The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
People complain a lot about amd drivers , but I have friends on NVidia cards and they constantly have problems and since they are all on older cards they have to wait a very long time for NVidia to get around to fixing performance on those cards for newer games (if they ever do)

With AMD it seems to be a consistant month wait for drivers tuned for the new card which isn't a bad wait. I also have a line up of amd cards going back to the 5850 and they are still supported and still gaining in performance with newer drivers.

So I think the NVidia drivers are better is just a myth that was at one time perhaps true but has simply become part of the zetigiest
 
Nivdia seem to be better at squeezing extra performance out of their cards. I believe their driver team is many times the size of AMD's.

If you go back to look at how older cards perform with never games - once driver team optimisation no longer factors in - AMD cards appear to age better. I've seen a compilation of scores for older games (from around launch) vs newer, and AMD gains several percent. Nvidia really seem to double down on games that are commonly used for bechmarking.

On another note, Bulldozer will be four years old in october. The "high end" AM3 chips haven't even been updated to match their younger and higher IPC brothers. It's sad that even a 10% IPC boost and ~30% die shrink couldn't give them something worth trying to sell. :(
 
Nivdia seem to be better at squeezing extra performance out of their cards. I believe their driver team is many times the size of AMD's.

If you go back to look at how older cards perform with never games - once driver team optimisation no longer factors in - AMD cards appear to age better. I've seen a compilation of scores for older games (from around launch) vs newer, and AMD gains several percent. Nvidia really seem to double down on games that are commonly used for bechmarking.

On another note, Bulldozer will be four years old in october. The "high end" AM3 chips haven't even been updated to match their younger and higher IPC brothers. It's sad that even a 10% IPC boost and ~30% die shrink couldn't give them something worth trying to sell. :(

I don't think bulldozer for desktop ever got a die shrink ?

I think it was smart of them to just stop and stay low end with their cpu's until they came up (hopefully) something better like the new chip that's launching next year. Lets just hope the new process at GF works well for them and the chip comes in on time with no problems lol.
 
I think it was smart of them to just stop and stay low end with their cpu's until they came up (hopefully) something better like the new chip that's launching next year.

But if you're going to make those FX processors for desktop, wouldn't it pay to make them as none-awful as possible?

32nm was built for speed where as 28nm is more of a compromise, but even so AMD have been able to crank out 4+ gHz (with turbo) processors on 28nm, and at lower speeds of 3~3.5 ghZ they're now massively more power efficient and with the IPC gains around the same speed Bulldozer. Significantly smaller, far more power efficient and seemingly a much better bet for servers seems like a better position than they're in now, but ...

... is the cost of engineering the new chip (even with the IP and much of the layout done) just too high, or are AMD contractually obliged to keep making 32nm chip and left with no option other than "get sued"?

Lets just hope the new process at GF works well for them and the chip comes in on time with no problems lol

I'm already getting bad feels of a paper launch in late 2016. :(
 
I don't think AMD has said anything other than "2016". In AMD speak, "year N" usually means "the second half of year N", and sometimes "paper launch in December of year N with retail availability in early year N+1"…
And that is when the product isn't delayed.

That said, AMD usually releases CPUs in the summer.
 
But if you're going to make those FX processors for desktop, wouldn't it pay to make them as none-awful as possible?

32nm was built for speed where as 28nm is more of a compromise, but even so AMD have been able to crank out 4+ gHz (with turbo) processors on 28nm, and at lower speeds of 3~3.5 ghZ they're now massively more power efficient and with the IPC gains around the same speed Bulldozer. Significantly smaller, far more power efficient and seemingly a much better bet for servers seems like a better position than they're in now, but ...

... is the cost of engineering the new chip (even with the IP and much of the layout done) just too high, or are AMD contractually obliged to keep making 32nm chip and left with no option other than "get sued"?



I'm already getting bad feels of a paper launch in late 2016. :(

I would think it was their socket am3+ that was the problem. They would have needed to design a modern socket to go with the newer chips.


If I were AMD i'd be praying the zen cores were very power efficient to team up 4 of them with a dx12 gpu in a single 5-15watt apu and try to break into the lower end tablet market and 4k set top box market.
 
I don't think AMD has said anything other than "2016".

I do think you missed some articles, at least couple of them.

AMD CEO Hints to Release Date of K12 And x86 Sister Core – Targets Q1 2016

Read more: http://wccftech.com/amd-ceo-hints-release-date-k12-based-products-targets-q1-2016/#ixzz3e8j9P7kd

However, rival AMD is also launching its Zen core in early 2016 posing a challenge to both Skylake and Broadwell CPUs

http://www.valuewalk.com/2014/10/intel-broadwell-e-production-q1-2016/

Perhaps what’s more exciting is the fact that Broadwell-E will likely not go uncontested. With news of AMD’s Zen core telling of an early 2016 launch

Read more: http://wccftech.com/intel-broadwell-e-production-in-2016/#ixzz3e8jhe1ci
 
I do think you missed some articles, at least couple of them.

