Rx Vega hits SiSoft Sandra, 35% faster than gtx 1080.

[yian88]

11 hours ago, themctipers said:

V E G A 

V E G A 

V E G A

 

 

wait shit , i bought my 1070 for $600

 

 

shitshitshitshit aaaa

Its OK, Vega will probably have driver issues, power issues and other bugs. So you actually payed for quality software support aswell not just the card.

[VagabondWraith]

5 hours ago, AluminiumTech said:

I just realized.

 

Both have 56 out of 60(?) SMs enabled. For some reason I thought that the Titan XP was fully enabled....

28/30. 128 CUDA cores per SM* 28 SMs = 3584 CUDA cores. GP100 has 60 SMs where 56/60 are enabled with 64 CUDA cores per SM.. 56* 64 = 3584. The rest are DP cores. GP102 is essentially the gaming version of GP100 minus the DP cores.

[Citadelen]

17 hours ago, Coaxialgamer said:

-snip-

You can't really compare the die sizes between Vega and Pascal, Vega has features like the HBCC and two shader pipelines (one is primitive and the other is programmable) which all take up die space, these have to be programmed for to get the full benefit, which leads to the chip looking as though it's using die space badly.

[themctipers]

1 hour ago, yian88 said:

Its OK, Vega will probably have driver issues, power issues and other bugs. So you actually payed for quality software support aswell not just the card.

I paid $600 to get a DOA card 

 

shitegg took a month to ship it to me, and for me to ship it back to them for them to ship it back to me. 

 

Novideo geeforce experience has crashed more than I can tell you, at stock speeds. It also requires me to sign up for a fucking account, and their UI is 45fps mode. 

 

Im telling you right now that the 1070 cards have stupid voltage limits. From my understanding with my card, it's stuck at 1.06v. Forever. 

nvidia fucking encrypted their bios, and now requires a signature to flash a new bios with pascal. ???

its always hitting its power limits with voltage and consumption, going to 122% while gaming..

 

[Delicieuxz]

2 hours ago, themctipers said:

I paid $600 to get a DOA card 

 

shitegg took a month to ship it to me, and for me to ship it back to them for them to ship it back to me. 

 

Novideo geeforce experience has crashed more than I can tell you, at stock speeds. It also requires me to sign up for a fucking account, and their UI is 45fps mode. 

 

Im telling you right now that the 1070 cards have stupid voltage limits. From my understanding with my card, it's stuck at 1.06v. Forever. 

nvidia fucking encrypted their bios, and now requires a signature to flash a new bios with pascal. ???

its always hitting its power limits with voltage and consumption, going to 122% while gaming..

Some 1070s have unlocked voltages, like the EVGA FTW and Classified versions. I wonder if the FTW BIOS can be extracted from a FTW, and successfully applied to a non-FTW EVGA 1070.

[themctipers]

Just now, Delicieuxz said:

Some 1070s have unlocked voltages, like the EVGA FTW and Classified versions. I wonder if the FTW BIOS can be extracted from a FTW, and successfully applied to a non-FTW EVGA 1070.

i have the regular zotac one (no special name and generic) , so no unlocked voltage for me sadly.

it was only $100 more!

$100 after that, and i could've gotten a 1080 (actually including tax for the 1070, it would be over the price of a 1080 here in Canada)

[RagnarokDel]

14 hours ago, Memories4K said:

This myth needs to die

The Fury X was outperforming the Titan XM at or beyond 5k. I cant find the link but I believe it was on anantech

[Valentyn]

13 minutes ago, RagnarokDel said:

The Fury X was outperforming the Titan XM at or beyond 5k. I cant find the link but I believe it was on anantech

Dunno about 5K, but at 4K it was matching it rather well.

 

 

[Coaxialgamer]

3 hours ago, Citadelen said:

You can't really compare the die sizes between Vega and Pascal, Vega has features like the HBCC and two shader pipelines (one is primitive and the other is programmable) which all take up die space, these have to be programmed for to get the full benefit, which leads to the chip looking as though it's using die space badly.

True, but i could also argue that they are using a denser node. 

[vanished]

11 hours ago, Delicieuxz said:

I don't think the reasoning has to do with its speed, but with efficiency. It sounds to me like Raja is saying that games request and are allocated a lot more VRAM than they actually use (Raja suggests most games don't use more than 50% of what they request and receive allocation for), but that with Vega, the handling of VRAM allocation is improved, so that a game is only allocated exactly what it makes use of, thereby freeing up all the VRAM that has traditionally been tied up in allocation, while not actually being needed by the game that's requested it.

 

 

Well that would be interesting

[DildorTheDecent]

1 hour ago, Delicieuxz said:

Some 1070s have unlocked voltages, like the EVGA FTW and Classified versions. I wonder if the FTW BIOS can be extracted from a FTW, and successfully applied to a non-FTW EVGA 1070.

BIOS editing would be easy. Just copy the clocks and power/voltage table. And flashing the BIOS directly would work too. 

 

Doubt there's much in the way of unlocked GTX 1070s though. Since the there's no 1070 Classy. 1080 would be the better bet for an unlocked card. 

[VerticalDiscussions]

I wanna see when they are releasing a competitor for the Titan lineup that is :P. Vega RX1x2 for 800-900€ and being faster than even the 1080 Ti would be pretty cool, but i guess theyr attention shifted when they said the 395x2 was being aimed at professional usage only, with a very high price.

 

On the other hand, am i the only one who is curious to see if Nvidia could revive the GTX Titan Z, as a Pascal card? Nearly 6000 cuda cores with todays architecture efficiency and clock speeds... would be a crazy beast in SLI

[xAcid9]

10 hours ago, RagnarokDel said:

The Fury X was outperforming the Titan XM at or beyond 5k. I cant find the link but I believe it was on anantech

Yeah, Titan XM is ~30% slower compare to Fury-X at 5k. 

 

10fps vs 14fps   

 

/s

[Drak3]

On 3/13/2017 at 8:02 PM, Memories4K said:

This myth needs to die

@juri-han@Ryan_Vickers

 

Here's the seed of truth to the myth:

 

AMD's HBM implementation on Fiji includes firmware that will attempt to/ refuse the allocation of VRAM if it is not used for data at all. The idea is that the refusal of this unnecessary allocation frees up memory and allows it to be used much more effectively - delivering results that could take more VRAM to achieve on AMD and NVidia implementations of GDDR throughout the years, and sometimes, it works really well.

 

But here the narrative derails from that statement: how much allocation varies greatly. A handful of games can allocate 200% of what they actually need, some might allocate a meager 1% more than they need, and there's a wide range in between. Games that aren't unnecessary VRAM hogs, or don't use more that 4 (and upcoming 8)GB of VRAM AFTER allocation, benefit nearly nothing from AMD's HBM implementation, and that 4GB, and upcoming 8GB, will see the same results of GDDR equivalents when paired with comparable GPUs.

[Tech Inquisition]

On 14/03/2017 at 4:26 PM, zMeul said:

I don't think we'll ever see the GP100 on desktop anytime soon, if ever - it will be a bitch to cool

Introducing... the Quadro GP100 > http://www.nvidia.com/object/quadro-graphics-with-pascal.html#utm_source=shorturl&utm_medium=referral&utm_campaign=quadro-pascal

[zMeul]

47 minutes ago, Tech Inquisition said:

Introducing... the Quadro GP100 > http://www.nvidia.com/object/quadro-graphics-with-pascal.html#utm_source=shorturl&utm_medium=referral&utm_campaign=quadro-pascal

235W ... yeah, that's seriously underclocked to meet thermal stability