Amd R290X early benchmarks

[Xzibit]

According to the original source, the 290X being benched was running at it's maximum stable clock of 1020/5000. While the production parts will undoubtedly overclock higher, it also tends to suggest that the card at 1020 core is overclocked past reference speed.

They might be pulling a Tahiti all over again.

290X when OC'd beats Titan and uses less power in doing so.

wait for Nvidia to respond then release 290X GHZ edition..

Either way it looks like a good card if it can beat 780 & Titan while being OC'd and still using less power then Titan

The early 7970 were able to reach 1000-1200 ghz but they sucked power while doing so and if this can stay under TITAN stock power usage while delivering more performance and costing much less

Hmmm....

If this is Hawaii XT is there a Hawaii XT2 ?

 

[Naito]

That isn't nvidia's next round, that is just them finally releasing a year old GPU to the consumer market.



Kepler came out 3 months after Tahiti, that isn't half a year.

And it was so good the mid-range GPU outperformed AMD's top Tahiti.

Kepler was so far ahead that nvidia had no need to release their high end GPU and just sat on it for a year or more.

The sad thing is that this looks like it might be AMD's best shot and it barely beats Titan.

That means that all nvidia might have to do is a slight clock speed bump with maybe a slightly better reference design with a stronger power setup and they'll easily be able to continue to compete.

Of course they'll finally have to lower the price on titan to something reasonable too.

But then they'll just release whatever they've been working on for the past year or more and charge $1000 or that because AMD won't have anything to compete with it.

Gotta say I agree with most points here. AMD releasing lackluster hardware allowed Nvidia to release originally mid-range Kepler against the Tahiti to maintain competition in that segment, whilst allowing them [Nvidia] to price gouge in the enthusiast segment with products such as GTX 780/Titan.

 

[LAN_deRf_HA]

They've supposedly been sitting on a fully enabled GK110 chip since Kepler launched, so all they'd need to do is drop that and shuffle price points around. Nvidia has been coasting since they first saw 7970 benchmarks. I'd expect their next architecture to be a massive leap with all the R&D time it's gotten, but let's face it they'll just keep pulling their punches since AMD can't challenge them.

 

[HumanSmoke]

The back plate is also half vent, half connectors. The picture posted in the OP has a DVI ports stacked on another like on my 6870s. So yes, the cooler looks the same. That's about it.

That was my point. I doubt that the shroud in the OP's post is the final design for the R9-290X. I would also doubt that the production item will ship with those jumpers that are next to the chokes.
And no, I wasn't suggesting that the pictures in the OP's post were of a HD 7970- just to make that clear.
As for the red PCB. Some reference HD 7970's shipped that way, so no big surprise there.

They've supposedly been sitting on a fully enabled GK110 chip since Kepler launched, so all they'd need to do is drop that and shuffle price points around.

Probably a fair assumption. Nvidia have a few combinations to work with. A fully enable 2880 core Titan Ultra at the $999 price point, dropping the Titan down to $650-750 (depending upon whether they do/do not split the model into 6GB and 3GB versions) and the GTX 780 down into whatever the R9-290X ships at. There is also the possibility that Nvidia could bifurcate the line by allying the GK 110 with 7Gb/sec memory as per the GTX 770/760. Throw in the dual and triple BIOS and/or relaxed voltage that EVGA are shipping and there's more than a little scope for tweaking the lineup.
The interesting question is what Nvidia plan to do against the inevitable second-tier AMD salvage part. It will obviously be superior to the GTX 770.

 

[the54thvoid]

It's still quite weird people choose sides over gfx cards. The facts are pretty self evident. GK110 (as K20x) was released in Nov 12. Titan is the same card with some bits disabled. Given (as i was rightly corrected in a previous thread) Titan is roughly 25% faster than the 7970Ghz (a billion years old now) you'd expect the new card to beat Titan.

Fact is, if it doesn't, then that's really bad for AMD on the PR front. Despite the fact I own a Titan, I honestly want AMD to beat it. No fan boy nonsense required, just good tech economics.

Problem is the price. AMD released the 7970 very high. If they have a card that beats the 780 they'll easily launch in that price range.

