Report: AMD Developing New Kabini and Piledriver Chips

Toggle sidebar Toggle sidebar
Status
Not open for further replies.
28 nm at last !!
Now keep the IGPU kicking and beat Intel processors.

Also, with PS4 and Xbox720 coming soon with 8-core AMD, I suppose multicore support will be more common and AMD will benefit from that.
 
The only 2 things wrong with the FX chips are IPC latency, Windows Scheduler and wattage. However, the latter doesn't seem to be a huge issue unless your going for overclocking. Turning off Core Parking seems to alleviate some scheduling issues and allow the cores to remain active seeing that windows does not know the different between Hyperthreading and modulated cores. (please correct me if I'm wrong)

Personally I'm waiting for steamroller seeing that its the last reported CPU from AMD to use AM3+.
 
[citation][nom]tomc100[/nom]Please God, let AMD make a profit.[/citation]
I think you mean "competitive product." Profits typically follow suit.
 
Dude, what's with all the downrating going on. It just some people cheering up for a neat development.
Fucking Tom's filled with fanboys
 
[citation][nom]mousseng[/nom]I think you mean "competitive product." Profits typically follow suit.[/citation]

They do have some competitive products. The fx-6300 for instance, is amazing at its price point. Beats intel (in the same price range) in any task that is threaded well. it goes for between 120-140$. Not the best for gamers, but for many other tasks it cant be beat for under 150$
 
Well damn, I just bought a FX 6300 :-\ same thing happened to me when I bought my Galaxy Note 2 and TWO days later they announce the Galaxy S4!*sigh* screw keeping up with the " Jones 's", it's waay too costly! 😛
 
[citation][nom]azraa[/nom]Dude, what's with all the downrating going on. It just some people cheering up for a neat development.Fucking Tom's filled with fanboys[/citation]
It's the down voting trolls. They seem to be unusually prevalent lately, and that's saying a lot. The voting system is rarely used for its intended purpose, but despite how often it's misused Tom's still continues to openly support it. I really think commenting worked better back when the voting system was broken.
 
[citation][nom]thecynicalmonk[/nom]Well damn, I just bought a FX 6300 :-\ same thing happened to me when I bought my Galaxy Note 2 and TWO days later they announce the Galaxy S4!*sigh* screw keeping up with the " Jones 's", it's waay too costly! 😛[/citation]

The FX-6350 is a 125W TDP and more expensive, and they all likely OC to the same point, so I'd say little reason to go 6350 anyway.

 
[citation][nom]dragonsqrrl[/nom]It's the down voting trolls. They seem to be unusually prevalent lately, and that's saying a lot. The voting system is rarely used for its intended purpose, but despite how often it's misused Tom's still continues to openly support it. I really think commenting worked better back when the voting system was broken.[/citation]

Downrating makes someone's post look bad, even though it has a very good point. Like you said, downrating troll, and the system is very broken.
 
I hope AMD's CPU business hangs in there and competes with Intel at the high end again... an Intel monopoly sound rather expensive for the consumer and will stagnate innovation. I'm not a fanboi, but in a 2 horse race, you have to support the underdog to maintain competition and benefit all.
 
Pertaining to the mobile news only, being 50% better than the last? Im actually not impressed, unless its really on par with a i3.. let alone a i5 mobile chip from last generation. i dont see the reason to buy them other than the fact that they might cost less or that the user is ignorant to performance/games beyond 13xx x 7xx resolutions.

i mean were not talking abot a a-10 here its saying better than brazos. They said thin notebooks and i think "ultrabook" i already see previews of 600$ haswell laptops/ultrabooks.

the i3 is already in alot of tablets. im a little tired of everyone supporting that amd isnt going to compete at all in any bracket of performance except for ultra low end in the mobile market.
 
[citation][nom]childofthekorn[/nom]The only 2 things wrong with the FX chips are IPC latency, Windows Scheduler and wattage. However, the latter doesn't seem to be a huge issue unless your going for overclocking. Turning off Core Parking seems to alleviate some scheduling issues and allow the cores to remain active seeing that windows does not know the different between Hyperthreading and modulated cores. (please correct me if I'm wrong)Personally I'm waiting for steamroller seeing that its the last reported CPU from AMD to use AM3+.[/citation]

IPC latency? Were you drinking when you made that up? It makes no sense.

They basically castrated the integer units for the Bulldozer and Piledriver, and then added another one to make a pair. Consequently, the only way to get good integer throughput is to use two threads per core, or you just have a lousy processor. But, of course, the decoder can't keep up with two threads at the same time, so basically you have this crummy processor that performs worse at everything compared to the Intel processors.

On top of that, AMD still hasn't learned how to make a cache. They have a puny 16K L1, but it's still slow, and makes the processor much more dependent on the L2 cache. The L2 cache is catastrophically slow, and is only exceeded in ineptness by their L3 cache implementation.

In short, AMD's Bulldozer/Piledriver chips suck donkey balls.

So, if you meant their cache latency is killing the processors, and exacerbating already poor IPC caused by castrating the integer unit, I'd agree.

It's a remarkably poor implementation, being quite large, very slow, and consuming massive amounts of power. It took a lot of engineering talent to make a modern processor that bad. Piledriver helped, but the basic design and slow cache limit this lousy processor severely.

Steamroller could fix a lot of sins with a better decoder, restoring the ALU in each integer unit, and making a cache system that works. The latter of which is nearly impossible, since AMD has no idea how to implement an efficient cache, but maybe with Jim Keller there's hope. I'm pretty sure they'll improve the first two, but if the cache continues to blow, it's going to severely limit the performance of the processor. I'd settle for even a moderate improvement, which is possible since the BD/PD has significantly lower performance than the Thuban had. Just get us back to Thuban, and it will help a lot.


 
[citation][nom]ta152h[/nom]IPC latency? Were you drinking when you made that up? It makes no sense.They basically castrated the integer units for the Bulldozer and Piledriver, and then added another one to make a pair. Consequently, the only way to get good integer throughput is to use two threads per core, or you just have a lousy processor. But, of course, the decoder can't keep up with two threads at the same time, so basically you have this crummy processor that performs worse at everything compared to the Intel processors.On top of that, AMD still hasn't learned how to make a cache. They have a puny 16K L1, but it's still slow, and makes the processor much more dependent on the L2 cache. The L2 cache is catastrophically slow, and is only exceeded in ineptness by their L3 cache implementation. In short, AMD's Bulldozer/Piledriver chips suck donkey balls. So, if you meant their cache latency is killing the processors, and exacerbating already poor IPC caused by castrating the integer unit, I'd agree. It's a remarkably poor implementation, being quite large, very slow, and consuming massive amounts of power. It took a lot of engineering talent to make a modern processor that bad. Piledriver helped, but the basic design and slow cache limit this lousy processor severely.Steamroller could fix a lot of sins with a better decoder, restoring the ALU in each integer unit, and making a cache system that works. The latter of which is nearly impossible, since AMD has no idea how to implement an efficient cache, but maybe with Jim Keller there's hope. I'm pretty sure they'll improve the first two, but if the cache continues to blow, it's going to severely limit the performance of the processor. I'd settle for even a moderate improvement, which is possible since the BD/PD has significantly lower performance than the Thuban had. Just get us back to Thuban, and it will help a lot.[/citation]
You forgot that Intel's compiler is extremely biased toward its own chips & many software companies use it. "Not genuine"...
 
Will this quad core sell as cheap as pentium class Ivy bridge?

Because the core 2 quad are still ridiculous expensive. Any new dual core from Intel isnt really an upgrade from core 2 duo 3GHz. Would be nice if they can be bench with core 2 quads. About time we get a quadcore @ intel's dual core pricing.
 
Is there much of an improvement from the fx6100 (I have a 6100 at 4.0ghz) to the 6350 ?
I was just planning on buying the 8350, I have a Asus Sabertooth 2.0 with the 6100 currently.
 
