AMD "Piledriver" refresh of Zambezi - info, speculations, test, fans

[demonkevy666]

Must be some bug in those tests. I do not belive that it can take on 2600K on any of CPU tests.

look again it not even close in the 2600K infact even the FX 4100 is faster then it 10% (I hope I'm not correct).... lack of L3 cache 10% slower seems kind of a high?

[informal]

look again it not even close in the 2600K infact even the FX 4100 is faster then it 10% (I hope I'm not correct).... lack of L3 cache 10% slower seems kind of a high?

FX4100 scores ~4000 pts in passmark. The trinity ES that was listed in the link works at 2.3Ghz with 3.2Ghz maximum Turbo and scores 5390pts. If we take that Trinity almost always runs at max. Turbo and that FX4100 has average clock of 3.7Ghz this leaves us with 6230pts for similarly clocked Trinity APU. This is approx. 50% better which is a bit on a high side IMO. Unless the test hit some bottleneck in BDver1,Trinity shouldn't be performing so much better than current Bulldozer parts. For comparison, 8150 that has 2x more cores/threads than FX4100 scores ~2x more : 8246pts. If we go by this ,a Trinity FX 8C part with similar clocks as 8150 should score more than 10500 pts... I doubt this.

[Kristoferr]

look again it not even close in the 2600K infact even the FX 4100 is faster then it 10% (I hope I'm not correct).... lack of L3 cache 10% slower seems kind of a high?

Check integer math. Thats the bug i was talking about.

FX4100 scores ~4000 pts in passmark. The trinity ES that was listed in the link works at 2.3Ghz with 3.2Ghz maximum Turbo and scores 5390pts. If we take that Trinity almost always runs at max. Turbo and that FX4100 has average clock of 3.7Ghz this leaves us with 6230pts for similarly clocked Trinity APU. This is approx. 50% better which is a bit on a high side IMO. Unless the test hit some bottleneck in BDver1,Trinity shouldn't be performing so much better than current Bulldozer parts. For comparison, 8150 that has 2x more cores/threads than FX4100 scores ~2x more : 8246pts. If we go by this ,a Trinity FX 8C part with similar clocks as 8150 should score more than 10500 pts... I doubt this.

This.

[ilkkahy]

So clear from cpu side a FAILDOZER.
AMD in 9 months was not capable to make some tweaks to Llano sillicon to make it do 3.2-3.6ghz(like Phenom II) with overclock to 4-4.2ghz. It should have been a much better choice.
Llano oc >= Trinity oc.


FAIL !!!

http://hwbot.org/submission/2193039_...50d_8027_marks
A8-3850 @3.321Mhz - 4577.

No wonder why Romania is such fail of a country if even smart people there express themselves with such idiotic manners. I dont necessarily disagree what you say but i think that the way you say things annoys a lot of people here. Do us all a favor and learn some manners please. But I guess you propably wont...

[haylui]

Check integer math. Thats the bug i was talking about.




This.

perhaps there is bug.
But I don't understand why CPU Floating Math, Phenom II x4 965 is higher than i7?

[informal]

perhaps there is bug.
But I don't understand why CPU Floating Math, Phenom II x4 965 is higher than i7?

Because Deneb has a very robust FPU. The problem is that it's not multithreaded so it loses to i7's SMT capabilities in MT aware software.
Bulldozer remedies this somewhat since it has 4 FP units that are 2 way SMT capable. That's why "8 core"(integer core) Bulldozer has a ~SB level performance in FP intensive workloads. 4 FP units vs 4 FP units,both 8 threads capable.

[BeepBeep2]

FX4100 scores ~4000 pts in passmark. The trinity ES that was listed in the link works at 2.3Ghz with 3.2Ghz maximum Turbo and scores 5390pts. If we take that Trinity almost always runs at max. Turbo and that FX4100 has average clock of 3.7Ghz this leaves us with 6230pts for similarly clocked Trinity APU. This is approx. 50% better which is a bit on a high side IMO. Unless the test hit some bottleneck in BDver1,Trinity shouldn't be performing so much better than current Bulldozer parts. For comparison, 8150 that has 2x more cores/threads than FX4100 scores ~2x more : 8246pts. If we go by this ,a Trinity FX 8C part with similar clocks as 8150 should score more than 10500 pts... I doubt this.

