[AT] AMD Kaveri APU Launch Details: Desktop, January 14th

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

SiliconWars

Platinum Member
Dec 29, 2012
2,346
0
0
Except that the DDR3 7750 has DDR3-1600 memory- Kaveri should use at least 2133MHz, hopefully 2400MHz. That will help alleviate the bandwidth bottleneck a little.

That's actually a lot, going from 25.6 GB/s to 38.4 GB/s. As bandwidth is the restriction here, I would expect to see this easily beat the 7750 with DDR3 1600 MHz. With DDR3 2400 MHz this will be the most balanced APU, graphically, that AMD has released by far. I'm expecting good things from power consumption.
 

Lonyo

Lifer
Aug 10, 2002
21,939
6
81
Also depends on what else those tests were run at. 1920x1080 isn't very helpful. Depends also on various quality settings. Some of those settings don't really make sense, like running AA on an IGP.

And obviously without actual FPS it's irrelevant. If it goes from playable to unplayable, that's a problem. If it goes from unplayable to even more unplayable, who cares?
This is one instance where the Hocp testing method makes sense. What's the highest settings it can get playable framerates at.

Also it might be more CPU limited with an AMD processor vs Intel, so the performance hit would be lower.
 

blckgrffn

Diamond Member
May 1, 2003
9,198
3,185
136
www.teamjuchems.com
Well, I am pleasantly surprised at the 3.7Ghz clock speed.

The power usage - especially at idle and mid loads - is really going to have an impact on the perception of this chip for ne. Given what we had heard about the process characteristics, it seems like this could be good at idle but ugly under load. This peak load number seems to get a lot of attention.

Will this chip get roasted in these forums for its load power usage?

I just ordered a set of this memory (DDR3 2133):

http://www.newegg.com/Special/ShellS...3-_-11122013_3

To go with a Kaveri APU (given that we don't get any nasty surprises in the meantime). I am feeling the itch to build a mitx HTPC with one of these chips... Normally I have just been buying whatever is cheapest and that has been 1333 or 1600.
 

Enigmoid

Platinum Member
Sep 27, 2012
2,907
31
91
That's actually a lot, going from 25.6 GB/s to 38.4 GB/s. As bandwidth is the restriction here, I would expect to see this easily beat the 7750 with DDR3 1600 MHz. With DDR3 2400 MHz this will be the most balanced APU, graphically, that AMD has released by far. I'm expecting good things from power consumption.

Assuming AMD's memory controller has improved.



Not seeing the scaling. At most in toms review is around 22% in games.
 

Ventanni

Golden Member
Jul 25, 2011
1,432
142
106
I was gonna say, isn't that an issue with AMD's current memory controller? While DDR3-2133 may provide a significant amount of bandwidth over DDR3-1600, AMD's memory controller isn't efficient enough to effectively use all of it; something Intel doesn't have as much of a problem with.

Isn't it scheduled for a redesign come Excavator?
 

monstercameron

Diamond Member
Feb 12, 2013
3,818
1
0
I was gonna say, isn't that an issue with AMD's current memory controller? While DDR3-2133 may provide a significant amount of bandwidth over DDR3-1600, AMD's memory controller isn't efficient enough to effectively use all of it; something Intel doesn't have as much of a problem with.

Isn't it scheduled for a redesign come Excavator?

the cache setup seems different, atleast L2, maybe this will help...also what if they took the memory controller from the gpu side rather than the cpu side??
 

Blue_Max

Diamond Member
Jul 7, 2011
4,227
153
106
Here's a neat idea... AMD should come up with a PCIe16 "daughter card" option for its APU series. Instead of sharing DDR3 system RAM, one can purchase a daughter-card to plug into the PCIe16 slot. It doesn't have a separate GPU on it, only extensions for the APU to improve it, and/or 1-2GB of dedicated GDDR5 memory.

...perhaps not the most cost-effective... but a fun thought, at least.
 
Last edited:

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
Here's a neat idea... AMD should come up with a PCIe16 "daughter card" option for its APU series. Instead of sharing DDR3 system RAM, one can purchase a daughter-card to plug into the PCIe16 slot. It doesn't have a separate GPU on it, only extensions for the APU to improve it, and/or 1-2GB of dedicated GDDR5 memory.

...perhaps not the most cost-effective... but a fun thought, at least.

With HSA wonder if it would be possible for hybrid crossfire where the iGP shares the GDDR5 memory on the discrete GPU...
 

Erenhardt

Diamond Member
Dec 1, 2012
3,251
105
101
Here's a neat idea... AMD should come up with a PCIe16 "daughter card" option for its APU series. Instead of sharing DDR3 system RAM, one can purchase a daughter-card to plug into the PCIe16 slot. It doesn't have a separate GPU on it, only extensions for the APU to improve it, and/or 1-2GB of dedicated GDDR5 memory.

...perhaps not the most cost-effective... but a fun thought, at least.

funny, because it is exact opposite of what they want to do (Unified memory)
 

monstercameron

Diamond Member
Feb 12, 2013
3,818
1
0
funny, because it is exact opposite of what they want to do (Unified memory)

the memory would be as latent for the gpu as would the cpu if they would have to access it via the pcie bus, so that could hurt latency sensitive workloads but it seems workloads that desire bandwidth could see benefits...
 