AMD CEO Hints to Release Date of K12 And x86 Sister Core – Targets Q1 2016

Read more: http://wccftech.com/amd-ceo-hints-release-date-k12-based-products-targets-q1-2016/#ixzz3e8j9P7kd

However, rival AMD is also launching its Zen core in early 2016 posing a challenge to both Skylake and Broadwell CPUs

http://www.valuewalk.com/2014/10/intel-broadwell-e-production-q1-2016/

Perhaps what’s more exciting is the fact that Broadwell-E will likely not go uncontested. With news of AMD’s Zen core telling of an early 2016 launch

Read more: http://wccftech.com/intel-broadwell-e-production-in-2016/#ixzz3e8jhe1ci

This is about K12 more than Zen, and K12 has been delayed. Plus, those slides don't say anything about "early" 2016.
 
This is about K12 more than Zen, and K12 has been delayed. Plus, those slides don't say anything about "early" 2016.

Sorry, I think you do not understand something.

Yes, K12 is delayed even further than Zen, for sometime in 2017 now.
While, it looks like that at the moment Zen is scheduled for Q3 2016.

Which doesn't change the fact that ORIGINALLY both were planned for (early) 2016 release.

No matter how unpleasant it sounds for you.
 
I know this must be a bit like reading tea leaves, but do we have any idea if AMD are really likely to deliver Zen next year?

I don't see how much more they can squeeze out of 28nm and Bulldozer, and Jaguar is EoL. Without some kind of presence next year, I don't see where AMD will be able to rebuild from.
 
I know this must be a bit like reading tea leaves, but do we have any idea if AMD are really likely to deliver Zen next year?

Yes, AMD should deliver as planned. If it doesn't happen, that means there are terrible troubles. No one would benefit in such a situation.

Why are you so scared that there should be necessarilly some troubles?
 
Yes, AMD should deliver as planned. If it doesn't happen, that means there are terrible troubles. No one would benefit in such a situation.

Why are you so scared that there should be necessarilly some troubles?
I'm not going to tell you to stop it with the noise again, but rather will stop it myself (again).

function: really likely is difficult to establish, as the design and context are not without risk. It's opaque to us how the 14nm process they intend to use is faring in practice, and how adequate it is for the sort of MPU they want to bring. Furthermore, this should be AMD's first SMT implementation, so validation there is an unknown. They also appear to be changing their cache hierarchy from the one that they have been using for ages (in no small part because they were confident in their ability to validate it), which adds further murkiness. Historically, AMD's woes tended to came as they were moving into validation, as opposed to when they were presenting slides, and the former might be in pretty early stages at best at this moment. We are likely to start hearing about how things are going later on during this year, one assumes.
 
function: really likely is difficult to establish, as the design and context are not without risk. It's opaque to us how the 14nm process they intend to use is faring in practice, and how adequate it is for the sort of MPU they want to bring. Furthermore, this should be AMD's first SMT implementation, so validation there is an unknown. They also appear to be changing their cache hierarchy from the one that they have been using for ages (in no small part because they were confident in their ability to validate it), which adds further murkiness. Historically, AMD's woes tended to came as they were moving into validation, as opposed to when they were presenting slides, and the former might be in pretty early stages at best at this moment. We are likely to start hearing about how things are going later on during this year, one assumes.

Thanks. So it appears that there's a good amount to be concerned about, but we may begin to know how things are going later this year.

New microarchitecture on a new process - the opposite of Intel's tick-tock and something that didn't work out so well last time. But I guess they're rather up against it and don't have much choice.
 
Thanks. So it appears that there's a good amount to be concerned about, but we may begin to know how things are going later this year.

New microarchitecture on a new process - the opposite of Intel's tick-tock and something that didn't work out so well last time. But I guess they're rather up against it and don't have much choice.
Indeed, there might've been little choice if the plan is to manage some degree of competitiveness. Note that the commonly accepted current wisdom (which is based on some leaked roadmaps that might be true) is that to begin with Zen will not trickle down into the APU lineups, and that those will see some XV refresh being brought about, still on 28nm. So to some extent they should have a "tried & true" part that that ships without much pain ready, although it is difficult to work out much enthusiasm for what can only be regarded as rather long in the tooth. It should, however, brink higher-wattage desktop SKUs, which could be appealing for those that opted for e.g. the 7850K. On the other hand, much of this hinges on what can fit pretty tight budgetary constraints, and I can imagine quite a few accounting / financial axe-men running around looking at projects unfit for life / which should be axed.
 
Status
Not open for further replies.
Back
Top