First of all Passmark used to be considered a joke of a benchmark, and it really hasn't changed all that much...

Also, are we talking about overall system score? Regardless, I think you're right...

[demonkevy666]

Because Deneb has a very robust FPU. The problem is that it's not multithreaded so it loses to i7's SMT capabilities in MT aware software.
Bulldozer remedies this somewhat since it has 4 FP units that are 2 way SMT capable. That's why "8 core"(integer core) Bulldozer has a ~SB level performance in FP intensive workloads. 4 FP units vs 4 FP units,both 8 threads capable.

I read that sandys where 4 256 bit but can do two at time because of the hyper threading
while bulldozers are 128 bit for 8 and 256 for 4.

[tbone8ty]

http://vr-zone.com/articles/amd-to-s...ll-/15564.html

good read

[Piledriver]

Thats 11% or more CPU points with 3800/4200mz vs 2900mhz on 3850K.
Faildozer all over again.

Yes 12% with about 30% higher core frecuency. And if it ends like BD which meant major power consumption and TDP increase at a slight overclock, what then.
A bad IPC is a bad IPC no mather you put it.
Let say oc on Trinity from 3.8ghz to 4.5ghz, that's about 20% but on Llano performance it's like raising with 7% the frecuency.
Let's compare again so A8 3850 at 3250Mhz(+12%) should equal Trinity at 3.8ghz(don't even count TURBO).
Furthermore A8 3850 at 3500Mhz( 3250Mhz + about 7%) should be equal with Trinity Cpu at 4.5ghz .
The IGP part it's irrelevant, it's clear a good path as in desktop Gpu's.

So clear from cpu side a FAILDOZER.
AMD in 9 months was not capable to make some tweaks to Llano sillicon to make it do 3.2-3.6ghz(like Phenom II) with overclock to 4-4.2ghz. It should have been a much better choice.
Llano oc >= Trinity oc.


FAIL !!!

http://hwbot.org/submission/2193039_...50d_8027_marks
A8-3850 @3.321Mhz - 4577.

Some things are very simple to understand, but then, there are very simple people, and you need to explain those things that are very simple.

So here we go,

IPC means instructions per clock... PER CLOCK... one more time, PER CLOCK. So that means that clock is one important element.

Whining about how trinity it's faster because it has more clock, it's stupidity at his best. I don't understand how people that think they are knowledgeable enough to participate in tech discussions, don't know that high frequencies were in fact AMD's goal.

This is like some company that builds submarines, decides to start building planes, and then you have some moron customers going "Yeah those new planes, they don't travel underwater as good as their submarines"


So Bulldozer need more frequency? No sherlock... and who cares?
I guess if AMD made a 10GHZ CPU that had the absolutely worst IPC ever, but thanks to the high frequency was the best CPU around, you still would have some bright "tech engineers" going, "bu-bu-but at the same frequency it's slower than Lano."

The only thing that matters is... for the same amount of money, buying price and electric bill, Trinity >>>> Lano, in everything, CPU and GPU.

[Andi64]

Everybody here knows that both things matter. The problem is, as you said, BD is a high frequency architecture that won't pass 4.6/4.7Ghz on average, the same as a Sandybridge, while having... idk, 45%? let's say, lower IPC, and consuming a lot more.

They NEED to improve IPC, cause I don't see these airplanes hitting 8Ghz as they need to match the competition, but I actually see them sinking in water

I'm talking mostly over the enthusiast part based on Piledriver, of course.

[Piledriver]

Everybody here knows that both things matter. The problem is, as you said, BD is a high frequency architecture that won't pass 4.6/4.7Ghz on average, the same as a Sandybridge, while having... idk, 45%? let's say, lower IPC, and consuming a lot more.

They NEED to improve IPC, cause I don't see these airplanes hitting 8Ghz as they need to match the competition, but I actually see them sinking in water

I'm talking mostly over the enthusiast part based on Piledriver, of course.

