AMD - What can they do to improve?

[humbug]

Get the cooling and power use improved/lowered and I could be tempted with something from the 300 series. Their cards have always been great value for money, but the cooling in particular has put me off trying them.

+1, performance Per-Watt needs to come up.

Actually it really does have to, for a single GPU 250 to 300 Watts is the limit of what AIB Air coolers are good for, the 290X and 780TI fell into that range.

If AMD want to get a significant performance margin over the 290X with the 300 Series the performance per-watt must be improved significantly or its just not going to happen.

 

[snips86x]

The biggest issue for me was the lack of driver support reference cards using vSync. When usinf vSync, the xore would fluctuate causing instability and would crash the display driver and/or cause a hard reboot.

While it was fixed using RadeonPro, it wasn't a priority!

 

[Zethor]

Currently, Nvidia is the champion and AMD is the challenger. To take the belt, it's not enough to be the champ's equal, you need to knock him out and that's exactly what AMD needs to do with their new cards.

 

[Marine-RX179]

Marketing, marketing and more marketing. They have been and are very good cards, but there is a perception that if you can get similar frame rates for slightly more money from nVidia then that's a better option - or the people who think gaming card=nVidia

IMO marketing alone will NEVER change people's image of Nvidia being better, regardless how much AMD improve their own image...what they need a "facts/action speaks louder than words".

What AMD need to pull off is similar to when they introduced the original Athlon CPUs, or what ATI did when they released the 5000 series, and give a huge system shock to the market.

Although I seriously doubt they can pull it off, but if they can somehow, someway pull off releasing their next card being significantly faster than the Titan X and cheaper, it would serious mess up Nvidia's grant plan of selling their fastest card the Titan X at crazy price level and grim it down become different products to fit into difference (overpriced) price bracket...it will throw them right back to the drawing board on what to do next rather than the milking they've been counting on.

 

[humbug]

IMO marketing alone will NEVER change people's image of Nvidia being better, regardless how much AMD improve their own image...what they need a "facts/action speaks louder than words".

What AMD need to pull off is similar to when they introduced the original Athlon CPUs, or what ATI did when they released the 5000 series, and give a huge system shock to the market.

Although I seriously doubt they can pull it off, but if they can somehow, someway pull off releasing their next card being significantly faster than the Titan X and cheaper, it would serious mess up Nvidia's grant plan of selling their fastest card the Titan X at crazy price level and grim it down become different products to fit into difference (overpriced) price bracket...it will throw them right back to the drawing board on what to do next rather than the milking they've been counting on.

How about the HD 4000 series?

Hopefully HBM will be the new GDDR5 shocker.

 

[Ayahuasca]

+1, performance Per-Watt needs to come up.

Actually it really does have to, for a single GPU 250 to 300 Watts is the limit of what AIB Air coolers are good for, the 290X and 780TI fell into that range.

If AMD want to get a significant performance margin over the 290X with the 300 Series the performance per-watt must be improved significantly or its just not going to happen.

It's funny you should mention that because at the time of the 970 release when people were praising it on performance per watt there were loads of members saying they couldn't give a monkeys how much power their cards use as long as it performs. So the point you've made is a good one as these 300w cards are pushing the limits for air cooling in the majority of peoples setups.

 

[Marine-RX179]

How about the HD 4000 series?

While the 4000 series did provided some shock in terms of value for money, but even the 4890 wasn't really fast enough to challenge the GTX280/285

 

[Rusty0611]

Nvidia has done the same...remember the big deal that they made out of AMD's framepacing "issue"? Now that it happens on their cards, it is "there's nothing to see here move along"

The framepacing issue is a tinny problem in contrast to the 970 memory controversy, but clearly AMD deserve more sticks than Nvidia.

As I said it's all petty. nVidia, AMD and Sony et al. But AMD aren't in the same position as nVidia so what I think he was getting at was that he'd rather they focus more effort on the things that matter to their customers. Stuff like that doesn't bring in more customers really either. And boy do AMD need some more customers at the mo.

For me, I buy what's best for me at the time but that's just my anecdotal take on it. The market has spoken.

 

[humbug]

While the 4000 series did provided some shock in terms of value for money, but even the 4890 wasn't really fast enough to challenge the GTX280/285

Ah ok.

 

[philo-sofa]

In terms of the next generation people here are right: AMD needs to ape Nvidia’s move towards efficiency and partnering in a ‘deep’ fashion with its fab (i.e. TSMC) to get the most out of each generation in terms of performance per watt and low fault rates on silicon. I’d like to see AMD producing ‘pipe-cleaner’ parts like they last did with the… HD 3770? Also agree that moving away from the compute focus in GCN 2.0 may be a good idea; that after all is a large part of what turned Nvidia’s 400 series space heaters into the ass-kicking 500 series. Perhaps go back to small die (optimise to where the largest volume sells) but go a bit more 'whole hog' than the Green team and release a double size piece of silicon too – challenge the performance halo Nvidia tends to win.

Process technology, die size and generations aside AMD needs to buck up their marketing ideas, lose the underdog mentality, perhaps reduce its tempo of driver ‘features’ (it's a trade-off and is recording really that important to the mass market?) and actually fix and stabilise its drivers. Moving off the monthly release schedule was a succesful start to this but it's tapered off. It would also benefit AMD massively to start making a fuss about their latest cards and dripping information out to create its own free marketing and associated ‘halo’. When it can present with well behaved drivers and present well they’ll be better able to leverage the fact it seems to handily win every other generation (4000 series, 5000 series, eventually the 7000 series and the R9 290 / 290X smackdown prior to Maxwell) to produce the comparable or better sales numbers the hardware so often deserves.