[citation][nom]tomfreak[/nom]Will this quad core sell as cheap as pentium class Ivy bridge? Because the core 2 quad are still ridiculous expensive. Any new dual core from Intel isnt really an upgrade from core 2 duo 3GHz. Would be nice if they can be bench with core 2 quads. About time we get a quadcore @ intel's dual core pricing.[/citation]

Dual core Celeron G1610 is available for about $50 and it's about 50% faster than a Core 2 Duo at ~3GHz. Intel has many options available that kick the crap out of Core 2 Duo and some of them have been around since Sandy Pentiums/Celerons came out.

[citation][nom]Teeroy32[/nom]Would it be worth upgrading from my FX-6100 to a FX-63xx or just save for a FX-83xx?[/citation]

Unless you are doing some highly-threaded work, it'd be better to skip out on the eight-core models for now and wait until the next generation to see what is available. The FX-6350, although better than the FX-6100, isn't so much better that I'd recommend upgrading to it unless you feel that what you currently have is inadequate. Although not the only advantage, the main advantage that the 6350 has over the 6100 seems to be clock frequency- overclocking a 6100 (even on stock voltage) should cover most of the lost ground.

[citation][nom]uglynerdman[/nom]Pertaining to the mobile news only, being 50% better than the last? Im actually not impressed, unless its really on par with a i3.. let alone a i5 mobile chip from last generation. i dont see the reason to buy them other than the fact that they might cost less or that the user is ignorant to performance/games beyond 13xx x 7xx resolutions. i mean were not talking abot a a-10 here its saying better than brazos. They said thin notebooks and i think "ultrabook" i already see previews of 600$ haswell laptops/ultrabooks. the i3 is already in alot of tablets. im a little tired of everyone supporting that amd isnt going to compete at all in any bracket of performance except for ultra low end in the mobile market.[/citation]

Are you seriously demanding that AMD's competitors for Atom get anywhere near Core i3s in performance and saying that if they fail to do so, they are a product with no purpose? WTF are you smoking? Brazos was already better than Intel's competition for it, Atom, and if Kabini is really that much better at 50% all-around better than Brazos while being incredibly more efficient and such, then it could be a product with a wide and lucrative range of markets to work with.
 
[citation][nom]childofthekorn[/nom]The only 2 things wrong with the FX chips are IPC latency, Windows Scheduler and wattage. However, the latter doesn't seem to be a huge issue unless your going for overclocking. Turning off Core Parking seems to alleviate some scheduling issues and allow the cores to remain active seeing that windows does not know the different between Hyperthreading and modulated cores. (please correct me if I'm wrong)Personally I'm waiting for steamroller seeing that its the last reported CPU from AMD to use AM3+.[/citation]You mean memory latency? AMD engineers looked into this, and determined that the percentage of performance gained from "fixing" that "problem" wouldn't yield as much performance benefits as spending resources and transistors on other improvements. They seem to be correct, as Piledriver provided a good boost to performance even though cache latency is unchanged.
 


I don't think he/she meant IPC latency and regardless, cache latency is a huge issue for AMD. It is true that their memory latency issue is only minor, but their cache latency issue is one of their biggest performance problems. That AMD changed some other things and improved performance in no way proves this wrong. That manual overclocking of AMD's cache greatly improves performance in most heavy workloads and that Intel has a huge performance advantage with a huge cache performance advantage support this quite conclusively.

That AMD's L3 cache capacity seems almost irrelevant for most tasks (compare FX-4170, FX-4300, and A10-5800K CPU performance with otherwise same performance systems, the 0MiB L3, 4MiB L3, and 8MiB L3 differences seem irrelevant desptie being the only big performance difference between the three CPUs because they all have nearly identical CPU performance) also supports this quite well in that AMD's L3 is just too slow to make a difference in most situations.
 
[citation][nom]mousseng[/nom]I think you mean "competitive product." Profits typically follow suit.[/citation]
Actually, what he means is, keep AMD profitable, so AMD wll still be standing, and keep producng competetive prduct.
The end justifies the means though.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom