What's AMD doing on sep. 12th?

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
Llano is slightly slower than a Phenom II, a tiny bit.
JF-AMD said that the IPC increases compaired to the Phenom II.

The bulldozer mhz pr mhz, will be faster than a equal Phenom II, probably not by much but it wont be slower.

The fact that you can on air probably reach ~5ghz overclocking with these bulldozers is pretty good.

I realise you can do the same with a 2600k, and the 2600k will probably end up slightly faster in 4thread stuff, and the 8150 slightly faster with 8thread stuff.

Unfortunately ~10-20% of any IPC increase is lost from the CMT nature of the BD module. This is according to AMD's own pre-launch briefings of a module~=1.8 theoretical BD mono-cores. Is it reasonable to think they were able to squeeze more than a 10-15% IPC gain when comparing a BD module to 2 Phenom II cores? I can see why for desktop purposes they would delay BD until they had a decent supply of BD chips capable of 4+GHz turbo at target TDP as that would mean benchmark comparisons between the SB 2500K and 2600K would not embarrass AMD.
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
Better not be just re-announcing unlocked Llanos. Fingers crossed it's related to next gen Radeon or BD.

Thomas Siefert said this morning, that they will be demonstrating the next gen 28nm Southern Islands GPU tonight.
 

ransomlist

Member
Sep 12, 2011
46
0
0
Thomas Siefert said this morning, that they will be demonstrating the next gen 28nm Southern Islands GPU tonight.

:wub: sums it up. I'm on a 9600GT right now- being able to change that to a 7870/7950 before the end of November would be stellar. Screw Nvidia :D
 

toyota

Lifer
Apr 15, 2001
12,957
1
0
:wub: sums it up. I'm on a 9600GT right now- being able to change that to a 7870/7950 before the end of November would be stellar. Screw Nvidia :D
I sure hope your cpu is not as low end as the 9600gt or you will be mighty disappointed with a 7870/7950.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
I sure hope your cpu is not as low end as the 9600gt or you will be mighty disappointed with a 7870/7950.

I upgraded from a 5870 to a 6970 and the performance increase was tremendous. I'd say the 7870 would be in the same ballpark as the 6970 performance wise.

This is pretty exiting news, this means that Tahiti XT isn't far off , and AMD has a *huge* lead on nvidia. Now its time for all the haters to post speculative bullcrap about how tahiti performs before its even released, woohoo.
 

toyota

Lifer
Apr 15, 2001
12,957
1
0
I upgraded from a 5870 to a 6970 and the performance increase was tremendous. I'd say the 7870 would be in the same ballpark as the 6970 performance wise.
first of all every review site would disagree as the 6970 is hardly any faster than a 5870 in most cases. but then again you have also said your 2500k had to be oced just for games to play smoothly. next what does that have to do with what i said to him about having a older/slower cpu?
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
first of all every review site would disagree as the 6970 is hardly any faster than a 5870 in most cases. but then again you have also said your 2500k had to be oced just for games to play smoothly. next what does that have to do with what i said to him about having a older/slower cpu?

Hardly any faster, if you say so pal. Aside from any review site that uses the most retarded ccc settings i've ever seen, I do know that many of my games had a 20-30 fps jump from 5870 to 6970, but if you feel like spreading crap, have at it. :cool: And without exception I could run everything with full detail, with far less loss of performance.

Lets keep the bullcrap going, tell us about how 7970 is going to be a disappointment and will barely outperform 5870...i'm expecting it. We have tons of threads with speculative bullcrap about the bulldozer, minuswell not stop there. Keep going. We need more intel fanboys and nvidia fanboys spreading garbage -- if that makes you feel better about your dated nvidia hardware.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,379
126
Hardly any faster, if you say so pal. Aside from any review site that uses the most retarded ccc settings i've ever seen, I do know that many of my games had a 20-30 fps jump from 5870 to 6970, but if you feel like spreading crap, have at it. :cool: And without exception I could run everything with full detail, with far less loss of performance.

Lets keep the bullcrap going, tell us about how 7970 is going to be a disappointment and will barely outperform 5870...i'm expecting it. We have tons of threads with speculative bullcrap about the bulldozer, minuswell not stop there. Keep going. We need more intel fanboys and nvidia fanboys spreading garbage -- if that makes you feel better about your dated nvidia hardware.

You're barking up the wrong tree. Toyota is a very fair minded guy, he recommends and has honest opinions on the AMD GPUs just like the Nvidia ones.

