[Rumor (Various)] AMD R7/9 3xx / Fiji / Fury

[Stuka87]

So how trustworthy are these benchmarks from videocardz? I for one am not putting much faith in them. I just don't understand why with all the previous benchmark leaks that nobody "believes" anymore, its all of a sudden fact with the latest benchmark of an unreleased GPU.

Anyways, I'm so excited for the 16th!

They are generated, so not trust worthy at all. No actual test was run.

 

[IEC]

still kinda afraid that they didn't even have the card booting 2 weeks ago. that means no launch in June IMHO

The alleged sample had no BIOS on it simply to discourage theft and consequent leaking of benchmarks.

 

[MrTeal]

The alleged sample had no BIOS on it simply to discourage theft and consequent leaking of benchmarks.

It sounds weird, but I suppose theft could happen there. Google News had a story today that the manuscript for the sequel to the 50 Shades of Grey book got stolen. I'm sure if someone was really ballsy and stupid they might try to swipe the card from Computex. No bios solves that potential problem quite easily.

 

[cen1]

Yeah, no way that card was not bootable 2 weeks ago. It was probably in a working state for months.

 

[el etro]

This whole quoting and use of Furmark as ANYTHING useful, other than maybe power supply stability, really just shows how uninformed the person is, and that they probably have no clue what they're talking about.

This.

 

[tential]

This whole quoting and use of Furmark as ANYTHING useful, other than maybe power supply stability, really just shows how uninformed the person is, and that they probably have no clue what they're talking about.

I don't even get why peak power consumption is quoted. All I care about is average power consumption during gaming.
Peak draw is only useful as a omg just in case my psu should be able to handle this load incase something happens. I'm not even sure because I am not knowledgeable 100% confidence about the subject. I just don't use furmark or any benchmarking tool as a measure of performance because I don't benchmark daily. I run games daily. If benchmarks were closer to gaming performance I'd use them.


I used benchmark to mean benchmark software as I'm on mobile and too lazy to fledge out sentence.

 

[dzoni2k2]

Why would they send a non working WCE Fiji to repi. Lets not be silly.

https://twitter.com/repi/status/601739457960763392

 

[Zstream]

still kinda afraid that they didn't even have the card booting 2 weeks ago. that means no launch in June IMHO

I think it's more of a deterrent for leaks and what not. I'm honestly not sure how AMD has managed to keep a lid on everything, but I think it's really cool.

Either way, I don't see myself upgrading from my trusty r9 270 anytime soon. I"m just in the minority that thinks it's cool that AMD is releasing HBM and a water cooling solution at the same time.

 

[Glo.]

Same TitanX 3D Mark means better game performance.

And as for R9 390X, if they have the same 3dMark Extreme score as the GTX980 it also means better gaming performance.

 

[Creig]

 

[DarkKnightDude]

If that's Fiji, really hits me how small it looks compared with my 770.

 

[Skurge]

http://videocardz.com/56236/amd-radeon-fury-x-pictured-some-more

 

[soccerballtux]

The alleged sample had no BIOS on it simply to discourage theft and consequent leaking of benchmarks.

sweet. seems legit

 

[.vodka]

Hmm, the Fury card in that link hasn't got the tubes braided.. It also seems the part that's silver colored in the braided card is black in the one that isn't, it doesn't look like a super reflective surface that acts like a mirror, for example to take a silver color in the braided card.

Maybe different versions of watercooled Fiji? Or maybe a sample for checking performance, that doesn't have the finishing touches...


The size reminds me of a 7600GT I had a while back. Crazy to see how HBM has helped reduce the PCB size.

 

[Creig]

I can't wait to see a teardown so we can get a look at that waterblock. GPU, VRM's, memory. Everything should be covered. Lower GPU and VRM temps should help contribute to lower power consumption.

 

[jamesgalb]

I find it funny that all these videocardz.com leaked benchmarks we see lately always include an OCed 980ti and OCed 980, and no other OCed cards, in order to show more NVidia cards at the top.

I take their AMD coverage with a grain of salt. They seem to be trying to de-mystify the AMD launch imo.

 

[swilli89]

I think it's more of a deterrent for leaks and what not. I'm honestly not sure how AMD has managed to keep a lid on everything, but I think it's really cool.

Either way, I don't see myself upgrading from my trusty r9 270 anytime soon. I"m just in the minority that thinks it's cool that AMD is releasing HBM and a water cooling solution at the same time.

Nah, you're not in the minority. People with some weird fear or reason to crap all over this launch are the loudest. Many many people are excited and silently lurking/reading waiting for reviews. Fiji is freaking exciting!

 

[utahraptor]

Hmm, the Fury card in that link hasn't got the tubes braided.. It also seems the part that's silver colored in the braided card is black in the one that isn't, it doesn't look like a super reflective surface that acts like a mirror, for example to take a silver color in the braided card.

