develop a controller that could support them all.

That's not really necessary: the core transaction scheduling is the almost the same for all anyway. The only difference is the PHY side. Almost all memories system implantations that I've seen in there last decade separate the scheduler from the PHY.
From a scheduler point of view, a single HBM stack would look like bunch (8?) GDDR5 chips. Nothing else.

The base die for HBM could also integrate part of the controller or possibly a minimally active interposer.

That would make very little sense.

Power/latency of the HBM should be far easier to drive than the GDDR5.

The latency of all DRAM technologies of the same generation is essentially the same.

 

[CSI PC]

Really a GTX 970 to a GTX 980 same chip but the 980 is clocked higher with more parts enabled. What's the power difference when running a game? Well, looking at Anandtech in Crysis 3, it's ~4 watts difference.

How about something closer to mid-range. Using a GTX 760 versus a GTX 770. Again according to Anandtech there's a 17 watt difference.

Unless they are pushing it far past the knee of the power curve (like Fury X) there's not going to be anywhere near a 25+ watt difference. There's likely to be a 5-15 watt difference between the 470 and 480.

BTW - I was going to use the 270 versus 270x for AMD but at Anandtech the slower 270 actually consumed more power than the faster 270x.

I think Dave mentioned somewhere before that salvage chips used for lower card tiers often had worse power characteristics than chips used for the intended top end version. Meaning they often had worse perf/watt despite sometimes consuming less power. Or was it that there was far more variability with higher potential for lower perf/watt? It's been a while since I read that post.

So while unlikely, it's possible for 470 to consume more power than 480.

Regards,
SB

Yeah.
Worth noting one other aspect that will add to the roughly 15W difference will be the 470 being 4GB and the the 480 being 8GB and also with higher speed ram (maybe lol).
I appreciate there is also a 480 with 4GB, but the potential of this card really should be considered as primarily 8GB model (just like 390), especially when looking to run latest and future games with decent settings or resolution.

Cheers

 

[silent_guy]

mini-rant:
It doesn't make much sense to have no less than 3 threads about Polaris..

I don't disagree, isn't the first poster of a thread the one who created it? IOW: you?

Edit: Ok, mod spawn. Carry on!

 

[Anarchist4000]

From a scheduler point of view, a single HBM stack would look like bunch (8?) GDDR5 chips. Nothing else.

The idea I was getting at was take a portion of the controller and be able to up the frequencies to properly interface with GDDR when an interposer isn't present. Use one chip for multiple designs instead of respinning it with a different memory controller.

That would make very little sense.

Idea here being channel bonding or the ability to remap the lanes to some degree. If it was even required, but it's been a while since I've messed with memory designs.

The latency of all DRAM technologies of the same generation is essentially the same.

The difference being the proximity to the controller. Driving a signal over a matter of a few mm instead of cm. Part of why HBM uses less power to start. It should buy you a little headroom being closer without complicated clocking for the HBM.

 

[sir doris]

I don't disagree, isn't the first poster of a thread the one who created it? IOW: you?

Think it was spawned by a MOD.

 

[AlNom]

Could have been Red Kryptonite.

 

[Ryan Smith]

A non-serious(?) answer would be that none of them got the slides they now have permission to disseminate.

Confirming that I didn't get the deck until this afternoon. I'm assuming here that AMD had some on-site briefings at E3 for those press who were in attendance.

 

[xpea]

I'm a bit speechless. Poor performance overall. Waiting for independent reviews because I don't see where goes the "biggest CGN change in history" AMD was hyping...
Otherwise, very annoying to see Polaris 10 inside RX 480 but Polaris 11 inside RX 480M. bad AMD, bad

 

[xpea]

Confirming that I didn't get the deck until this afternoon. I'm assuming here that AMD had some on-site briefings at E3 for those press who were in attendance.

seriously ?
Won't say bad things because they are people with jobs behind it, but damn, AMD marketing team is so [censored]
well, at least they are consistent in their mediocrity

 

[Orion]

I'm a bit speechless. Poor performance overall. Waiting for independent reviews because I don't see where goes the "biggest CGN change in history" AMD was hyping...
Otherwise, very annoying to see Polaris 10 inside RX 480 but Polaris 11 inside RX 480M. bad AMD, bad

Poor performance? latest leaked 3dmark scores put it above an overclocked 980 and even a nano and that is reference boards using 100w with expectations of boards with 8 pin connectors, two 6 pin connectors, and even a 6 and 8 pin connector iirc.