Yes AMD and Intel goal it's to have the best\fastest combination of IPC and clock, and i'm pretty sure they both are always working around the clock to improve their CPU's. Now about the Lano vs Trinity, do you disagree that it's dumb, whining about a CPU being faster because it has more MHZ, when it was designed to be that way? How about bragging that it's faster despite the worse IPC? Works both ways i guess.

[Kristoferr]

Some things are very simple to understand, but then, there are very simple people, and you need to explain those things that are very simple.

So here we go,

IPC means instructions per clock... PER CLOCK... one more time, PER CLOCK. So that means that clock is one important element.

Whining about how trinity it's faster because it has more clock, it's stupidity at his best. I don't understand how people that think they are knowledgeable enough to participate in tech discussions, don't know that high frequencies were in fact AMD's goal.

This is like some company that builds submarines, decides to start building planes, and then you have some moron customers going "Yeah those new planes, they don't travel underwater as good as their submarines"


So Bulldozer need more frequency? No sherlock... and who cares?
I guess if AMD made a 10GHZ CPU that had the absolutely worst IPC ever, but thanks to the high frequency was the best CPU around, you still would have some bright "tech engineers" going, "bu-bu-but at the same frequency it's slower than Lano."

The only thing that matters is... for the same amount of money, buying price and electric bill, Trinity >>>> Lano, in everything, CPU and GPU.

Thats not kindergarden for you. But funny to read tho. Had a good laugh. Thank you for that.

Its true that only thing what matters is perf per watt and perf per cost. No one ever disagreed in the first place. The problem is that Trinity has too low CPU increase over Llano and probably has less OC potencial as well, due higher stock speed.
Im am however still hopeful that Vishera will have 20+% over bulldozer, which will definately be Sucsess with a big S.

[haylui]

Trinity got high score in Integer Math is because of enabling of working hardware divider?
http://www.planet3dnow.de/vbulletin/...d.php?t=404354

[cesariuth]

Official slides for Trinity

nEO_IMG_trinity-1.jpg.jpeg
nEO_IMG_trinity-2.jpg.jpeg
nEO_IMG_trinity-3.jpg.jpeg
nEO_IMG_trinity-4.jpg.jpeg
nEO_IMG_trinity-5.jpg.jpeg
nEO_IMG_trinity-6.jpg.jpeg

http://vr-zone.com/articles/amd-trin...on-/15716.html

[FlanK3r]

Wow, Im looking forward for this...Desktop Trinity will be fun Im thinking and later Vishera...Who knows next improvements

[tbone8ty]

Trinity got high score in Integer Math is because of enabling of working hardware divider?
http://www.planet3dnow.de/vbulletin/...d.php?t=404354

can someone explain this? hard to fallow since its in another language

[informal]

The bug that was triggered by passmark's integer test in case of BD and K10 is not triggered in Piledriver's case. Trinity(PD based) scores much better since the HW divider is running the code. K10 and BD didn't run the code on the divider unit and that's why the score was lower.

[FlanK3r]

JUly in desktop is not possible...,maybe in August or September

[Maxforces]

Never ending delaney

[FlanK3r]

its not delay, it was planned years ago...one year after new platform (APU Q2, desktop CPU Q3/Q4 - now will be maybe one-2 moth earlier than year before, its not bad)

[trueblue]

its not delay, it was planned years ago...one year after new platform (APU Q2, desktop CPU Q3/Q4 - now will be maybe one-2 moth earlier than year before, its not bad)

Does you have info on the rumored new chipsets (1090, 1070 and new southbridge) and their release date?
I have noticed a lot of dropping prices on am3+ boards with many models being deactivated at different sites.

[FlanK3r]

I dont know man...But I wish some Crosshair VI board with some BIOS features as Maximus V Gene (RAM profiles etc)

[Oese]

http://amdfx.blogspot.de/2012/05/amd...a10-4655m.html

dunno if it has been posted before somewhere.. looks good. 2m/4t trinity @ 2.0 ghz 2.3 turbo matches llano 4c @ 1.9 ghz 2.3 turbo (which doesnt work) at 10w less tdp..

[BeepBeep2]

2.3 GHz Trinity matches 1.9 GHz Llano at better TDP? No way!

4.6 Ghz Trinity better match 3.8 GHz X4 then