Llano vs Sandy Bridge Compute Capacity

[Aerwidh]

@JPQY: nobody is denying that SB has a higher IPC than Llano (= K10.5+) and in this test the SB CPU is even running at much higher clk speeds than the Llano APU. so yes, the SB sys has alot more CPU power than the Llano sys shown here.

but here´s what AMD tries to proof: for most mainstream users, and by that i mean surfing the web, watching videos, listen to music, work on a few pics and do office, the Llano notebook APU does a better job than a SB i7-2600 desktop CPU+GPU combo. just because Llano has enough CPU power for most tasks, while the SB lacks enough GPU power.

and im quiet shure that a quad-core K10.5+ Llano can handel your chess game too.

Well put

[Motiv]

how many OpenCL apps do you use on a regular basis? I know that I use NONE. in fact I have NEVER used an OpenCL accelerated app for... anything really. And I'm a power user... The chances of an average home user that would be using the type of system demoed, using a single GPU accelerated app (let alone 3) is extremely small. The only apps that I have seen that a home user would use that support GPU based acceleration are based on CUDA.

I suspect if you're into video editing, then OpenCL is a big thing. With Apple especially pushing it, I suspect art packages will succumb to openCL too (Adobe no doubt will be, as can be seen with Flash now happily GPU integrated).

It's not what currently uses it, it's what will start to use it.

[Nintendork]

video editing springs to mind.

The more apps that could use OpenCL successfully, the better it is for AMD. If you only need to run flash, win7 and the web, then anything better than an Atom will do. however, people do more than that, even if once in a blue moon.

For example Sony Vegas, now supports OpenCL (firegl, radeon, fusion).

[desnudopenguino]

This is starting to remind me of the old x64 and multi-core arguments... "there's nothing that utilizes it, so why are they doing it?" arguments were all over the place, but now who runs a single-core computer? who runs 32 bit OS's (other than benching and people with old computers)? Software developers develop for the hardware (generally speaking)... It probably won't be a big difference to start off, but look at how AMD pioneered 64 bit and multi-core back in the day.

[Chumbucket843]

Lol, how much computing power does chess need exactly? and would there be any tangible benefit to upping compute power...

*from the perspective of a chess nerd*

if you want to get technical chess needs a lot of compute power. there are 10^120 possible games after 80 moves. checkers has been solved by computers but chess will likely never be solved and as of now has only been solved for 6 pieces on the board.

modern chess engines calculate around 15-20 moves ahead for most positions in less than 30 seconds and see almost every tactic which has really changed the game for humans. it's quite complicated to explain why but you can take my word for it. it has revolutionized the chess world in the last 10 years.

[Manicdan]

^i suck at chess on the PC, rock at it in person.
its touch to read the mind of a machine running at a few jiggahz

[Manicdan]

^i suck at chess on the PC, rock at it in person.
its touch to read the mind of a machine running at a few jiggahz

[M.Beier]

The majority of desktops sold don't have graphic cards in them, so I'll disagree with this statement.

But isnt the Zacata plenty for these users anyway?

[[XC] hipno650]

How do you tell a difference in web browsing from AMD to Intel? Have you eliminated any other factors in the 2 systems?

I build allot of home systems for customers. So think of it this way. Two systems. They have the same hard drive, the same RAM, one is using Intel CPU, chipset and GPU the other using AMD CPU chipset and GPU, both using windows 7 FRESHLY installed with the same updates and other programs installed. It has been my experience that in these situations the Intel system responds faster and smoother in general in the OS and in say IE or Chrome responsiveness. (of course the speed at which the page is loaded is almost 100% dependent on the ISP coverage)

Well youre right that theres not much in opencl/gpgpu atm.
However i think thats outstanding what amd does.Because when theres starts to be large enough user base.It will start being implemented.
Something in the line, "build it ,and they will come" philosophy.And we all are gonna cash in on that, desktop cards are going to get much more use also.
Plus much more demanding UI`s for next gen operating systems etc etc.
And i think photoshop/ movie maker types of apps are going o be one of the first ones to utilize this.

you know I do have to hand it to AMD for really pushing it but I think it will be at-least one more generation before it is commonly adopted. but i guess someone needs to push or it will never happen... I just think they focused far to much on it compared to trying to close the CPU gap. As for UI or video acceleration. the new Intel integrated GPU's is more than enough...

I do a fair amount of video editing using lots of Adobe based product and a few others and to be honest they have been saying they will make use of GPGPU for years now and even now it is nothing to write home about. and the big thing is 90% of what I have seen is currently or is being transfered over to CUDA. some of these programs do not recommend using AMD cards because of the heavy CUDA and Nvidia influences that they are starting to incorporate.

good on AMD for pushing GPGPU out but I don't expect to see any reason to switch to them this generation as most apps won't be optimized for it yet. I hope though I'm wrong.

[MrMojoZ]

But isnt the Zacata plenty for these users anyway?

Zacate? No, the overall performance deficit would be noticeable between Llano and Zacate in everyday usage scenarios. Zacate may be enough, but it isn't plenty enough for mid-range on up.

[Sherman Tank]

I build allot of home systems for customers. So think of it this way. Two systems. They have the same hard drive, the same RAM, one is using Intel CPU, chipset and GPU the other using AMD CPU chipset and GPU, both using windows 7 FRESHLY installed with the same updates and other programs installed. It has been my experience that in these situations the Intel system responds faster and smoother in general in the OS and in say IE or Chrome responsiveness. (of course the speed at which the page is loaded is almost 100% dependent on the ISP coverage)

Placebo...

[MrMojoZ]

Placebo...

Heh, "Hey guys this P4/K10 loses in benchmarks but it sure is smooth!!!!!".

[Caparroz]

I build allot of home systems for customers. So think of it this way. Two systems. They have the same hard drive, the same RAM, one is using Intel CPU, chipset and GPU the other using AMD CPU chipset and GPU, both using windows 7 FRESHLY installed with the same updates and other programs installed. It has been my experience that in these situations the Intel system responds faster and smoother in general in the OS and in say IE or Chrome responsiveness. (of course the speed at which the page is loaded is almost 100% dependent on the ISP coverage)

Wasn't AMD the "king of smootheness"? Now Intel is faster and smoother? AMD is doomed...

Anyway, numbers/benches or it didn't happen...

[ajaidev]

Latest video on youtube shows Llano dual core @ 0:48 mark vs Zacate E350.
Mobile Llano has model number A4 3300M and IGP has model number 6480. There is roughly 7 watts of difference under full load so my guess is that Llano is 25W mobile part.
Llano has 50% more GPU compute performance than Zacate,meaning it has roughly 80x1.5=120 SPs. CPU cores are also notably faster,even if they run at 1.6Ghz too.Not a bad tradeoff ,7W more power for 50% better GPU and definitely better pure CPU performance.

Yep very interesting the difference is very obvious am quite happy that AMD did that much with the 7W TDP. Intel also did a lot with the new 2100 series SNB's same is the case here more for less

[BeepBeep2]

You also bring up a good point. I can tell the difference in day to day tasks (yes I can tell the diff in CPU speed from an AMD system to say an i7 in web browsing) but most people will not... if they fire up Photoshop Elements or Movie maker they sure will though...

So you are saying you can tell that your browser loads pages faster on your 2600k at 4.5 than an X6 1100T at 4.4?

Browser responsiveness is dependent on HDD speed, latency and throughput. Maybe you are seeing the difference between two different HDD's or ICH10 vs SB850, in which we all already know ICH10 beats the out of.

The majority of people I know are still on P4 and A64...some have Core 2 Duo and only a few people I know have bought PC's in the last year have a stock 920 or similar.

Anyone looked at that video closely in 720p?
SPi finishes Loop 1 on Llano at 2m 00.493s @ 1.8-2.5 Ghz
SPi finishes Loop 1 on Sandy Bridge at 1m 26.287s @ 3.4-3.8 Ghz

...thats slower, but the Llano chip has a 45w TDP and only 4MB L2

Either one of two things are going on -
1. AMD has MUCH faster cache at 32M and or K10.5 cores respond much better without being starved of cache (1MB per core vs 512KB for Deneb/Thuban)
2. i7 is crippled by having 8 HyperPi vs 4 running at once, despite it bouncing off of 100% load every few seconds.

