• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

AMD APUs: 5870 speed in 2014?

Obsoleet

Obsoleet

Platinum Member
What are your thoughts on this, do you think it's possible according to what they have for a roadmap? I'm thinking my next CPU upgrade will be to Broadwell, but if AMD can match a 5870 with an APU I'll do that instead. Do you think this is going to happen?
 
Obsoleet said:
What are your thoughts on this, do you think it's possible according to what they have for a roadmap? I'm thinking my next CPU upgrade will be to Broadwell, but if AMD can match a 5870 with an APU I'll do that instead. Do you think this is going to happen?
Click to expand...

Eventually it'll happen, though 2014 looks a bit early for that.
 
Not gonna happen. Processnode shrinks is essentially the only way forward in that matter for AMD. And Llano/Trinity is miles away from 5870. You gonna be extremely lucky if you even get 5750 performance in 2014.

Realisticly I would say HD5670 performance. Or 3x Llano performance.
 
Last edited:
I would say that this is possible but it's hard to say how likely it is. The GPU power of APUs has been going up really fast.
 
I can't imagine it. System memory bandwidth isn't even close to being enough even if by some miracle the integrated GPU had the number crunching power.
 
I would say its possible since we already have about 5 five year old performance (3870) from llano and the 5870 came out almost 3 years ago.
 
I tend to hate these charts. But it might show the issue very well.

http://www.tomshardware.com/charts/2011-entry-level-graphics/Mafia-2-medium,2876.html

Even Llano aint even close to a HD5550. And Trinity would be lucky to get near there.

Now try and look up the chart, and tell me again if you think HD5870 performance is even remotely possible in 2 years.

If AMD could do that, and Intel as well for that matter. Then discrete GFX would be dead in 4-5 years.

Its easy to hype and hope. Hard to deliver. People should have learned by now that it always fail.
 
Last edited:
ShintaiDK said:
I tend to hate these charts. But it might show the issue very well.

http://www.tomshardware.com/charts/2011-entry-level-graphics/Mafia-2-medium,2876.html

Even Llano aint even close to a HD5550. And Trinity would be lucky to get near there.

Now try and look up the chart, and tell me again if you think HD5870 performance is even remotely possible in 2 years.

If AMD could do that, and Intel as well for that matter. Then discrete GFX would be dead in 4-5 years.
Click to expand...
The power would have to triple. Seeing as intel recently doubled their APU performance with Ivy Bridge, I would say that tripling performance within a few years is definitely possible.
 
SickBeast said:
The power would have to triple. Seeing as intel recently doubled their APU performance with Ivy Bridge, I would say that tripling performance within a few years is definitely possible.
Click to expand...

Intel didnt double their "APU" performance. And APU is just a silly name AMDs marketign division invented.

Ivy is 40-60% at best better than Sandy (With a shink dedicated to the IGP.). But we talk a very weak IGP to begin with. Not to mention the bandwidth issue aint as strong there yet.

Trinity cant even beat HD5550. Llano far from it. Yet in 2 generations you hope for HD5870? Not to mention HD5870 had what, 150GB/sec memory bandwidth? Even with DDR4-3200 you gonna peak at around 50GB/sec shared with the CPU.
 
I would say on the Core only side possible. The memory bandwidth side, doubt it. Now, if AMD's cpu's at that time used DDR5 ram (Or provided a discreet fame buffer) then that could make getting to 5870 speeds more likely.
 
SickBeast said:
The power would have to triple. Seeing as intel recently doubled their APU performance with Ivy Bridge, I would say that tripling performance within a few years is definitely possible.
Click to expand...

