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 ...

 

[Cazalan]

Such bad timing for AMD. Announcing availability of the new chips on the same day Apple has a slew of product announcements.

 

[gamerk316]

Highlights the problem: At the same speed, PD still stinks compared to IB. PD is basically relying on its speed advantage at this point.

Problem is, PD still doesn't OC any higher then Intel does, so the advantage is limited to stock clocks.

I'm going to say this one final time: The overwhelming majority of applicationd do not, and will not scale. Hence the entire BD architecture, for the desktop, was flawed from the very beginning.

 

[Cazalan]

On to FM2 not only do gaming results with discrete rather show up FX, I found out that Photoshop and Adobe will have regular updates for FM2 which better allocates resources to the iGPU component for performance that makes the fastest x86 processor look antiquated, the future is fusion enough with this FX crap.

Those updates also boost performance with discrete GPUs. It's not just for APUs. It helps the HPC market which NVidia/AMD/Intel are fighting for.

The future is fusion, but it still has several generations to catch even a $100 discrete.

 

[de5_Roy]

^^ isn't that the software developers' responsibility? i mean, to make the software scale to higher number of core within the coding limitations.
edit: @gamerk316's post.

 

[viridiancrystal]

I'm fairly impressed by Piledriver thus far. It competes well with Ivy i5/i7 in most applications, the exception of course being single threaded workloads. Gaming is a nice step forward, It beat out the PII 980 in every bench on the Tom's review, I think.

The important thing is that PD is better in every single aspect over BD. There isn't a single thing that is worse about it. IPC, clocks, heat, power, It was all improved overall.

Glad to see some better pricing at launch this time. The 4 core variants will be very good options over i3s, considering their gaming performance will be almost identical to the 8350 in most games.

I'm going to hold off this gen, mostly because I cant afford much of anything, and I would rather work around what I have now for the long term (Sound card and SSD will be next, just because of their longevity and constant benefits). I may go for Steamy if it is on AM3+.

Hours after we find out how PD is, we all already look forward to the next arch. True enthusiasts.

 

[noob2222]

anandtech benches 😉
8530 vs 1090t
cinebench r11.5
single 1.1 vs 1.08
multi 6.89 vs 5.72
so it is near ~20% better than 1090t in multi threading, thats as per my prediction 😉 😀

2500k vs 8350
http://www.anandtech.com/bench/Product/697?vs=288

3220 vs 4300
http://www.anandtech.com/bench/Product/700?vs=677
Cinebench 11.5 - Multi-Threaded :- 3.24 vs 3.29 😱 (i need an eye specialist :heink)

but here comes the worst one
http://www.twcableuntangled.com/wp-content/uploads/2010/09/shocked-kid-300x225.jpg
5800k vs 4300
http://www.anandtech.com/bench/Product/700?vs=675
Cinebench 11.5 - Multi-Threaded :- 3.24 vs 3.31 😱 (i need a psychiatrist 😱 )
http://kinleycain.com/wp-content/uploads/2012/05/shocked_cat1.jpg

Don't get stuck on cinebench, you can run it 5 times and come up with miniscule differences. Instead on he 5800k vs 4300, look at the games...

DAO - 116 to 97
DoW II - 64 to 51
WoW - 81 to 73
SC II - 43 to 41

Does l3 cache help? 4-25% improvement.

So whats one possible explanation on cinebench?

We've always done these processor graphics performance comparisons using DDR3-1866

APUs get faster memory because of the IGP.

 

[esrever]

FM2 is ok for the low end but it's locked at 4 cores.
You need the 6 and 8C variants to get closer to i5/i7 performance.

They can't scale the iGPU much more without a major die shrink and that's 2 years away. This is why Steamroller is focused on CPU enhancements (fixes) over GPU increase.

FM2 makes perfect sense for laptop/netbook/AIO but no sense for desktop.

Kaveri is going to have gcn and unified memory space which will should help with gpgpu and general graphics performance. It is expected to have steamroller cores and built on 28nm instead of 32nm. They can up the transistor count by 1.3 and be in the same die area and power consumption range.

 

[noob2222]

http://hardocp.com/images/articles/1350948339W82hNIgGd3_3_3.png

http://hardocp.com/images/articles/1350948339W82hNIgGd3_3_4.png

Highlights the problem: At the same speed, PD still stinks compared to IB. PD is basically relying on its speed advantage at this point.

Problem is, PD still doesn't OC any higher then Intel does, so the advantage is limited to stock clocks.

I'm going to say this one final time: The overwhelming majority of applicationd do not, and will not scale. Hence the entire BD architecture, for the desktop, was flawed from the very beginning.