[generics_user]

I build allot of home systems for customers. So think of it this way. Two systems. They have the same hard drive, the same RAM, one is using Intel CPU, chipset and GPU the other using AMD CPU chipset and GPU, both using windows 7 FRESHLY installed with the same updates and other programs installed. It has been my experience that in these situations the Intel system responds faster and smoother in general in the OS and in say IE or Chrome responsiveness. (of course the speed at which the page is loaded is almost 100% dependent on the ISP coverage)

placebo, when intel went away from their old FSB there was no difference at all between amd and intel in terms of resposiveness

in everyday usage i noticed no difference at all between my old Xeon W3520 @ 4ghz; the 1055t @ 4ghz and my current stop-gap i7 860 @ 2.8ghz
everything with 2 128gb raid 0 Ultradrive GX SSDs

(i got the 1055T for rendering so it was an upgrade from the W3520 )


i'm sure that my new i7 2600 is going to feel just the same in non-work related usage

[XRL8]

you know I do have to hand it to AMD for really pushing it but I think it will be at-least one more generation before it is commonly adopted. but i guess someone needs to push or it will never happen... I just think they focused far to much on it compared to trying to close the CPU gap. As for UI or video acceleration. the new Intel integrated GPU's is more than enough...

I do a fair amount of video editing using lots of Adobe based product and a few others and to be honest they have been saying they will make use of GPGPU for years now and even now it is nothing to write home about. and the big thing is 90% of what I have seen is currently or is being transfered over to CUDA. some of these programs do not recommend using AMD cards because of the heavy CUDA and Nvidia influences that they are starting to incorporate.

good on AMD for pushing GPGPU out but I don't expect to see any reason to switch to them this generation as most apps won't be optimized for it yet. I hope though I'm wrong.

Youre probably right again ,about this generation.Thing is, pc`s and laptops are being used 3-5 years on average, so if its going to get pushed 2-3 years down the line, hardware must be here NOW.And i have to tell, that really most people doesnt need MUCH cpu power, they need ENOUGH, especially on laptops/netbooks.Same thing goes for GPU, its just that no integrated gfx up until Llano had something thats enough.Everytime i have to explain friends/family/clients WHY their brand new notebook with newest intel stuff "just aint enough" to play a game i get so frustrated over this.And they dont understand why they "have to" have discret gfx...
But it looks this combo is not going to be fast, but enough to do most stuff normal people do.
As for the GPGPU, i believe CUDA will go the way of the glide, theres really no place for it in the desktop segment.Its directcompute and opencl in the future.
As for the UI, i was talking about the future, not vista`s aero.
As for the smoothness, its kinda difficult.Im SURE that XP64 UI is damn much faster than 7`s .Especially when im using NET based software under 7, i literally can see the window be "drawn" .My friends tell me im not well tho :P .
Anyhow, there were tests made, that made clear that in some instances, human eye can "kind of" see up to 500FPS, in short bursts and not whole detail.

[Hornet331]

Either one of two things are going on -
1. AMD has MUCH faster cache at 32M and or K10.5 cores respond much better without being starved of cache (1MB per core vs 512KB for Deneb/Thuban)
2. i7 is crippled by having 8 HyperPi vs 4 running at once, despite it bouncing off of 100% load every few seconds.

You forgt the 3rd option:
i7 has a additional handicap of running also the cfd simulation at the same time. (probably 8 threads).

About the "smoother" BS... honestly thats always so much bs... I have noticed no difference between any of my systems till the A64 days and thats mostly due to the fact that since that point I started using a raptor...

The only time when I'll start to see a more "smoother feeling" again, is when i switch to a SSD.

[BeepBeep2]

You forgt the 3rd option:
i7 has a additional handicap of running also the cfd simulation at the same time. (probably 8 threads).

About the "smoother" BS... honestly thats always so much bs... I have noticed no difference between any of my systems till the A64 days and thats mostly due to the fact that since that point I started using a raptor...

The only time when I'll start to see a more "smoother feeling" again, is when i switch to a SSD.

Looks to me the cfd simulation is a GPU only load

[Hornet331]

Yes for amd, the intel HD3000 doesn't support opencl right now, so it runs on pure cpu mode.

[Serra]

*from the perspective of a chess nerd*

if you want to get technical chess needs a lot of compute power. there are 10^120 possible games after 80 moves. checkers has been solved by computers but chess will likely never be solved and as of now has only been solved for 6 pieces on the board.

I don't think anyone is talking about endgaming Chess on Llano. Unless your last name is Karpov though I think even the slowest Llano will be more than a match for 99% of people. I hold the #13 position on Facebook chess worldwide [or I did.. I've been gone so long I'm not sure I show up globally anymore] and can attest that my old Core 2 Duo in my laptop is far better than I am (of course, the FB chess ranking system is very broken... I'm really not that good, just good enough to know how much better than me the computer is).

[[XC] hipno650]

Wasn't AMD the "king of smootheness"? Now Intel is faster and smoother? AMD is doomed...

Anyway, numbers/benches or it didn't happen...

I was never one to believe the "AMD is smoother stuff" ever since C2D I have always felt that Intel Systems have "seemed" faster. also I'm sure we both know that it's impossible to benchmark "smoothness"

So you are saying you can tell that your browser loads pages faster on your 2600k at 4.5 than an X6 1100T at 4.4?

Browser responsiveness is dependent on HDD speed, latency and throughput. Maybe you are seeing the difference between two different HDD's or ICH10 vs SB850, in which we all already know ICH10 beats the out of.

The majority of people I know are still on P4 and A64...some have Core 2 Duo and only a few people I know have bought PC's in the last year have a stock 920 or similar.

Anyone looked at that video closely in 720p?
SPi finishes Loop 1 on Llano at 2m 00.493s @ 1.8-2.5 Ghz
SPi finishes Loop 1 on Sandy Bridge at 1m 26.287s @ 3.4-3.8 Ghz

...thats slower, but the Llano chip has a 45w TDP and only 4MB L2

Either one of two things are going on -
1. AMD has MUCH faster cache at 32M and or K10.5 cores respond much better without being starved of cache (1MB per core vs 512KB for Deneb/Thuban)
2. i7 is crippled by having 8 HyperPi vs 4 running at once, despite it bouncing off of 100% load every few seconds.

thats an extreme comparison. of course a 4.4ghz 1100T system is going to seem as fast as anything because it is has a massive single threaded boost over stock. I was referring to consumer grade stuff as in like core i3/i5 stuff and lower end Phenom and Athlon quads... i'm not saying the difference is massive or anything and as many people have pointed out it might be placebo.... totally possible but I also notice a small diff between my 2600k at stock and OCed to 5GHZ... the differene however is MASSIVE on say something like a netbook compared to say even a normal laptop (CPU usage over 70% opening IE or anything) you might be right about the chipset difference as well though. put it's part of the whole AMD vs Intel package...

placebo, when intel went away from their old FSB there was no difference at all between amd and intel in terms of resposiveness

in everyday usage i noticed no difference at all between my old Xeon W3520 @ 4ghz; the 1055t @ 4ghz and my current stop-gap i7 860 @ 2.8ghz
everything with 2 128gb raid 0 Ultradrive GX SSDs

(i got the 1055T for rendering so it was an upgrade from the W3520 )


i'm sure that my new i7 2600 is going to feel just the same in non-work related usage

you might be right but I'm talking mostly on lower end and lower clock speed CPU's