Ventanni

Golden Member
Jul 25, 2011
1,432
142
106
Here's a neat idea... AMD should come up with a PCIe16 "daughter card" option for its APU series. Instead of sharing DDR3 system RAM, one can purchase a daughter-card to plug into the PCIe16 slot. It doesn't have a separate GPU on it, only extensions for the APU to improve it, and/or 1-2GB of dedicated GDDR5 memory.

...perhaps not the most cost-effective... but a fun thought, at least.

Neat idea, but wouldn't that far exceed the bandwidth that a PCI-E 16x slot would allow?
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,752
14,783
136
<edited after explanation, just trying to keep this thread pure>
 
Last edited:

Ventanni

Golden Member
Jul 25, 2011
1,432
142
106
We were talking about the bandwidth issue the 512-core GPU portion of the Kaveri chip would face, and a few member's suggestions of how AMD may be able to get around it (hence the PCI-E daughter board suggestion). We also looked at the performance difference between it and a DDR5 version of the same board. There's also speculation that maybe AMD also redesigned the memory controller, but that likely isn't going to come until Excavator. Of course with any new chip and its new process, there's also question of things like power consumption and any leaked performance numbers.

In an attempt to make sure we're staying on topic, I looked for anything related to video design in this thread. I don't see anything where anyone mentioned anything about video design.
 

unon

Junior Member
Jan 2, 2013
21
0
61
Are you sure AMD was to improve the memory controller in xv? Because some of the planned 2014 xv apu features made it into this kaveri.. GCN 2.0 and context switching wasn't there in original kaveri.
 

rtsurfer

Senior member
Oct 14, 2013
733
15
76


Correct me if I am wrong, but I think this slide hints that AMD is no longer interested in making more traditional "Big Cores".

I think FX-8350 will be the last AMD processor with the "Big cores" :'''(, from now on it will be HSA.

More GPU cores and less CPU cores.....??

Edit:
Source
 
Last edited:

monstercameron

Diamond Member
Feb 12, 2013
3,818
1
0
hmm, what about mixed cores a la big.LITTLE?

2 SR big cores [1 module]
4 bobcat cores
512 gpu style cores
+
opencl and boom!

single threaded perf and multithreaded perf in one package the size of richland
 

mikk

Diamond Member
May 15, 2012
4,173
2,211
136
Except that the DDR3 7750 has DDR3-1600 memory- Kaveri should use at least 2133MHz, hopefully 2400MHz. That will help alleviate the bandwidth bottleneck a little.


Not really because the bandwidth is shared with the CPU in a game.
 

inf64

Diamond Member
Mar 11, 2011
3,764
4,223
136
Correct me if I am wrong, but I think this slide hints that AMD is no longer interested in making more traditional "Big Cores".

I think FX-8350 will be the last AMD processor with the "Big cores" :'''(, from now on it will be HSA.

More GPU cores and less CPU cores.....??

Edit:
Source

That article has one problem
Both SR and Excavator ARE big cores. They are not "small" core family. So, no , AMD is still doing the big cores. They are not doing the more cores approach on desktop though.

They opted to boost the IPC and eliminate bottlenecks. Who cares if Excavator will be 2M/4?TD) in 2015 IF it would to perform like 4M/8T 8350+? I wouldn't certainly care, it would be even better marketing-wise . Excavator core seems to be going after just that, building up the individual core strength in order to roughly match up with Haswell/Broadwell generation. I say roughly since it's really hard to do it via only IPC, AMD will opt for also high clock design and some HSA features. But if Ex. core comes at ~90% of Haswell's IPC , or roughly at IB+ territory, it won't mater if Skylake will be +10 or so % more than Haswell, or if it will support AVX3 (I doubt even AVX2 will be supported and give benefit in mainstream software by that time). It won't matter since both will be so fast nobody would care about 10 or 20% difference. They would care about feature set, price and iGPU performance(what the chip can do in their workloads).
 
Last edited:

rtsurfer

Senior member
Oct 14, 2013
733
15
76
That article has one problem
Both SR and Excavator ARE big cores. They are not "small" core family. So, no , AMD is still doing the big cores. They are not doing the more cores approach on desktop though.

They opted to boost the IPC and eliminate bottlenecks. Who cares if Excavator will be 2M/4?TD) in 2015 IF it would to perform like 4M/8T 8350+? I wouldn't certainly care, it would be even better marketing-wise . Excavator core seems to be going after just that, building up the individual core strength in order to roughly match up with Haswell/Broadwell generation. I say roughly since it's really hard to do it via only IPC, AMD will opt for also high clock design and some HSA features. But if Ex. core comes at ~90% of Haswell's IPC , or roughly at IB+ territory, it won't mater if Skylake will be +10 or so % more than Haswell, or if it will support AVX3 (I doubt even AVX2 will be supported and give benefit in mainstream software by that time). It won't matter since both will be so fast nobody would care about 10 or 20% difference. They would care about feature set, price and iGPU performance(what the chip can do in their workloads).

Thats what I was thinking, even though my first post might lack some clarity.

I think 2M/4C is gonna be the max number of Big cores we are going to see, rest is going to be GPU.

Oh and just for clarity, that isn't an 'article' by some tech site, it is an official slide from AMD 's conference that is going on right now.


Also, makes me wonder, what are they going to do with the server market..??

I think 'Warsaw' should carry the server market through 2014 and by 2015 the HSA accelerated applications might perform well enough that we wouldn't need a large amount of cores..?

Which is a bit too optimistic. I think they are going to leave the super computing performance market for Intel. Until, GPU accelerated applications are fast enough to outperform traditional CPU applications..
 
Last edited:
Status
Not open for further replies.
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |