AMD: Volcanic Islands R1100/1200 (8***/9*** series) Speculation/ Rumour Thread

[jimbo75]

Nor heat and power too ! I seem to recall AMD releasing a video mocking GTX 480 for exactly the same stuff 290X is doing !

Drawing nearly twice the power of the 5870 while performing 10% faster probably had something to do with that. The 290 is drawing what, 10% more power while performing the same. And it's cheaper as well!

The comparison isn't remotely valid. Also calling AMD's turbo half-baked is completely ridiculous. It's likely accounting for a large part of the performance gains and this will surely be seen with proper coolers.

 

[DavidGraham]

The comparison isn't remotely valid.

It's actually worse , GTX 480 did all of that and was faster in the end , 290X draws even more than 780Ti on all cylinders and still end up slower !

 

[Gipsel]

Pure marketing because they didn't have their own turbo solution yet.

The "turbo" or "boost" vocabulary is pure marketing. All solutions always clock at the maximum qualified clockspeed until they hit some limit. That's true for the original Powertune, the improved "boost" version and also the new (non deterministic) version we got now as well as all versions of nV's Boost (always non deterministic). It's just a question of how many powerstates are available, how (fast) one change between them, how conservative or aggressive these limits are set (the first Powertune versions, also the boosting ones stayed almost always at full clocks outside of FurMark, exception was maybe the 7950 with boost), and (depending on the kind of used limits) if the resulting behaviour is deterministic or not.

 

[jimbo75]

It's about degrees. The 480 was literally drawing almost double the power of the 5870.

83% more power on average - http://tpucdn.com/reviews/NVIDIA/GeForce_GTX_480_Fermi/images/power_average.gif

If the 290X ends up 10% slower than the 780 Ti, with 10% more power draw it's not even close to that disaster.

There has to be some tradeoffs, AMD has a smaller die and less transistors so you get the performance for less. OC'd versions with proper coolers will be better in performance and power too.

edit - just saw the reviews, power consumption is about equal on AT. 10% less performance for around the same power (with 1GB more memory)? This is somehow AMD's Fermi?

Hexus even has the 780 Ti drawing 10% more.

 

[trinibwoy]

The cooler sucks but apart from that I fail to see how AMD's implementation of Turbo is "half-baked". If anything it's far more sophisticated that NVIDIA's. It's just marketed differently, and perhaps poorly, but that's a different matter.

By implementation I'm referring to the cards, the silly quiet mode and the poor cooling. Powertune itself is fine. Like I said they'll get it right next time.

The "turbo" or "boost" vocabulary is pure marketing.

Use whatever vocabulary you like. The default behavior of the product and features marketed to the consumer have changed. The 7970ghz edition wasn't "up to 1Ghz but really only 850Mhz unless you increase the fan speed".

 

[DavidGraham]

Anyhow, Obviously this has turned into an armed race for clocks! Both architectures are too close to each other spec wise !

-number of Shaders is nearly identical
-Kepler has more TMUs
-Hawaii has more ROPs
-memory bandwidth almost the same

Kepler has a slight advantage in rasterization rate.. that aside, both end up more or less equal. As a result clocks became more vital than ever. The one who can get ahead in clock speed is the one caching in all the chips!

 

[gamervivek]

It's actually worse , GTX 480 did all of that and was faster in the end , 290X draws even more than 780Ti on all cylinders and still end up slower !

It easily bests it at 4k


http://www.hardware.fr/articles/912-22/recapitulatif-performances.html

A xtx version with improved drivers(never mind mantle) would easily top the overclocked version of 780Ti. Though that would be most likely a 2014 affair, the custom cards will have to do for now.

 

[UniversalTruth]

I think they should leave this meaningless race for a frame or two more and focus all their resources on true next generation cards- 20 nm, Please!

 

[no-X]

It's actually worse , GTX 480 did all of that and was faster in the end , 290X draws even more than 780Ti on all cylinders and still end up slower !

780Ti is faster than 290X in silent mode, but its power consumption isn't lower. 780Ti has lower power consuption than 290X in turbo mode, but it's barely faster (2 % faster according to ComputerBase, 8 % faster according to TechPowerUp and 2 % slower according to Hardware.fr - all in 2560*1xxx).

290X equipped by an alternative cooler (e.g. Accelero Xtreme III) is cheaper, faster and even more silent than 780Ti.

 

[LittleJ]

780Ti is faster than 290X in silent mode, but its power consumption isn't lower. 780Ti has lower power consuption than 290X in turbo mode, but it's barely faster (2 % faster according to ComputerBase, 8 % faster according to TechPowerUp and 2 % slower according to Hardware.fr - all in 2560*1xxx).

290X equipped by an alternative cooler (e.g. Accelero Xtreme III) is cheaper, faster and even more silent than 780Ti.

290X should certainly be more modest on power consumption in silent mode with the clock decreases.