The 299$ options should probably match or exceed 1070 in certain dx12 applications.

 

[gamervivek]

That's wccftech and quite incongruent with the scores given by AMD themselves.

 

[Orion]

That's wccftech and quite incongruent with the scores given by AMD themselves.

AMD themselves? did they break NDA schedule?

According to some of the comments the provider of the bench has history of being questioned and later proving being legit.

 

[SimBy]

So wccftech decided their source is more reliable than AMDs own numbers, interesting Not saying those numbers are impossible on custom OC models though. But I highly doubt this are reference cards.

 

[Orion]

So wccftech decided their source is more reliable than AMDs own numbers, interesting

Numbers on 3dmark firestrike? 480? perhaps you mean leaks? if it is leaks it is easy simply by forcing it to run at base clock to lower the score.

Or do you mean amd has done some benchmark pr with the 480?

PS

if some of those benches linked are correct, it seems the 480 sli > 1080 sli in ashes.

 

[SimBy]

Numbers on 3dmark firestrike? 480? perhaps you mean leaks? if it is leaks it is easy simply by forcing it to run at base clock to lower the score.

Or do you mean amd has done some benchmark pr with the 480?

AMD provided Steam VR test score of 6.3. That's stock 290 level. I really can't see how you get such a huge discrepancy between high Firestrike wccftech score and low Steam VR score. Makes 0 sense.

So either AMD is taking us all for fools, wccftech source is unreliable or this are actually legit scores but with superclocked custom cards.

 

[Orion]

AMD provided Steam VR test score of 6.3. That's stock 290 level. I really can't see how you get such a huge discrepancy between high Firestrike wccftech score and low Steam VR score. Makes 0 sense.

well if these are real(probably a big if), two possibilities, something funky in the way steam calculates their score or perhaps amd used base clock.

 

[xpea]

all official AMD sliders here:
http://www.hardware.fr/news/14685/radeon-polaris-rx-470-rx-460-apres-rx-480.html
French article, but English slides are at the bottom of the page

 

[Frenetic Pony]

So, which is it, 150 watts or 100? (And why is there such a range?)

If it's drawing 100 while running 3dMark Firestrike Ultra the RX 480, normalized to the 1080's tdp of 180 watts, gets a benchmark of 6,000 while the 1080 gets 5,000. If it's pulling 150 watts then it's a normalized score of 4,100. Why is there such a huge range, and even how, and is it PR bullshit? I dunno, but there's the math of it.

 

[xEx]

mini-rant:
It doesn't make much sense to have no less than 3 threads about Polaris.. where one is about Polaris in general, the other is about one Polaris 10 (RX 480) and the other is about one Polaris 10 and one Polaris 11 (RX470 + RX460).

Think it was spawned by a MOD.

Yes it was. I have no idea why since we dont see a thread for 970, 980m980ti.1070,1070FE, etc....

 

[xEx]

So, which is it, 150 watts or 100? (And why is there such a range?)

If it's drawing 100 while running 3dMark Firestrike Ultra the RX 480, normalized to the 1080's tdp of 180 watts, gets a benchmark of 6,000 while the 1080 gets 5,000. If it's pulling 150 watts then it's a normalized score of 4,100. Why is there such a huge range, and even how, and is it PR bullshit? I dunno, but there's the math of it.

its TDP is 150W. the card itself consumes 100W of electricity while gaming.