The fact is that the 5850 and 5870 were massively successful, great GPUs, and the 6xxx series didn't offer a compelling reason to upgrade outside of those needing very high amounts of memory. I am a relatively happy owner of an XFX 2GB 6950, but I acknowledge that it's kind of a tiny jump from the 58xx series. Particularly sad was the say 5870 to 6870, where you'd actually go backwards lol.
 

toyota

Lifer
Apr 15, 2001
12,957
1
0
Hardly any faster, if you say so pal. Aside from any review site that uses the most retarded ccc settings i've ever seen, I do know that many of my games had a 20-30 fps jump from 5870 to 6970, but if you feel like spreading crap, have at it. :cool: And without exception I could run everything with full detail, with far less loss of performance.

Lets keep the bullcrap going, tell us about how 7970 is going to be a disappointment and will barely outperform 5870...i'm expecting it. We have tons of threads with speculative bullcrap about the bulldozer, minuswell not stop there. Keep going. We need more intel fanboys and nvidia fanboys spreading garbage -- if that makes you feel better about your dated nvidia hardware.
what is wrong with you? LOOK at ALL reviews and you will see on average a 6970 is about 15% faster than a 5870. spreading crap? again you are the person that claims your 2500k had to be oced because games stuttered using stock speeds. you are delusional.

and I still cant figure out why you quoted me and told me about your 5870 to 6970 upgrade when it had NOTHING to do with my comment to ransomlist about his cpu.
 
Last edited:

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
You're barking up the wrong tree. Toyota is a very fair minded guy, he recommends and has honest opinions on the AMD GPUs just like the Nvidia ones.

The fact is that the 5850 and 5870 were massively successful, great GPUs, and the 6xxx series didn't offer a compelling reason to upgrade outside of those needing very high amounts of memory. I am a relatively happy owner of an XFX 2GB 6950, but I acknowledge that it's kind of a tiny jump from the 58xx series. Particularly sad was the say 5870 to 6870, where you'd actually go backwards lol.

I think what confused people is that AMD changed their model numbers with the 68xx series. 68xx was never intended to be the successor, 69xx series was cayman, 68xx wasn't. AMD has a long history of doing stuff that makes absolutely no sense --especially model numbers-- so who knows what the 7xxx series will be like.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,379
126
I think what confused people is that AMD changed their model numbers with the 68xx series. 68xx was never intended to be the successor, 69xx series was cayman, 68xx wasn't. AMD has a long history of doing stuff that makes absolutely no sense --especially model numbers-- so who knows what the 7xxx series will be like.

No disagreement there.

I hope the 7xxx series will be good.
 

Arkadrel

Diamond Member
Oct 19, 2010
3,681
2
0
Unfortunately ~10-20% of any IPC increase is lost from the CMT nature of the BD module. This is according to AMD's own pre-launch briefings of a module~=1.8 theoretical BD mono-cores. Is it reasonable to think they were able to squeeze more than a 10-15% IPC gain when comparing a BD module to 2 Phenom II cores? I can see why for desktop purposes they would delay BD until they had a decent supply of BD chips capable of 4+GHz turbo at target TDP as that would mean benchmark comparisons between the SB 2500K and 2600K would not embarrass AMD.


Yes!

Why cant people understand that?

JF-AMD was asked about IPC, and he said the bulldozer has higher IPC than the Phenom II.

Basically take a 4ghz overclocked Phenom II x4, and compair it to a 4ghz Bulldozer (4 core), and the bulldozer will be faster, because it has higher IPC!

Im beginning to think AMD trying to explain the modual approch has backfired... cuz everyone seems to think it means they ll have lower IPC than a phenom II, when that isnt the case.

People are just stuck with that hypothetical bulldozer that isnt a modual, vs a modual bulldozer and the 85% figour they mentioned. NOT UNDERSTANDING that bulldozer IPC > Phenom II, even with modual approch.

How do we know that? will because JF-AMD answeared it, after getting frustrated with people mis-understanding the modual concept, and saying that it ll have lower IPC than a phenom II.




*IF* you where able to overclock a Phenom II to 5.0ghz, it would still be slower than a Bulldozer @5ghz.
However that isnt even a real situation because Phenom II's cant reach 5ghz overclocks on air.

So bulldozer cpu is gonna out perform the Phenom II's.
How anyone could think otherwise is beyound me, why make a new cpu if its slower than your last one? makes sense? No.
 