On the CPU side maybe try and poke GloFo to vaguely keep up with Intel on process tech; something neither GloFo, nor AMD’s manufacturing arm before them have never acheieved. But primarily AMD just needs to get Zen out the door and have it work well.

 

[bobboberts]

Surely some of the main issues - heat and loud - are purely because of the one advantage they usually have - price. To satisfactorily address those issues would presumably necessitate a jump in price, and if their cards started getting closer to nVidia prices that would hurt them because of the perception that nVidia cards are worth the premium price. If anything, a jump in AMD prices would make nVidia cards look more attractive a proposition.

 

[Ayahuasca]

Noise was only really an issue with the reference coolers, you can get loud aftermarket coolers on nvidia cards as well, sometimes it's even the more expensive models that are louder.

Heat shouldn't necessarily be higher just because of low prices either. You have to remember at the time of release the reference 290x was around the same price as the 980.

 

[Rusty0611]

Yeah and as well the heat is what it is. The cooler just efficiently (or not) removes it from the GPU but that heat is still there. Either in your case or in your room.

 

[philo-sofa]

While the 4000 series did provided some shock in terms of value for money, but even the 4890 wasn't really fast enough to challenge the GTX280/285

That isn't really representative of reality. The 4890 was a tad slower than the 285 but it also cost about half as much to make and thus was far cheaper than the 280/285 to buy and more profitable to sell. We often forget that Nvidia completely gave up on high end GPUs for a few months between the 200 series and 400 series. For a while there was just no point trying to use the 200 series versus the HD 4000 series; they just stopped making high end GPUs.

Which brings to mind another point; whatever happened to the 'decap ring' that allowed higher clocks via more stable volts on the 4890? Does everyone use that now or soemthing?

 

[eddyr]

In terms of the next generation people here are right: AMD needs to ape Nvidia’s move towards efficiency and partnering in a ‘deep’ fashion with its fab (i.e. TSMC) to get the most out of each generation in terms of performance per watt and low fault rates on silicon. I’d like to see AMD producing ‘pipe-cleaner’ parts like they last did with the… HD 3770? Also agree that moving away from the compute focus in GCN 2.0 may be a good idea; that after all is a large part of what turned Nvidia’s 400 series space heaters into the ass-kicking 500 series. Perhaps go back to small die (optimise to where the largest volume sells) but go a bit more 'whole hog' than the Green team and release a double size piece of silicon too – challenge the performance halo Nvidia tends to win.

Process technology, die size and generations aside AMD needs to buck up their marketing ideas, lose the underdog mentality, perhaps reduce its tempo of driver ‘features’ (it's a trade-off and is recording really that important to the mass market?) and actually fix and stabilise its drivers. Moving off the monthly release schedule was a succesful start to this but it's tapered off. It would also benefit AMD massively to start making a fuss about their latest cards and dripping information out to create its own free marketing and associated ‘halo’. When it can present with well behaved drivers and present well they’ll be better able to leverage the fact it seems to handily win every other generation (4000 series, 5000 series, eventually the 7000 series and the R9 290 / 290X smackdown prior to Maxwell) to produce the comparable or better sales numbers the hardware so often deserves.

On the CPU side maybe try and poke GloFo to vaguely keep up with Intel on process tech; something neither GloFo, nor AMD’s manufacturing arm before them have never acheieved. But primarily AMD just needs to get Zen out the door and have it work well.

They are making a push into the pro/supercomputing market and seem to be making some inroads there I believe, which is part of the point to GCN as well as their long-term APU coprocessor goals. Not sure they could lose compute capability without co-developing a less compute focused arch like Nvidia. Somewhat amusing that Nvidia and AMD somewhat took up each others design strategies. Except Nvidia decided to co-develop 2 arch variants to address different markets while AMD scaled just the one.

 

[Dave2150]

One huge issue is that the majority of the community regards AMD GPU's as being the budget, or cheaper option. They believe that NVIDIA cards are much better because they are more expensive, and fail to actually look at benchmarks that show AMD have been very competitive upto Maxwell release.

NVIDIA command so much of the market share now, last report I read showed it was around 75%. That's a huge share of the market, and will only get worse unless AMD produce a new generation of GPU's that are far faster than NVIDIA.

I believe that's the only way they can hope to claw back market share - for every skew in their new GPU line to be unequivocally faster than NVIDIA, whilst being comparable in heat, power consumption.

I highly doubt AMD can hope to do this, so I foresee NVIDIA going to 80 or even 90% share of the GPU market in the next few years.

The fact that so many games are now getting 'gameworks' features is also a nail in the coffin for AMD.

My next GPU will probably have to be an NVIDIA model - which is quite sad for me as I've had AMD GPU's starting from the 9800pro many years ago

 

[eddyr]

^

Is that market share of units shipped per quarter or an aggregate? not sure I've seen the latter. Not that I've bothered to look.

 

[SiDeards73]

Great responses so far folks. Seems like most of us are of the same opinion. Part of me wishes AMD would ditch either GPU or cpu to focus on just one and become great at It, however lack of competition in the market for either would be a disaster.

 

[Final8y]

Great responses so far folks. Seems like most of us are of the same opinion. Part of me wishes AMD would ditch either GPU or cpu to focus on just one and become great at It, however lack of competition in the market for either would be a disaster.

If they drop the GPU side then you are only left with NV, if they drop the CPU side then you are only left with Intel, so no dropping one side is not a good idea.

Its the sum of all AMD parts that keeping them going and keeping Intel and NV in check to some degree.

 

[Stanners]

If they drop the GPU side then you are only left with NV, if they drop the CPU side then you are only left with Intel, so no dropping one side is not a good idea.

Its the sum of all AMD parts that keeping them going and keeping Intel and NV in check to some degree.

This. We need AMD strong to keep the others inline.