Youre probably right again ,about this generation.Thing is, pc`s and laptops are being used 3-5 years on average, so if its going to get pushed 2-3 years down the line, hardware must be here NOW.And i have to tell, that really most people doesnt need MUCH cpu power, they need ENOUGH, especially on laptops/netbooks.Same thing goes for GPU, its just that no integrated gfx up until Llano had something thats enough.Everytime i have to explain friends/family/clients WHY their brand new notebook with newest intel stuff "just aint enough" to play a game i get so frustrated over this.And they dont understand why they "have to" have discret gfx...
But it looks this combo is not going to be fast, but enough to do most stuff normal people do.
As for the GPGPU, i believe CUDA will go the way of the glide, theres really no place for it in the desktop segment.Its directcompute and opencl in the future.
As for the UI, i was talking about the future, not vista`s aero.
As for the smoothness, its kinda difficult.Im SURE that XP64 UI is damn much faster than 7`s .Especially when im using NET based software under 7, i literally can see the window be "drawn" .My friends tell me im not well tho :P .
Anyhow, there were tests made, that made clear that in some instances, human eye can "kind of" see up to 500FPS, in short bursts and not whole detail.

for most of the customers I deal with I will always recommend a faster CPU over a faster GPU every day of the week as they are far more likely to boot up movie maker or photoshop then WoW or any other game for that matter. for any one with ANY inkling of gaming usage I ALWAYS recommend a dedicated GPU. this may change with Lano but the trouble is that most more casual games that people may play on a laptop are going to be more CPU bound as far a simulation time is concerned. For example, two of the most popular games for PC right now, WoW and Starcraft 2 are both fairly CPU bound. in which case the Lano GPU might cut it for say meduim/high settings in StarCraft 2 but once a ton of units get going the CPU will die. In which case they are better off spending the extra $100 to get an i7 lappy with a dedicated GPU. I will agree however in saying that for gaming all of the Intel IGP's are total trash...

I do 100% agree that opencl is the future but I see that future as being much farther off then you may think. while it is a "better" solution than CUDA it is newer to market and has a much smaller current dev base than CUDA. also I have heard it is harder to create an OpenCL app compared to a CUDA one. I think more and more people will develop for CUDA and in a few years decide to switch to OpenCL. Software moves pretty slow. Look how long 64bit OSes and 64bit CPU's have been around for, yet still 90% of apps don't natively support 64bit operations...

[mattkosem]

Maybe you are seeing the difference between two different HDD's or ICH10 vs SB850, in which we all already know ICH10 beats the out of.

Really? I routinely saw >1GB/sec seq reads on SB850 when I was using it, and have seen mention of a 600MB/sec limit on ICH10R plenty of places. I thought they were pretty even overall...no?

--Matt

[XRL8]

for most of the customers I deal with I will always recommend a faster CPU over a faster GPU every day of the week as they are far more likely to boot up movie maker or photoshop then WoW or any other game for that matter. for any one with ANY inkling of gaming usage I ALWAYS recommend a dedicated GPU. this may change with Lano but the trouble is that most more casual games that people may play on a laptop are going to be more CPU bound as far a simulation time is concerned. For example, two of the most popular games for PC right now, WoW and Starcraft 2 are both fairly CPU bound. in which case the Lano GPU might cut it for say meduim/high settings in StarCraft 2 but once a ton of units get going the CPU will die. In which case they are better off spending the extra $100 to get an i7 lappy with a dedicated GPU. I will agree however in saying that for gaming all of the Intel IGP's are total trash...

Well, there are two things you should consider.
Unfortunately, many if not most people, go to somekind of market and get a laptop by themselves or by the recommendation of the clerk (which does not care...) and they end up most often than not with a crap one,IGP one.
And second ,laptops with decent specs plus a intel cpu and a discreet gpu cost sizably more.I know that for USA and western europe thats not a BIG problem.However for the rest of the world, its more often than not hard to justify added cost.
Frankly, i dont get why intel lagged in gfx department so long, they have so much resources to throw at this obvious flaw, sandybridge is a big step up, however its still not enough.
As for the SCII and WOW, well i just dont agree that 4 core improved athlon X4 is too slow to be enough for these games, its not like were talking max details and constant 60fps here.But enough ;-) .

[Nikolasz]

wasnt intel all proud and bragging about their cache structure and tweaking?
amd seems to handle their cache and bandwidth a lot better, i give them that...

No one here seems not like to see intel GOING DOWN in this tests