Last edited:

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
I believe it's 2-3% higher than Athlon II, but below Phenom II.

Athlon II and Phenom II: K10.5. Llano: revised K10.5. Llano's version of K10.5 has 3-4% higher IPC than the one in the Athlon II and Phenom II. The Phenom II only has higher performance than the Athlon II in gaming and file compression, where the L3 cache comes into play.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
Yes!

Why cant people understand that?

JF-AMD was asked about IPC, and he said the bulldozer has higher IPC than the Phenom II.

Basically take a 4ghz overclocked Phenom II x4, and compair it to a 4ghz Bulldozer (4 core), and the bulldozer will be faster, because it has higher IPC!

Im beginning to think AMD trying to explain the modual approch has backfired... cuz everyone seems to think it means they ll have lower IPC than a phenom II, when that isnt the case.

People are just stuck with that hypothetical bulldozer that isnt a modual, vs a modual bulldozer and the 85% figour they mentioned. NOT UNDERSTANDING that bulldozer IPC > Phenom II, even with modual approch.

How do we know that? will because JF-AMD answeared it, after getting frustrated with people mis-understanding the modual concept, and saying that it ll have lower IPC than a phenom II.




*IF* you where able to overclock a Phenom II to 5.0ghz, it would still be slower than a Bulldozer @5ghz.
However that isnt even a real situation because Phenom II's cant reach 5ghz overclocks on air.

So bulldozer cpu is gonna out perform the Phenom II's.
How anyone could think otherwise is beyound me, why make a new cpu if its slower than your last one? makes sense? No.

It doesn't matter if it has 10% higher IPC, which it won't have. That is lost in multi-threaded applications because of the fact that two modules running equals 180% in comparison to running two normal cores, which then would be 200% performance.
 

Arkadrel

Diamond Member
Oct 19, 2010
3,681
2
0
It doesn't matter if it has 10% higher IPC, which it won't have. That is lost in multi-threaded applications because of the fact that two modules running equals 180% in comparison to running two normal cores, which then would be 200% performance.
Your still not understanding it! /face palm.

180% (modual) bulldozer >>>> 200% (not modual) Phenom II core's.

The Bulldozer even at 180% (20% lost) is MUCH faster than the Phenom II, without any performance lost from being in a modual.

JF-AMD has said that time and time again, and people keep mis-understanding it, mis-quoteing that silly figour they used when they tried to explain the bulldozer modual approuch.

WITH BULLDOZER, IPC increases over exsisting products! like the Phenom II.
Even with the modual thingy, its still faster mhz pr mhz than the phenom II.



It doesn't matter if it has 10% higher IPC, which it won't have.
JF-AMD's word, saying Bulldozer Core has better IPC > you saying it wont.

The guy works for AMD, He must know what hes talking about. And No I dont think hes lieing either, which means your wrong when your saying IPC drops.

Your mis-quoteing, and Mis-understanding a theoretical number used to explain how the modual approch works and why they went for that build. Instead of not shareing resources.
 
Last edited:

Arkaign

Lifer
Oct 27, 2006
20,736
1,379
126
Huh? If what I'm understanding is correct :

A 2-module, 4-'core' BD @ 4ghz would be roughly equal to a 4-core Phenom II, because each module is only approximately 1.8 cores, but we expect a modest IPC gain with BD over PhII? So I hope the IPC gain is more substantial. 8 mediocre cores would be lame compared to 4 kickass cores for my uses.
 

GaiaHunter

Diamond Member
Jul 13, 2008
3,732
432
126
It doesn't matter if it has 10% higher IPC, which it won't have. That is lost in multi-threaded applications because of the fact that two modules running equals 180% in comparison to running two normal cores, which then would be 200% performance.

A BD module has more computational resources than a phenom II dual core, with the exception of the shared resources, which are reworked to handle two cores instead of one.

A phenom II isn't 200%. 2 BD modules with the 2nd integer core shaved off in each module is the 200%. When you shave the 2nd integer core off a BD module what you have is a core with more than twice the resources of a stars core.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
Your still not understanding it! /face palm.

180% (modual) bulldozer >>>> 200% (not modual) Phenom II core's.

The Bulldozer even at 180% (20% lost) is MUCH faster than the Phenom II, without any performance lost from being in a modual.

JF-AMD has said that time and time again, and people keep mis-understanding it, mis-quoteing that silly figour they used when they tried to explain the bulldozer modual approuch.

