AMD Piledriver rumours ... and expert conjecture

[Reynod]

We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...

 

[jdwii]

I'm not saying its not possible for Amd to get 30-40% performance jump in 1 generation(steamy) i'm just saying its not possible without Up to statements LOL.

 

[noob2222]

I'd say around 15% at least.

for games no Read above, there was a link to a very interesting Intel archi article.

http://ixbtlabs.com/articles3/cpu/intel-ci7-123gen-p3.html

and the verdict is 136 fps to 142 woohoo 4.4%

Even at its highest gains, ivy is barely hitting 15% faster than nehalem in its video encoding. the overall difference was 12%.

I'm not saying its not possible for Amd to get 30-40% performance jump in 1 generation(steamy) i'm just saying its not possible without Up to statements LOL.

see above, even Intel can't make a 10% jump for games through 4 archies. Nahalem, Lynnfield (not shown), sandy bridge, and ivy bridge. If your saying 30% flat out ... you will fail.

 

[viridiancrystal]

for games no Read above, there was a link to a very interesting Intel archi article.

http://ixbtlabs.com/articles3/cpu/intel-ci7-123gen-p3.html

and the verdict is 136 fps to 142 woohoo 4.4%

Even at its highest gains, ivy is barely hitting 15% faster than nehalem in its video encoding. the overall difference was 12%.




see above, even Intel can't make a 10% jump for games through 4 archies. Nahalem, Lynnfield (not shown), sandy bridge, and ivy bridge. If your saying 30% flat out ... you will fail.

In brute performance, no. Remember to consider that they have also dropped their TDP by 48 watts while making performance improvements, however minimal. That considered, efficiency is WAY up in those four generations.

 

[sarinaide]

The i7 990X is the best DT processor Intel have ever made, it was a true top end beast that does pretty well today. Sold mine for $350(around that) but it was time to move on.

The hoopla going around about how far AMD is behind, well I don't see anything in Phenom II's performance that didn't rival first gen core i processors up untill the hexcores and while BD is slightly down on PII it is up in other regards namely core crunching requirements. I would hazzard a guess Intel is at best 15% faster overall.

 

[fazers_on_stun]

Guys we can bounce off the walls at each other all day long and nothing will be gained, appart from early reports of Piledriver changes over Zambezi nobody knows anything, AMD for one have been very tight lipped about anything so right now its just complete speculation. What did come out was AMD stated they are targetting at the very least a 15% performance per clock improvement per release and they did indicate they were quite pleased with Vishera's progress.

As to the ultimate king of processors, probably not, probably never again when you consider AMD's level of technology compared with Intel it is highly unforeseeable that AMD will ever defeat Goliath in a matchup of pure brute force, but in a article released on THG recently about AMD "Vision" AMD have stated themselves they are not looking to slug it out on one aspect of CPU architecture. The future is not single threaded performance.

So we have gone on for 129 pages highlighting the only negitive on Zambezi but what about the promise, how about a little love 😛.

Hmm, all I've seen is "15% performance improvement" - nothing about per clock (i.e., IPC). Considering the few changes that PD is gonna have compared to BD, I'd suspect maybe 5% IPC and 10% clock (3.6 ----> 4.0 GHz base clock). If indeed AMD hits its targets that is..

 

[fazers_on_stun]

From http://seekingalpha.com/article/852961-amd-the-sinking-chip-finding-bottom?source=msn

The next problem with AMD is the amount of insider selling activity. In the past year, by insiders, or those who are required to file with the SEC when trading AMD, 280,000 shares have been sold and only 8,000 have been purchased. Either the insiders know something, or they're going broke. Or both: the two could be related. Jokes aside, the truth is that it's not actually a big deal, as CEO Read Rory alone still owns 260,000 shares, and other officers are still holding above 200,000 shares as well.

This would be something to watch in the near future - if more company stockholders start selling their shares, I would be worried..

 

[sarinaide]

Hmm, all I've seen is "15% performance improvement" - nothing about per clock (i.e., IPC). Considering the few changes that PD is gonna have compared to BD, I'd suspect maybe 5% IPC and 10% clock (3.6 ----> 4.0 GHz base clock). If indeed AMD hits its targets that is..

Well we only know what the changes are to the architecture, but we don't know what its performance is. A 400mhz bump at well less power is maybe 5-7% clocks and the rest IPC.

Those that I liaise with suggest the AMD flagship will be around the same per core performance as a i7 960-970 with slightly better power consumption. Memory wise still behind Intel but its a marginally better IMC than on the Zambezi part.

Reading reviews on Guru3D and others a very good review on the 1100T vs 8120 and 1100T vs 2500K in general if accepted;

8120 - 8-10% slower per core than a 1100T which is about 5-8% slower than a 2500K per core. The FX 8120 - 22% faster IMC than the 1100T but about 25% slower than Intels IMC. FX 8120 faster than both the 1100T and 2500K in core crunching multithreading by some distance but also behind in single core execution.

I would be happy if the median is split; That is 5-8% faster per core than a Thuban and around 8-10% slower than a SB/IB. Also I would be happy with about a 15% difference in IMC.

Powerwise the 8120 stock is about 5% more efficient than the 1100T but neither is particularly good relative to Intel parts so any help here is acceptable.

 

[cgner]

Guys we can bounce off the walls at each other all day long and nothing will be gained, appart from early reports of Piledriver changes over Zambezi nobody knows anything, AMD for one have been very tight lipped about anything so right now its just complete speculation. What did come out was AMD stated they are targetting at the very least a 15% performance per clock improvement per release and they did indicate they were quite pleased with Vishera's progress.

As to the ultimate king of processors, probably not, probably never again when you consider AMD's level of technology compared with Intel it is highly unforeseeable that AMD will ever defeat Goliath in a matchup of pure brute force, but in a article released on THG recently about AMD "Vision" AMD have stated themselves they are not looking to slug it out on one aspect of CPU architecture. The future is not single threaded performance.

So we have gone on for 129 pages highlighting the only negitive on Zambezi but what about the promise, how about a little love 😛.

well its a discussion page, so when nothing is revealed, we at least speculate 😛

for games no Read above, there was a link to a very interesting Intel archi article.

http://ixbtlabs.com/articles3/cpu/intel-ci7-123gen-p3.html

and the verdict is 136 fps to 142 woohoo 4.4%

Even at its highest gains, ivy is barely hitting 15% faster than nehalem in its video encoding. the overall difference was 12%.




see above, even Intel can't make a 10% jump for games through 4 archies. Nahalem, Lynnfield (not shown), sandy bridge, and ivy bridge. If your saying 30% flat out ... you will fail.

totally. If AMD does not disappoint this time, and actually delivers 15% improvement, they will be on par.

In brute performance, no. Remember to consider that they have also dropped their TDP by 48 watts while making performance improvements, however minimal. That considered, efficiency is WAY up in those four generations.

ok that still does not help gamers who have lag with FX 4100 cpu's. Efficiency is of secondary concern to gamers.

 

[fazers_on_stun]

^ 3.6GHz + 10% ~= 4.0GHz, so such a clock bump would indeed be about 10%..

 

[sarinaide]

About 10.5%. It will need to split the median between 1100T and I5/I7 in some respects, anything more will be a bonus.

 

[cgner]

^ 3.6GHz + 10% ~= 4.0GHz, so such a clock bump would indeed be about 10%..

Zambezi chips were underclocked from the start. I don't know if it was a marketing trick to leave room for "improvement" but that bump would not be bad. Kinda what intel has been doing for a few years now.

About 10.5%. It will need to split the median between 1100T and I5/I7 in some respects, anything more will be a bonus.

It would need to outdo 1100T in single core performance by a healthy margin to be a gaming CPU.

 

[viridiancrystal]

Hmm, all I've seen is "15% performance improvement" - nothing about per clock (i.e., IPC). Considering the few changes that PD is gonna have compared to BD, I'd suspect maybe 5% IPC and 10% clock (3.6 ----> 4.0 GHz base clock). If indeed AMD hits its targets that is..

Well we only know what the changes are to the architecture, but we don't know what its performance is. A 400mhz bump at well less power is maybe 5-7% clocks and the rest IPC.

Those that I liaise with suggest the AMD flagship will be around the same per core performance as a i7 960-970 with slightly better power consumption. Memory wise still behind Intel but its a marginally better IMC than on the Zambezi part.

Reading reviews on Guru3D and others a very good review on the 1100T vs 8120 and 1100T vs 2500K in general if accepted;

8120 - 8-10% slower per core than a 1100T which is about 5-8% slower than a 2500K per core. The FX 8120 - 22% faster IMC than the 1100T but about 25% slower than Intels IMC. FX 8120 faster than both the 1100T and 2500K in core crunching multithreading by some distance but also behind in single core execution.

I would be happy if the median is split; That is 5-8% faster per core than a Thuban and around 8-10% slower than a SB/IB. Also I would be happy with about a 15% difference in IMC.

Powerwise the 8120 stock is about 5% more efficient than the 1100T but neither is particularly good relative to Intel parts so any help here is acceptable.

Still confused as to how we are loosing IPC from Trinity -> Piledriver

well its a discussion page, so when nothing is revealed, we at least speculate 😛



totally. If AMD does not disappoint this time, and actually delivers 15% improvement, they will be on par.




ok that still does not help gamers who have lag with FX 4100 cpu's. Efficiency is of secondary concern to gamers.

Bringing AMD into the mix changes thing a little bit. I was only talking Intel 😛.

 

[esrever]

Still confused as to how we are loosing IPC from Trinity -> Piledriver

Some people will think Vishera will be worse than Zambezi just because they want it for some reason.

 

[sarinaide]

Nobody really knows what DT trinity or Piledriver will bring IPC wise, maybe its 5% maybe its 15%.

 

[-Fran-]

Not to crack your heads and all, but...

All the speculation going around is "stock", right?

Put OC in the equation and AMD is left even further behind. The OC potential from the 2500k, 2600/2700k and the IB siblings is still way more than AMD's best efforst in terms of perf per watt and even perf alone. I won't say anything about SB-E, cause that thing is in another planet, haha.