... ya ok, .. 5.2 ghz is really the same speed as 4.7 ghz ... last I checked, thats 500 hmz faster.

http://www.overclockersclub.com/reviews/amd_fx8350/3.htm

 

[blackkstar]

http://hardocp.com/images/articles/1350948339W82hNIgGd3_3_3.png

http://hardocp.com/images/articles/1350948339W82hNIgGd3_3_4.png

Highlights the problem: At the same speed, PD still stinks compared to IB. PD is basically relying on its speed advantage at this point.

Problem is, PD still doesn't OC any higher then Intel does, so the advantage is limited to stock clocks.

I'm going to say this one final time: The overwhelming majority of applicationd do not, and will not scale. Hence the entire BD architecture, for the desktop, was flawed from the very beginning.

The [H] review is god awful. I hate to use the dreaded car analogy, but it's kind of like this. AMD has an engine that makes power at higher RPM and Intel has one that makes power at lower RPM. This is like running the engines at the same RPM and saying the one that needs the higher RPM blows. [H]ardOCP really hates AMD and I am absolutely not surprised to see them doing something like this.

Lets put an FX 8350 at 5.2ghz and compare it to an Ivy at max clocks of 4.6ghz. I would imagine things would look much better for the AMD, although not the greatest.

[H]ardOCP is *** and it always has been and it always will be. I got banned from their forums when they released the 7970 review since they compared it to an overclocked GTX 580 that cost a lot more than the 7970 and then said "7970 is only 15% faster than GTX 580 IT SUCKS!" and I called them out. Appearantly, being a member for 4 years means nothing if you call them out on their bullshit and they have no qualms with banning you if you call them out on anything.

I have a really good feeling that Kyle and Brent are laying down banhammers to anyone who says that FX clocks higher than Ivy and they're not meant to be compared clock for clock; also that it's BS to compare an overclocked CPU to a stock one and then say the stock one sucks.

[H]ardOCP and [H]ardForum is a place for circle jerking wankers to all jerk each other off about how good they feel that they bought overpriced hardware that they're never going to use to it's full potential. Example of [H] shitposting: http://hardforum.com/showthread.php?t=1723285

I don't know if you guys can pick up on it, but I ******* hate [H], Kyle, Brent, and the majority of the *** posting retards in their forum.

 

[noob2222]

As biased as Xbit is they did a decent job of looking at what PD changed, looks like primarily a tweak to the FPU

 

[viridiancrystal]

The [H] review is god awful. I hate to use the dreaded car analogy, but it's kind of like this. AMD has an engine that makes power at higher RPM and Intel has one that makes power at lower RPM. This is like running the engines at the same RPM and saying the one that needs the higher RPM blows. [H]ardOCP really hates AMD and I am absolutely not surprised to see them doing something like this.

Lets put an FX 8350 at 5.2ghz and compare it to an Ivy at max clocks of 4.6ghz. I would imagine things would look much better for the AMD, although not the greatest.

[H]ardOCP is *** and it always has been and it always will be. I got banned from their forums when they released the 7970 review since they compared it to an overclocked GTX 580 that cost a lot more than the 7970 and then said "7970 is only 15% faster than GTX 580 IT SUCKS!" and I called them out. Appearantly, being a member for 4 years means nothing if you call them out on their bullshit and they have no qualms with banning you if you call them out on anything.

I have a really good feeling that Kyle and Brent are laying down banhammers to anyone who says that FX clocks higher than Ivy and they're not meant to be compared clock for clock; also that it's BS to compare an overclocked CPU to a stock one and then say the stock one sucks.

[H]ardOCP and [H]ardForum is a place for circle jerking wankers to all jerk each other off about how good they feel that they bought overpriced hardware that they're never going to use to it's full potential. Example of [H] shitposting: http://hardforum.com/showthread.php?t=1723285

I don't know if you guys can pick up on it, but I ******* hate [H], Kyle, Brent, and the majority of the *** posting retards in their forum.

It would fare worse, at stock the clock difference is 500Mhz, and raising them both only slims the % between them, not to mention the widening power gap.

 

[Cazalan]

[H]ardOCP is *** and it always has been and it always will be.

That [H] review was just really odd. Two graphs for each family of processor. Not taking price into account at all. They used to write some decent articles there, now it's just meh.

Vishera isn't stellar but I think some chips will make it into the "Best build under XYZ $" lists.

 

[tonync_01]

Still AMD must discontinue AM3+ and put all the money into APU's for DT and Mobility, as the architecture evolves so will the APU. Kaviri APU's expected to have nearly 3x the bandwidth on the IGPU, faster memory, possible on die memory, revised architecture focusing heavily on core efficiency and latencies, it is very possible that the APU may even beat a FX Steam Roller part in pure gaming metrics, which is rather scary. I sold my AM3+ with the view that its long gone and dead to me now it has no market when a A10 is just as good on half the power.

