Desktop Trinity benchmarks are up.

[pantsaregood]

Ya know what else I can't use Smiles because I not allowed smilies I bad to the bone and a hacking fool. I take it you wanted me to highlight the quote and than press the quote selected text . I choose not to do it that way as your already complaining about my structure. There are no forum rules on structure . So just for you a copied and pasted your quote to add to your difficulties in reading my post . I bet your going to say you can read the bible and comprehend its meaning . Not likely. Your frequency is vary low you need more love in your life and less fear. You may call me any dang thing you like it has no effect on me. Its not like I hate AMD products . I hate AMD management I hate IBM for forcing intels hand at allowing AMD x86. IBM the monoply so big government stays away from them. A company so powerful they had the US gooberment break up AT&T just in time to make IBM the BIG Player in the Computer market. The company that supposedly invented the transitor in 1946(Bell Labs) . Cough at that date. 47 more likely. Roswell in 1947. After AT&T was broken up my phone bill doubled Thanks for nothing USA gooberment.

are you real

 

[inf64]

are you real

Oh he is very real.Unfortunately for this forum.
Best advice is : ignore him.

 

[NostaSeronx]

Sadly, Bulldozer will meets its demise after Steamroller. Jaguar-derivates will replace Bulldozer.

AMD FX(2011) -> 8 cores
AMD A(2012) -> 4 cores
AMD FX(2012) -> 8 cores
AMD A(2013) -> 4 cores
AMD FX(2013) -> 10 cores
AMD A(2014) -> 8 cores(8 Jaguar+ cores)
AMD FX(2014) -> 20 cores(20 Jaguar+ cores)
AMD A(2015) -> 12 cores(12 Jaguar-derived cores)
AMD FX(2015) -> 32 cores(32 Jaguar-derived cores)

---
Don't ask I'm NostaSeronx.

 

[podspi]

I don't think so. I think AMD has already proved the point that more, weaker cores are not the way to go.

 

[NostaSeronx]

I don't think so. I think AMD has already proved the point that more, weaker cores are not the way to go.

Bobcat in no way is a weak core.

It's the only AMD Architecture that clock to clock matches to Intel architectures. I'm surprised no one tried 1 module/1.6 GHz vs 2 core/1.6 GHz.

Bulldozer Module + L2 => 30.9 mm² (32nm)
Bobcat Dual Core + L2 => ~16 mm² (40nm)

It's kind of obvious that Bobcat will go the route of Merom. History repeats.

---
E-350 vs FX-8150 clock to clock/core to core Passmark via algorithms
CPU Mark (Composite Average)
E-350: ~800
FX-8150: ~900
Integer Math (Millions of Operations per second)
E-350: ~150
FX-8150: ~170
Floating Point Math (Millions of Operations per second)
E-350: ~640
FX-8150: ~600
Find Prime Numbers (Thousands of Primes per second)
E-350: ~150
FX-8150: ~180
SSE / Multimedia (Million Matrices per Second)
E-350: ~3.5
FX-8150: ~3.4
Compression (kBytes processed per second)
E-350: ~930
FX-8150: ~1190
Encryption (MBytes transferred per second)
E-350: ~3.1
FX-8150: ~3.2
Physics (Frames per Second)
E-350: ~50
FX-8150: ~60
String Sorting (Thousands of String per second)
E-350: ~600
FX-8150: ~800

E-350 2c 1.6 GHz/DDR3 1066 (7-7-7-20)? MHz/Labtop - ~9Ws 2c CPU 40-nm
FX-8150 8c 1.6 GHz/DDR3 1333 (7-7-7-21)? MHz/Desktop - 13-18W? 2c CPU 32-nm
---

Only a few people know the changes of Jaguar... http://www.hotchips.org/ but it will make an appearance at HotChips

 

[Riek]

but you forget SSE, AVX, FMA, XOP, and other instruction support.
you forget BD is designed for higher frequencies while bobcat is not so running at the same frequency is unfair.
Look at the 17-18W parts... to compare if you'd like.