Anyway, my point is: AMD might not be playing in the "enthusiast" arena anymore, but they still will have to play the "best bang for the buck" when introducing PD. And this means, price below 300 for the 83xx and just a tad better perf per watt number to be considered a "good product". We all know that, to beat SB or IB they'd need a miracle, since those can reach 4.5Ghz with ease without needing a nuclear reactor in your house.

So... What gives?

Cheers!

 

[obsama1]

THe 8350 is at 4GHz stock, so getting to 4.5GHz should be trivial.

 

[-Fran-]

Yes, that's going to be expected as a "reasonable OC" for PD (at least, on my part), but the IPC in PD better show that, with 4.5Ghz, it will be better than a PhII at 4Ghz (since that is the "reasonable OC" for them).

I really hope Toms does an in-depth clock per clock comparison across Ph, PhII, Zambezi, Vishera, Westmere (maybe, since they're hard to OC), Sandy and Ivy processors. And Llano with Trinity as well, for kicks.

I'm sure that a lot of other folks around would love to see the big OC number PD brings to the table and how that performance stands against a heavily OCed PhII.

Power figures would be interesting for a fine tune in Perf per Watt (like they did with some CPUs a while ago).

Cheers!

 

[obsama1]

PD is going to look pretty good. Not going to kill IB, but probably 5-10% better than PII. I use Intel, but I know all of us with our AMD brethren are waiting with bated breath.

 

[jdwii]

PD is going to look pretty good. Not going to kill IB, but probably 5-10% better than PII. I use Intel, but I know all of us with our AMD brethren are waiting with bated breath.

I'm going to wait until steamy is out i'm a pretty big casual gamer and honestly my 1100t is just fine i use my A8 laptop and its fine even. When i do real work on my desktop(encoding) its real fast.

 

[king smp]

I am very happy with my 4ghz 1100T and plan to wait
the tempation for Ivy Bridge is strong but not needed or justifiable

 

[skitz9417]

i hope i can upgrade to an fx pd cpu if they are any good so i can upgrade for a anm phemon x4 840

 

[mayankleoboy1]

where is the Kepler thread ?

http://www.phoronix.com/scan.php?page=news_item&px=MTE3OTA

The opensource Nvidia driver starts basic support of GK110. Since Nvidia does not officially suppport theses opensource drivers, it means that the devs have got hold of a card from somehwere.......


Edit, OT : anyone else getting sick of the huge iPhone5 discussions and speculations on the web ?

 

[esrever]

where is the Kepler thread ?

http://www.phoronix.com/scan.php?page=news_item&px=MTE3OTA

The opensource Nvidia driver starts basic support of GK110. Since Nvidia does not officially suppport theses opensource drivers, it means that the devs have got hold of a card from somehwere.......


Edit, OT : anyone else getting sick of the huge iPhone5 discussions and speculations on the web ?

kepler thread got canned since it was too much flaming.

 

[fazers_on_stun]

Nobody really knows what DT trinity or Piledriver will bring IPC wise, maybe its 5% maybe its 15%.

If it's 15% then total improvement would be at least 25%, due to the extra 10% of the higher stock clock.

Unless AMD marketing is a total fail then I wouldda thunk they'd be shouting that from every ad and rooftop 😀

 

[cgner]

Not to crack your heads and all, but...

All the speculation going around is "stock", right?

Put OC in the equation and AMD is left even further behind. The OC potential from the 2500k, 2600/2700k and the IB siblings is still way more than AMD's best efforst in terms of perf per watt and even perf alone. I won't say anything about SB-E, cause that thing is in another planet, haha.

Anyway, my point is: AMD might not be playing in the "enthusiast" arena anymore, but they still will have to play the "best bang for the buck" when introducing PD. And this means, price below 300 for the 83xx and just a tad better perf per watt number to be considered a "good product". We all know that, to beat SB or IB they'd need a miracle, since those can reach 4.5Ghz with ease without needing a nuclear reactor in your house.

So... What gives?

Cheers!

THe 8350 is at 4GHz stock, so getting to 4.5GHz should be trivial.

Yes, that's going to be expected as a "reasonable OC" for PD (at least, on my part), but the IPC in PD better show that, with 4.5Ghz, it will be better than a PhII at 4Ghz (since that is the "reasonable OC" for them).

I really hope Toms does an in-depth clock per clock comparison across Ph, PhII, Zambezi, Vishera, Westmere (maybe, since they're hard to OC), Sandy and Ivy processors. And Llano with Trinity as well, for kicks.

I'm sure that a lot of other folks around would love to see the big OC number PD brings to the table and how that performance stands against a heavily OCed PhII.

Power figures would be interesting for a fine tune in Perf per Watt (like they did with some CPUs a while ago).

Cheers!

You folks need to remember that stock clocks do not necessarily raise overclock walls. For example there are Phenom II CPU's that come with 3.7 ghz and those that come with 3.0 ghz from the factory. The chips are virtually identical though, so they both top out around 4.2 ghz.