I agree with this as long as they start making the 8 and 6-core APUs. Is there any info on whether this will occur with steamroller or excavator?

 

[-Fran-]

can't happen and won't happen, not enough room on the chip..
another process shrink or two perhaps.?
(22nm on the same size PD chip is what I meant..)

If what AMD said about the plans on SR are true, then they'll fuse most of the GPU with the CPU, so the necessary die will be less.

SR was supposed to be made in 28nm as well, right?

Cheers!

 

[cgner]

Phenom II from 2009 still pwnz it in single core performance. looooooooooooooooooooool

 

[dogman_1234]

Just heard about PD today. Finally came out!

 

[jdwii]

Well kaviri is supposed to be 28nm so they will have more space for the iGPU and the CPU

PD seems to be quite an improvement over BD but not worth the upgrade, I don't see a 10-15% increase in performance a necessary upgrade.

I will get one of this once they are available in my country and give my brother my old Phenom 9650 that in a lot of games struggles to get over 35 FPS while having CPU core at 100%

Like watching the videos and reading over 10 sites full of benchmarks the 8350 is easily 15-20% better then my CPU out of the box while using a little less power consumption what's weird is gaming is one area they did a lot better in and you can clearly see a 10-15% improvement which probably has something to do with FP benchmarks that show a 15-20% improvement. Piledriver is definitely Amd's processor in all accounts this time. Like i said Amd only said a 10-15% improvement anyways which is more than what we saw with the Phenom II x6 1100T to 8150 IMHO.

 

[jdwii]

So, what's the consensus on performance improvement over BD?

S/A says it's 15% total, so that would be 5% IPC and 10% clocks, which is what I said a bunch of pages back 😀..

Actually clock speed is not always representative to performance i would say Clock speed didn't make a whole 10% boost in performance. IPC went up 7-10%(on average) as posted with several benchmarks. Kinda like clocking my CPU to 3.9Ghz did not make it 18% faster on handbrake(program that stresses CPU 95+%) it made it more like 10% faster. There's reasons why this is true if i could take a guess it would be the Branch prediction hit rate shrinks as clock rates go up.

 

[amdfangirl]

Just heard about PD today. Finally came out!

That's what RSS feeds are for.

 

[-Fran-]

Finally!

New CPU!

...

An i7 2700k at $220usd in Microcenter with a $50usd discount in MoBos (Asus P8Z77-V PRO for me at $150usd) =X!

Cheers! xD!

 

[cgner]

^ewwww

 

[amdfangirl]

Finally!

New CPU!

...

An i7 2700k at $220usd in Microcenter with a $50usd discount in MoBos (Asus P8Z77-V PRO for me at $150usd) =X!

Cheers! xD!

Very nice .

 

[sookster54]

I agree with this as long as they start making the 8 and 6-core APUs. Is there any info on whether this will occur with steamroller or excavator?

6 core APU might be a possibility, but not anymore than that. Honestly for MOST things in today's world 4 cores is plenty. I like the APU but the Phenom II X6 still rocks it, I'm probably going with the i5 3570K and make the APU an HTPC after new year's.

 

[sarinaide]

Steamroller is dedicated to massive IPC and latency overhauls in architecture so a 4 Core APU is more than enough, its not the cores but the integration between the CPU and GPU parts....again talk of double the iGPU bandwidth and better which is what will deliver the ultimate performance. I have a good feeling that sub $120 cards will soon be rendered obselete.

 

[earl45]

Infaltion is nothing more then the result of increased consumer spending. More money being spent justifies highter prices for goods and services. Lack of inflation is one of the first indicators of a recessionary period.

Point being, if only a small portion of the populatuon increases spending, you get inflation, which reduces the buying power of everyone else. Hence why rising wages are a necessity for a growin economy, as you'll stagnate quickly if wages don't follow inflation.



Sorry, thats a BS argument that is not backed by historical facts. Its an argument made by whichever political party is attempting to gain control of government that has no basis in facts. This is especially notable, since the President has little to no power to directly affect the economy (theres this thing called Congress).

Regarding supply and demand, there is currently low demand, thus supply is or will be low, thus there are lay-offs. Case in point to get back on topic, AMD will be downsizing 15% of it's global workforce due to the slowdown in PC sales as tablets are becoming more popular and smartphone sales are on the rise. To a lesser extent, Intel is laying of a "small percentage" of it's McAfee division.

Hence why any attempts to solve the economic problems by making business more profitable is doomed to fail. Due to lack of demand, business will just keep the profits, rather then hire to meet demand that currently doesn't exist.

+1