 

[NostaSeronx]

but you forget SSE, AVX, FMA, XOP, and other instruction support.
you forget BD is designed for higher frequencies while bobcat is not so running at the same frequency is unfair.
Look at the 17-18W parts... to compare if you'd like.

They have the same pipeline length...15 stage pipeline.

Bulldozer was aimed only at high performance.
Bobcat was aimed at low power.

Knowing FD-SOI we all know Low Power is also good at High Power.

Bobcat supports up to SSE3. I said Jaguar is the derived architecture not Bobcat. I would guess Jaguar will at least add support for SSE4.1/SSE4.2/128b AVX/128b XOP. FMA has limited uses for the general market of Jaguar but the Jaguar-derived architecture that will replace Steamroller might use it.

Pentium D(NetBurst) ~3.2 GHz(95W TDP) got replaced by Core 2 ~2.3 GHz(65W TDP) which is a 0.9 GHz difference. Intel then said 40% less power for 40% more performance than Pentium D. In the real world that placed Intel's Core 2 right next to, to better than Athlon X2.

 

[IntelUser2000]

Bobcat in no way is a weak core.

It's the only AMD Architecture that clock to clock matches to Intel architectures. I'm surprised no one tried 1 module/1.6 GHz vs 2 core/1.6 GHz.

Uhh, no to both.

Bobcat is 10-15% behind the original K8, aka the Athlon 64. Bulldozer is 10-15% behind Deneb/Thuban, which is based on the enhanced K8L.

Meaning there's ~20% difference in IPC, which isn't trivial at all in CPU architectures. Bulldozer can also clock nearly 3x higher. While its true Bobcat has smaller die size and uses less power, scaled down enhanced Bulldozer can do just as well in power department. That's called Trinity A6. IPC-wise against Intel chips, Bobcat would be 30-50% behind Conroe based Celeron, which again is 30-50% behind Ivy Bridge.

(BTW Bulldozer has an ~18 stage pipeline)

Trinity A6: 17W TDP, with 2 cores clocking 50-100% higher than Bobcat.

TSMC's 40nm also may look worse in density than AMD's 32nm. In reality its not, because they are one of the best for density. Lower power/lower frequency architectures also tend to have smaller die sizes.

Because Bobcat and Bulldozer is so different, comparing the two is like comparing Taxis and Bicycles.

 

[NostaSeronx]

Bobcat is 10-15% behind the original K8, aka the Athlon 64. Bulldozer is 10-15% behind Deneb/Thuban, which is based on the enhanced K8L.

I used Passmark it says differ.

(BTW Bulldozer has an ~18 stage pipeline)

15 stage..The late Chuck Moore told us already.(and Mike Bulter).

Trinity A6: 17W TDP, with 2 cores clocking 50-100% higher than Bobcat.

E2-1800 2 cores/1.7 GHz on 40-nm vs A6-4455M 2 cores/2.1 GHz on 32-nm. Yah, 50-100% higher clock...

TSMC's 40nm also may look worse in density than AMD's 32nm. In reality its not, because they are one of the best for density. Lower power/lower frequency architectures also tend to have smaller die sizes.

Uh, two bobcat cores + each l2 is ~16 mm² while one trinity module + l2 is ~30.9 mm² and they provide about the same performance around the same clock.

I'm saying Jaguar-derived architecture will replace Steamroller's successor.

If Bobcat was put on 28-nm... straight die shrink no Jaguar...
-Bobcat 28-nm-
~8mm² (2 cores + L2)
~2-2.5 GHz

Up to SSSE3 performance clock to clock/core to core Bobcat should be within ~5% of Bulldozer.

 

[Arkadrel]

what I noticed:


1) Support for AVX helps bolster Trinity’s floating-point performance.
Very noticeable in the Sandra 2012 benchmarks.
Its "Int. Mpix/s" basically trippled (x3) from Llano, while its Float Mpix/s score whent up 25%.

2) Trinity includes acceleration for AES encryption and decryption.
Again performance shoots up (x5) (500%) while doing AES256 (GB/s) bench in Sandra 2012.

3) Slight performance boost in CPU bound tasks across the board (not by huge amounts)
Not by huge amounts.... around 10% or so compaired to Llano.