Most reviews are done by leaving default NV power and temp targets.....aka NV silent mode. Once settings are tuned to NV Uber mode like what hardware.fr is doing the card starts to stretch its legs a little. It will be interesting to see something like an MSI 290X lightning vs the 780Ti lightning in a couple months.

 

[DavidGraham]

780Ti is faster than 290X in silent mode, but its power consumption isn't lower. 780Ti has lower power consuption than 290X in turbo mode, but it's barely faster (2 % faster according to ComputerBase, 8 % faster according to TechPowerUp and 2 % slower according to Hardware.fr - all in 2560*1xxx).

290X equipped by an alternative cooler (e.g. Accelero Xtreme III) is cheaper, faster and even more silent than 780Ti.

That is not going to happen , Hawaii is a chip pushed to the extremes of it's process technology .. far more than GK110 is. The "equipped with" argument applies to both sides , so it really doesn't count at all.

And cherrypicking results from different sources then mixing them into a single conclusion is not accurate at all either. Computerbase shows the uber 290x consuming 100W more than 780Ti , hardware.fr has it at 50W more.

In short 290X is slower than 780Ti while being considerably hotter , noisier and more power hungry .

 

[LordEC911]

That is not going to happen , Hawaii is a chip pushed to the extremes of it's process technology .. far more than GK110 is. The "equipped with" argument applies to both sides , so it really doesn't count at all.

I would say they are equally pushed to their limits, just in different areas.
AMD pushed the clocks/power/heat as high as they could.
Nvidia pushed their chip design as far as they could and, apparently, went too far again.

Needing almost a year to release a non-neutered GK110 and then needing B1 to have the supply to release a consumer version.

Edit- Quality coolers will likely drop a good 20C off the max temps and save about the same in power consumption.

 

[DavidGraham]

I would say they are equally pushed to their limits, just in different areas.
AMD pushed the clocks/power/heat as high as they could.
Nvidia pushed their chip design as far as they could and, apparently, went too far again.

Needing almost a year to release a non-neutered GK110 and then needing B1 to have the supply to release a consumer version.

Good point.

 

[jimbo75]

@David

Yet Hexus has the Ti consuming 31W more and techpowerup has them basically identical during gaming. I dunno what's up with computerbase, are they using Furmark? The gaming power draw is very close.

I'm not sure how you can say the 290X is pushed to the extremes yet the 780Ti isn't. The 290X has a clearly inferior cooler which is making power draw appear worse than it is as well. Add another 1GB of VRAM and you might as well call it a wash.

The noise is all about the cooler also. In silicon terms they appear to be very close with perhaps even a very slight edge to Hawaii. Considering the area advantage they have you have to say AMD has completely closed the gap and then some.

 

[Dooby]

That is not going to happen , Hawaii is a chip pushed to the extremes of it's process technology .. far more than GK110 is. The "equipped with" argument applies to both sides , so it really doesn't count at all.

Hardly. Look at the original HardOCP 290X with the fan set to 100%. In Farcry 3 just the boost clock alone (without a "real" overclock) gets another 10-20% over the "uber" setting of 40% fan. With a decent HSF like ASUS CUII, there's plenty of performance left on the table for the 290X. A few posts back was a quote from a site who put a IceQ X2 from last generation on the chip, and it ran at 60-70C.

Whoever is in charge of the HSF for AMD should be fired, NOW! On the 7970 it was noisey and crap already. To not even bother with a decent HSF this time is simply criminal from a business perspective. They have an amazing chip, and have squandered so much good PR with a 3-4 yr old heatsink. Simply terrible.

 

[pjbliverpool]

290X equipped by an alternative cooler (e.g. Accelero Xtreme III) is cheaper, faster and even more silent than 780Ti.

No it wouldn't because most benchmarks place the 780Ti faster than the 290x in uber mode and when in uber mode the card is running consistently at its maximum clock speed. A better cooler will make it quieter than it is with the reference cooler but there's no guarantee it'll be quieter than a 780Ti - especially a Ti with the equivalent cooler.

Its only going to be faster (on average) than the Ti with a significant overclock which would again push noise and power consumption up. And then its only comparable to similarly overclocked Ti's which once again would likely tip the performance advantage to the Ti.

 

[jimbo75]

pjb - better coolers also lower power draw. There could be a saving of 20-30W with an after-market cooler, which in turn can be turned back into more performance.

These chips are incredibly close together in perf/W or I'll eat my hat.

 

[Nemo]

What the hell happened AMD?

 

[LordEC911]

A better cooler will make it quieter than it is with the reference cooler but there's no guarantee it'll be quieter than a 780Ti - especially a Ti with the equivalent cooler.

That will be an interesting comparison, I would like to see those results very much.

I was just thinking, shortly after 290x launch, that someone should mod the Titan cooler for Hawaii.

 

[Jawed]

So, basically what we have here is that AMD (or is that NVidia?) has given its partners a licence to charge an extra $150 for a 1.1GHz extreme-cooler card?

Should be a tasty choice.