I don't recall at all that IB doubled the video performance. Especially on average not even close. Anand majorly overblew that IB video trash. Its Intels 4th gen and it still can't even beat AMD's first gen. Llano still has a solid 25-35% lead on Intel. Not to mention basically better everything else.
 
formulav8 said:
I don't recall at all that IB doubled the video performance. Especially on average not even close. Anand majorly overblew that IB video trash. Its Intels 4th gen and it still can't even beat AMD's first gen. Llano still has a solid 25-35% lead on Intel. Not to mention basically better everything else.
Click to expand...
Llano's CPU is weak compared to what Intel has. Their GPUs are close to each other, plus Ivy has way better power consumption. I would take Ivy over Llano any day of the week.
 
ShintaiDK said:
I tend to hate these charts. But it might show the issue very well.

http://www.tomshardware.com/charts/2011-entry-level-graphics/Mafia-2-medium,2876.html

Even Llano aint even close to a HD5550. And Trinity would be lucky to get near there.

Now try and look up the chart, and tell me again if you think HD5870 performance is even remotely possible in 2 years.

If AMD could do that, and Intel as well for that matter. Then discrete GFX would be dead in 4-5 years.

Its easy to hype and hope. Hard to deliver. People should have learned by now that it always fail.
Click to expand...

You hate charts and yet you post one. I can post a couple too:

http://images.anandtech.com/graphs/5570_020810195333/21607.png

http://images.anandtech.com/graphs/graph4448/38843.png

Llano performance is a little slower than the 5570, which is about the same as 3870.
 
I think the better way to phrase this would be to say 5870-level compute performance. I doubt we'll that level of graphics performance in this time frame for memory bandwidth reasons (unless we see large eDRAM/more relevant embedded memory tech). That said, if start having this level of compute power in most every new system being sold, I think we'll finally see that GPGPU revolution we've been hoping for for years. Yes, I realize it isn't a magical cure-all and that we would need a paradigm shift in the algorithms used, but the benefits are just right there screaming "Hey look at me, I'm a massively parallel computing monster that can be very good at the tasks the rest of the system struggles with".
 
Last edited:
How much bandwidth is ddr4 going to bring to the table? If the keep the weaksauce CPU maybe they can get the same compute per.
 
ShintaiDK said:
Be careful not to pick the DDR2 version as alot tends to do with its limited 12.8GB/sec 😉

Its night and day difference with the GDDR3 or GDDR5 versions.
Click to expand...

Those were taken from two anandtech reviews so I wont go there.

How are you sure your results are credible? Were they using the same cpu for all their data points?
 
fishsauce said:
Those were taken from two anandtech reviews so I wont go there.

How are you sure your results are credible? Were they using the same cpu for all their data points?
Click to expand...

You already mix 2 different reviews yourself on different platforms. Not to mention the anandtech chart.
 
ShintaiDK said:
Depends, but by 2014-2015 we should have DDR4-3200. Or about 50GB/sec on a dualchannel setup.
Click to expand...

That sounds pretty decent, but isn't the memory bandwidth of a 5870 ~150GB/sec?
Perhaps they can offer two modes until system memory is up to that level, one where you don't install expensive "GPU" memory and another where you can pop in the side memory to have additional memory for the APU other than system.

1-2GB of 150GB/sec memory that granted 5870 speeds on a Trinity+ APU would be fine for me, and likely cheaper than the $250-500 for a high end GPU. This would be an ideal desktop for me.
 
I think 2015 is more likely.

By 2013, AMD aims at HD7750 level of performance with Kaveri. HD7750 is almost twice as slow as an HD5870.

I think that Kaveri won't be quite as powerful as the stand-alone desktop variant. If AMD doubles that performance by 2014, it still won't quite reach 5870's performance (unless games start to use more and more tessellation where HD5870 performs poorly). By 2015, I think we can say it's almost a done deal though. By then HD5870 will be a 6-year-old GPU. ^_^

At the same time by 2014-2015 we should have next generation games/game engines and GPUs twice as powerful as HD7970/680 today. By that point HD5870 level of performance won't be anything special, especially if tessellation takes off and games start to use > 2GB of VRAM.

Of course compared to Intel, AMD will be miles ahead. Even Trinity is about to mop the floor with Intel's IVB GPU.

1_m.jpg

Source
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top