4) Decent performance boost in GPU bound tasks (such as games)
Usually around the ~25% improvement compaired to Llano.

Diablo III @1920x1080 Low Quality settings, no AA ~40 fps.
WoW:Cataclysm (64bit) @1920x1080 Good Quality settings, no AA ~55 fps.
Batman:Arkham City @1920x1080 Good Quality settings, no AA ~31 fps.

This is with DDR3-1600's.... and these games are just borderline playable at 1920x resolution.
I guess faster ram could help, or lowering quality abit.

That said, still impressive to be able to play games at 1920x resolution on a APU's IGP.


I look forwards to see people overclock the IGP and CPU and giveing them fast ram,
just to see how far those IGP's can go when pushed

 

[IntelUser2000]

I used Passmark it says differ.15 stage..The late Chuck Moore told us already.(and Mike Bulter).E2-1800 2 cores/1.7 GHz on 40-nm vs A6-4455M 2 cores/2.1 GHz on 32-nm. Yah, 50-100% higher clock...Uh, two bobcat cores + each l2 is ~16 mm² while one trinity module + l2 is ~30.9 mm² and they provide about the same performance around the same clock.

I'm saying Jaguar-derived architecture will replace Steamroller's successor.

If Bobcat was put on 28-nm... straight die shrink no Jaguar...
-Bobcat 28-nm-
~8mm² (2 cores + L2)
~2-2.5 GHz

Up to SSSE3 performance clock to clock/core to core Bobcat should be within ~5% of Bulldozer.

-I stand corrected on the pipeline stages
-Passmark is horrible, Anandtech benches show Bulldozer has about ~20% IPC advantage over Bobcat
-A6 supports Turbo, that would put it still well above 30%
-As I said, lower speed architectures can be density optimized and TSMC is one of the best out there. That size likely won't scale if they make a version they clock at 3+ GHz.

 

[inf64]

Bobcat won't replace Bulldozer. AMD's plan for Bulldozer is public and the 2014-2015 generation is code named Excavator. That is not a Bobcat derivative but a Bulldozer one.

 

[Arkaign]

-I stand corrected on the pipeline stages
-Passmark is horrible, Anandtech benches show Bulldozer has about ~20% IPC advantage over Bobcat
-A6 supports Turbo, that would put it still well above 30%
-As I said, lower speed architectures can be density optimized and TSMC is one of the best out there. That size likely won't scale if they make a version they clock at 3+ GHz.

Passmark is indeed utterly useless. It's about as bad as Windows Experience Index, lol.

On topic, this Trinity still seems irrelevant to the desktop.

 

[pelov]

From [H]

Mike Butler, Senior Fellow Design Engineer, AMD - The latest architectural advancements from both AMD and our competitors have incorporated advancements from deeper pipelines. The pipeline within our latest "Bulldozer" microarchitecture is approximately 25 percent deeper than that of the previous generation architectures. That deeper pipeline is a key technology advancement, providing record breaking frequencies and performance improvements.

Additionally, the "Bulldozer" design inherently runs at a higher frequency for a given voltage than an alternative design would, and is thus a more power-efficient way of delivering performance – and we expect that performance will scale over time and as process maturity gains are realized.

I'm not sure whether that's a good thing or a bad thing.

While AMD has not officially disclosed Bulldozer pipeline length, from the misprediction latency it seems that it is about 50% longer then PhenomII one. In order to maintain acceptable performance, AMD significantly improved its branch prediction logic.

http://www.ilsistemista.net/index.p...n-whats-wrong-with-amd-bulldozer.html?start=7

You'd think it's much higher given the really bad misprediction penalties Johan pinned down in the server and architectural review of Zambezi.