If AMD release a Titan level card at half it's price I'd be absolutely gobsmacked. I think that is wishful thinking.

Roll on the AMD show

 

[Crap Daddy]

Problem is the price. AMD released the 7970 very high. If they have a card that beats the 780 they'll easily launch in that price range.

If AMD release a Titan level card at half it's price I'd be absolutely gobsmacked. I think that is wishful thinking.

Roll on the AMD show

Nvidia set the bar for high-end card at $650, I'm not talking about Titan since it was/is an exception and even if AMD will offer a card that beats the Titan it will be impossible for them to sell it at the same price point. But I can see a $650 price tag for the 290X. What Titan and the 780 proved is that there is a strong market for insanely priced cards and AMD cannot afford to price it much differently.

 

[HumanSmoke]

Nvidia set the bar for high-end card at $650, I'm not talking about Titan since it was/is an exception and even if AMD will offer a card that beats the Titan it will be impossible for them to sell it at the same price point. But I can see a $650 price tag for the 290X. What Titan and the 780 proved is that there is a strong market for insanely priced cards and AMD cannot afford to price it much differently.

AMD (read Rory...Read) has already demonstrated (t)he(y) isn't adverse to huge price tags. $999 for the HD 7990, $stupidprice for the FX-9590. The only time Rory shows pricing restraint is when he absolutely, positively needs to move a serious amount of units to offset the Monopoly-money halo parts.

 

[TheoneandonlyMrK]

They've supposedly been sitting on a fully enabled GK110 chip since Kepler launched, so all they'd need to do is drop that and shuffle price points around. Nvidia has been coasting since they first saw 7970 benchmarks. I'd expect their next architecture to be a massive leap with all the R&D time it's gotten, but let's face it they'll just keep pulling their punches since AMD can't challenge them.

What, , kepler wasn't a massive leap after a year and a half wait so your optimistic outlook is just that and nvidia are going to have to do what they can with a dear to make gk110 for a while as gk111 has been canned and maxwell using tsv connected edram on 20nm is still in daydream land with the ps4 and xbox ones next soc designs.
Ot benches matter to 10% of pc part buyers the rest just buy what they can afford and there mate dave recommend, priced right anything sells

 

[HumanSmoke]

as gk111 has been canned

You taking a leaf out of the Charlie Demerjian book? Announce a ficticious GPU then announce that the GPU has been cancelled.
Can't say I've seen any GK111 literature, let alone its cancellation. Got any links? I'm keen to see what I'm missing.

 

[Xzibit]

They've supposedly been sitting on a fully enabled GK110 chip since Kepler launched, so all they'd need to do is drop that and shuffle price points around. Nvidia has been coasting since they first saw 7970 benchmarks. I'd expect their next architecture to be a massive leap with all the R&D time it's gotten, but let's face it they'll just keep pulling their punches since AMD can't challenge them.

AMD Winner in Q2, Intel Up, NVIDIA Down, According to Jon Peddie Research

I don't think that's the direction you want to go while coasting and pulling your punches.

 

[HumanSmoke]

I don't think that's the direction you want to go while coasting and pulling your punches.

Well, it's pretty safe to say that Nvidia's overall graphics market share will fall considering they no longer make chipsets of any note and can't bloat their market share figures with APU numbers.
The discrete graphics business ( the one where they actually compete) is actually pretty favourable- especially given their high ASP's, and the cyclic fluctuations

Now that everyone's vented their dislikes and hung out their bait, how about a return to the subject at hand which I believe is an AMD card.

 

[TheoneandonlyMrK]

You taking a leaf out of the Charlie Demerjian book? Announce a ficticious GPU then announce that the GPU has been cancelled.
Can't say I've seen any GK111 literature, let alone its cancellation. Got any links? I'm keen to see what I'm missing.

i like many read a lot of sources and wouldnt put 100% faith in many but charlies prob right there due to nvidia focusing on tegra 4 lte, but being an electronics eng i can read betwixt the lines regarding fab node availabilty and what each company is trying to achieve.

It would be very nice(for us) if both amd and nvidia had the nodes and and packages they want right now but neither have exactly an ideal situ to work with.