Maybe different versions of watercooled Fiji? Or maybe a sample for checking performance, that doesn't have the finishing touches...


The size reminds me of a 7600GT I had a while back. Crazy to see how HBM has helped reduce the PCB size.

I believe they slightly alter each sample to help identify leaks. The position the screw heads are rotated, the location of the plastic ties on the braiding, type of braiding, cosmetic flaws ect...

 

[.vodka]

Ah, right. Yes, that would be an excellent way to identify the leaker. Heard AMD does it that way before... I didn't recall that a few posts back.

 

[Cookie Monster]

I wonder if the dual 8pin power connectors are a cause for concern..

However Im quite looking forward to this launch! A little less under a week to go.

 

[tviceman]

I can't wait to see a teardown so we can get a look at that waterblock. GPU, VRM's, memory. Everything should be covered. Lower GPU and VRM temps should help contribute to lower power consumption.

Except it didn't with the 295x2.

http://www.hardocp.com/article/2014/04/08/amd_radeon_r9_295x2_video_card_review/12#.VXj1Qc9VhBc

 

[Silverforce11]

I can't wait to see a teardown so we can get a look at that waterblock. GPU, VRM's, memory. Everything should be covered. Lower GPU and VRM temps should help contribute to lower power consumption.

If the water block covers VRMs as well, would be great. I think it should since its fanless on the card.

R295X2 has a small fan on the card to blow over the VRM area, so it was aircooled and probably very hot due to the power required.

 

[tviceman]

I find it funny that all these videocardz.com leaked benchmarks we see lately always include an OCed 980ti and OCed 980, and no other OCed cards, in order to show more NVidia cards at the top.

I take their AMD coverage with a grain of salt. They seem to be trying to de-mystify the AMD launch imo.

When it comes down to it, between Fury X and 980 TI, whichever has the best overall performance after OCing will be the card to get for prospective high end buyers. Since Fury isn't out yet, OC scores aren't available. It's nice to see 980 and 980 TI OC scores so we know how much Fury will need to stretch it's legs.

 

[TechyGeek]

I wouldn't buy Fury for 4k. Waiting for 400 series.

 

[RussianSensation]

I don't even get why peak power consumption is quoted. All I care about is average power consumption during gaming.
Peak draw is only useful as a omg just in case my psu should be able to handle this load incase something happens. I'm not even sure because I am not knowledgeable 100% confidence about the subject. I just don't use furmark or any benchmarking tool as a measure of performance because I don't benchmark daily. I run games daily. If benchmarks were closer to gaming performance I'd use them.


I used benchmark to mean benchmark software as I'm on mobile and too lazy to fledge out sentence.

That's a very good point too. There could have been a spike in 1 section of the review. I should also reconsider using Peak alone (perhaps show avg + peak as both points of reference) because as you said during gaming the GPU isn't loaded to 99-99.9% which means the Peak doesn't accurately represent what amount of power the graphics card actually uses over the entire gaming session. I am guilty of posting Peak usage myself so thanks for providing good constructive feedback about why certain methodologies posted on the forums may have their own weaknesses. I feel it's good to have these discussions out on the forum since after-all we are in this hobby together :thumbsup:

Except it didn't with the 295x2.

Good point; that was one of the key weaknesses of that design. This time with AMD bringing HBM closer to the die on the interposer, I wonder if it's possible for them to have a mini water-block of sorts?

That backplate looks too plain imo. I find it odd that it has no holes to dissipate heat and that it doesn't even have any logos or anything.

When it comes down to it, between Fury X and 980 TI, whichever has the best overall performance after OCing will be the card to get for prospective high end buyers. Since Fury isn't out yet, OC scores aren't available. It's nice to see 980 and 980 TI OC scores so we know how much Fury will need to stretch it's legs.

I still have my fingers crossed that AMD prices Fiji PRO at $499. It might be the sleeper out of all of them. Based on rumoured specs, it has identical ROPs, memory bus/HBM memory to the flagship card, with the differences coming down to 14% less shaders/TMUs and lower clocks. At that point if Fiji XT is $699 and Fiji PRO is $499, I am not entirely convinced 14% more performance in overclocked states (assuming the PRO and the XT overclock similarly) is worth the paying hundreds of dollars extra estimating how much more performance 14nm/16nm HBM2 GPUs will bring!

I do find it annoying/misleading when AT's GPU overclocking section has the card being tested overclocked going against all other cards running at stock. This creates misleading comparisons to be honest. It's going to be hard to compare 980/980Ti OC vs. Fiji PRO/XT OC in AT's review because the 980 OC numbers would be from old reviews, using older games on older drivers.