12 ALU stages and 18 FPU on K10.

The exact length of the pipelines is not known but it can be inferred that it has approximatelytwelve stages, based on the fact that the branch misprediction penalty is measured to 12clock cycles

http://www.scribd.com/doc/86649302/130/The-pipeline-in-AMD-processors

In short, despite the not-really-that-long pipeline (comparatively speaking as SB actually has a roughly equal pipeline length at ~15 [slightly higher for FPU]) so it's not necessarily the pipeline that's the issue but rather everything else

Trinity does look good, though. I was kind of surprised to see that it actually performs better as a desktop part than it does a laptop chip. Trinity looked pretty good in the mobile review but seeing the desktop review is actually a pleasant surprise. The SB/IB i3's cant go kicking it around as far as synthetic benchmarks go and Trinity actually surpasses it in multi-threaded scenarios. The on-die GPU is really awesome, though. 1080p gaming at moderate settings on an APU is incredible and Kaveri should be even better.

Don't compare Jaguar/Bobcat to Trinity/Steamroller, especially in TDP. The bobcat cores are great, but they also lack the graphical power of a Trinity APU, modern ISAs and they're based on a different architecture on a different process on a different node. You can't get any more... uh... different Scaling Bobcat up might not be feasible and most definitely does not guarantee any significant pushes in clock speed (Bulldozer is a prime example of that). Clock speed gains via shrink is something that we toss around like it ain't no thang but it's actually far more complicated than we assume and takes into account a whole variety of factors. It most definitely isn't a sure thing.

 

[podspi]

Bobcat in no way is a weak core.

It's the only AMD Architecture that clock to clock matches to Intel architectures. I'm surprised no one tried 1 module/1.6 GHz vs 2 core/1.6 GHz.

Bulldozer Module + L2 => 30.9 mm² (32nm)
Bobcat Dual Core + L2 => ~16 mm² (40nm)

It's kind of obvious that Bobcat will go the route of Merom. History repeats.

---
E-350 vs FX-8150 clock to clock/core to core Passmark via algorithms
CPU Mark (Composite Average)
E-350: ~800
FX-8150: ~900
Integer Math (Millions of Operations per second)
E-350: ~150
FX-8150: ~170
Floating Point Math (Millions of Operations per second)
E-350: ~640
FX-8150: ~600
Find Prime Numbers (Thousands of Primes per second)
E-350: ~150
FX-8150: ~180
SSE / Multimedia (Million Matrices per Second)
E-350: ~3.5
FX-8150: ~3.4
Compression (kBytes processed per second)
E-350: ~930
FX-8150: ~1190
Encryption (MBytes transferred per second)
E-350: ~3.1
FX-8150: ~3.2
Physics (Frames per Second)
E-350: ~50
FX-8150: ~60
String Sorting (Thousands of String per second)
E-350: ~600
FX-8150: ~800

E-350 2c 1.6 GHz/DDR3 1066 (7-7-7-20)? MHz/Labtop - ~9Ws 2c CPU 40-nm
FX-8150 8c 1.6 GHz/DDR3 1333 (7-7-7-21)? MHz/Desktop - 13-18W? 2c CPU 32-nm
---

Only a few people know the changes of Jaguar... http://www.hotchips.org/ but it will make an appearance at HotChips

D: I think that says more about Bulldozer than it does Bobcat...


I still think it makes better sense for AMD to fix up Bulldozer (Piledriver is a good first step, we'll see what Piledriver+ and Steamroller look like), just because Bobcat (unlike Core) isn't actually faster per clock than AMD's current CPU lineup. They would have to make up a lot more ground than Intel did with Core.

 

[ninaholic37]

Sadly, Bulldozer will meets its demise after Steamroller. Jaguar-derivates will replace Bulldozer.

AMD FX(2011) -> 8 cores
AMD A(2012) -> 4 cores
AMD FX(2012) -> 8 cores
AMD A(2013) -> 4 cores
AMD FX(2013) -> 10 cores
AMD A(2014) -> 8 cores(8 Jaguar+ cores)
AMD FX(2014) -> 20 cores(20 Jaguar+ cores)
AMD A(2015) -> 12 cores(12 Jaguar-derived cores)
AMD FX(2015) -> 32 cores(32 Jaguar-derived cores)

---
Don't ask I'm NostaSeronx.

Only a few people know the changes of Jaguar... http://www.hotchips.org/ but it will make an appearance at HotChips

I'm interested to see how this goes... will be checking back on the hotchips site August 27-29th as well.