 

[german103]

AMD Winner in Q2, Intel Up, NVIDIA Down, According to Jon Peddie Research

http://tpucdn.com/img/13-08-14/118a_thm.jpg

I don't think that's the direction you want to go while coasting and pulling your punches.

counting Intel CPU is uncalled for isn't it?
I mean who cares

edit: forgot about AMD APUs

 

[radrok]

I hope these aren't spot on, I hope they perform way more than this.

7 months for a tie? bleh.

 

[cdawall]

I hope these aren't spot on, I hope they perform way more than this.

7 months for a tie? bleh.

At times it doubles 7970 performance...what more do you want from amd? Give it a more mature driver and see how much better it does.

 

[micropage7]

nice card, but AMD should improve their driver too. make it simple and powerful

 

[radrok]

At times it doubles 7970 performance...what more do you want from amd? Give it a more mature driver and see how much better it does.

Would love a 20% on top of Titan

 

[Steevo]

All I want is a card that my waterblock mounts to, that doesn't cost a small fortune, and performs well enough to run games at 1080 60FPS with high eye candy. And I want GTA5 on PC.

 

[radrok]

All I want is a card that my waterblock mounts to, that doesn't cost a small fortune, and performs well enough to run games at 1080 60FPS with high eye candy. And I want GTA5 on PC.

If your waterblock is universal I don't see why it wouldn't

 

[newconroer]

AMD needs to achieve four things on their next lineup and moving forward :

A) Crossfire in windowed mode - gotta put this to rest, it's been years.
B) Easier over clock management. The use of similar voltage and auto power flex to that of Kepler would facilitate this.
C) Full support in every driver for mis matched cards in Crossfire configuration. e.g. 7950 + 7970
D) 4GB standard on all enthusiast cards.

I'd also like to see someone finally find a way to merge GPU RAM when used in Crossfire/SLI.
That or fit 6GB minimum onto a single card.

Guarantee if their cards perform in the ballpark, and they tackle at least A,B and C, then they'd win a huge share back from Nvidia.

betwixt the lines

That's a serious typo brother.

 

[TheGuruStud]

Gotta say I agree with most points here. AMD releasing lackluster hardware allowed Nvidia to release originally mid-range Kepler against the Tahiti to maintain competition in that segment, whilst allowing them [Nvidia] to price gouge in the enthusiast segment with products such as GTX 780/Titan.

It was impossible to produce those chips. They actually pulled an AMD (their CPUs mostly) and got ahead of themselves by making something too complex and big. It was at least a month before you even had a chance to buy the 1st kepler cards and 2 months, IIRC, before it reached the masses. Until 28nm was pretty mature, there was no way in hell nvidia could have ever made gk110 (from next to nothing wafer yields and not able to reach the needed clocks). To top it off, they removed all the compute power (dick move). But this has helped AMD a lot, so

AMD already learned that profitability takes a severe hit when you try doing that, so they've scaled back their operation to fit within the constraints of shitty TSMC and GloFo (and even that has to get delayed a lot). And they still got screwed. Tahiti never wasn't able to be made as power efficient as designed (killing all the 1300 MHz clock hopes by board partners). Fabs have been AMD's weak point forever. They don't have the capital to operate one (barely did before). There's nothing really wrong with either the CPU or GPU designs, but they can't physically be made correctly, so they get scaled back and stripped down (it was amazing that athlon 64 was made, even intel couldn't have done it haha - ever since has been a battle from design to production).

 

[HumanSmoke]

betwixt the lines

That's a serious typo brother.

Far from it. The word is correct and used in its proper context.

I'd also like to see someone finally find a way to merge GPU RAM when used in Crossfire/SLI.

Unified memory pool won't happen in AFR. AFR barely works consistently with discrete framebuffers without the major headache that pooling it would pose.

 

[SIGSEGV]

according to videocardz.com AMD Hawaii GPU has 2816 Stream Processors

 

[RCoon]

according to videocardz.com AMD Hawaii GPU has 2816 Stream Processors

That PCB VRM Phase looks totally retarded. 5+1, really? I'd like to see a stronger VRM on the 2990X

http://www.techpowerup.com/gpuz/ub828/

Even the memory clock looks wrong.

 

[erocker]

Pretty much the same layout as 7970. TDP isn't much different either.