WITH BULLDOZER, IPC increases over exsisting products! like the Phenom II.
Even with the modual thingy, its still faster mhz pr mhz than the phenom II.



JF-AMD's word, saying Bulldozer Core has better IPC > you saying it wont.

The guy works for AMD, He must know what hes talking about. And No I dont think hes lieing either, which means your wrong when your saying IPC drops.

Your mis-quoteing, and Mis-understanding a theoretical number used to explain how the modual approch works and why they went for that build. Instead of not shareing resources.

LOL. Are you serious?

If it had higher-than-Llano IPC it wouldn't be priced as it is. ALL signs point to IPC being no higher than Llano. All.

Even if they get 20% higher IPC, on multi-threaded applications there's still that 20% deficit, which can only be mitigated with more cores. IPC would still be lower than Sandy Bridge, plus HyperThreading gives Intel an advantage on average of 20% on multi-threaded workloads. Even if it has 20% higher IPC than Phenom II, it wouldn't be faster in multi-threaded by much in comparison to Sandy Bridge because of the modular design.
 
Last edited:

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
A BD module has more computational resources than a phenom II dual core, with the exception of the shared resources, which are reworked to handle two cores instead of one.

A phenom II isn't 200%. 2 BD modules with the 2nd integer core shaved off in each module is the 200%. When you shave the 2nd integer core off a BD module what you have is a core with more than twice the resources of a stars core.

I suggest you go back and read a bit. 180% speedup in multi-threaded workloads in comparison to 200% for two typical cores. Even if it has more resources, there's still that deficit.
 

GaiaHunter

Diamond Member
Jul 13, 2008
3,732
432
126
I suggest you go back and read a bit. 180% speedup in multi-threaded workloads in comparison to 200% for two typical cores. Even if it has more resources, there's still that deficit.

2 typical Bulldozer cores.

200% is 2 Bulldozer modules minus the integer core number 3 and number 4.

2 phenom II cores would be much weaker than 2 typical bulldozer cores (at least comparing the resources each core have).
 
Last edited:

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
2 typical Bulldozer cores.

200% is 2 Bulldozer modules minus the integer core number 3 and number 4.

2 phenom II cores would be much weaker than 2 typical bulldozer cores (at least comparing the resources each core have).


Go back and read again. What you're saying is erroneous, and what I'm mentioning has been known for a LONG time.
 

lau808

Senior member
Jun 25, 2011
217
0
71
ipc is per core obv, so jfamd never said, iirc, that 1 module is greater than 2 phII cores. hopefully not the case but it is very possible that bd has better than phII ipc yet 1 module = 2 phII cores or possibly slower (at same clock)
for ex:
phII ipc = 10
bd (with 10% ipc increase) = 11
2 phII cores = 20 (10+10)
1 bd core = 11
1 bd module = (11+11=22*1.8=19.8)
in this scenario, bd has 10% ipc increase, but is slightly slower than phII (module vs 2 cores) but to help this bd will have higher clocks and more cores

so the hope is that bd has more than 10% ipc increase. also amd went module to add more cores. 80% more performance with 12% more die space. it seems as if x2 users should "upgrade" to fx4, x4 user to fx6, x6 users to fx8.
 

GaiaHunter

Diamond Member
Jul 13, 2008
3,732
432
126
Go back and read again. What you're saying is erroneous, and what I'm mentioning has been known for a LONG time.

bulldozer-concept.jpg

bulldozer-module.jpg


It seems to me you grab "two full BD cores" and share some resources.

Doesn't seem Stars core at all.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
ipc is per core obv, so jfamd never said, iirc, that 1 module is greater than 2 phII cores. hopefully not the case but it is very possible that bd has better than phII ipc yet 1 module = 2 phII cores or possibly slower (at same clock)

Bingo.

1 Module BD with 25% increased IPC over Phenom II @ 80% module penalty vs. 2 core design = 1.25x IPC * 0.8 penalty = 1.0x.

You might bank that 25% IPC when you aren't sharing the module's resources if you run 2 threads on 4 cores, 3 threads on 6, or 4 threads on 8. But we don't have any info on this unfortunately.

If BD has 25% IPC increase over Phenom II in 1-4 threads apps, similar IPC to Phenom in 8 threaded apps + 4.8-5.0ghz overclocking, that would be a very, very fast processor. That would mean a 5.0ghz BD would be = 6.25ghz Phenom II X4 (in 4 threaded apps).
 
Last edited: