Sign in with
Sign up | Sign in
Your question

Does anyone think that AMD should release a lower power FX CPU?

Last response: in CPUs
Share
a b à CPUs
October 29, 2012 5:23:12 AM

Well, like the title states, I'm asking if anyone here thinks that AMD should offer a lower power FX CPU. With the FX-4300 at 3.8GHz base clock frequency and a 95W TDP, there's no way that I can see where AMD couldn't have an ~3GHz model at 65W that competes with the Pentiums by slaughtering them in threaded performance, even if while being slaughtered in single/dual-threaded performance in turn.





http://techreport.com/review/23750/amd-fx-8350-processo...

Really, it looks like such a CPU which would probably perform around an A8-3850 in CPU performance and it wouldn't consume too much more power than the top Pentiums that it beats even in gaming simply because two cores without HTT really don't cut it in gaming these days. AMD could use such a CPU on the AM3+ socket to give a lower end option that has a good upgrade path because putting such a CPU on the FM2 socket simply isn't going to give as much upgrade room as AM3+.
October 29, 2012 5:57:17 AM

I can see where you are getting at. And its a very valid point actually. Those ppl who already own an AM3+ motherboard with an FX or Athlons could really make use of a low power CPU, but then, since we have downclocked the CPU where its performance is really not more than 100$ worth...... it will eat the 4300 sales.....secondly......such a CPU will send the msg, that the AM3+ platform is only viable upto a 100$ CPU, anything more than that, and you are better off with an Intel CPU. I know, it really is the case now, but the new CPU will make this very apparent to the normal consumer and will highlight this fact. So, its better AMD doesnt do this........ or if they do have a new CPU like this, it better be near the price of the 4300. :) 
a b à CPUs
October 29, 2012 6:12:06 AM

All AMD has to do to get lower power chips is catch up in architecture.

32 nm technology can't compete in power with 22 nm technology.

As long as AMD stays far behind the curve in this, they will stay far behind the curve in power too.
Related resources
a c 186 à CPUs
a b À AMD
October 29, 2012 6:13:43 AM

Probably an athlon........
a b à CPUs
October 29, 2012 6:52:10 AM

wrazor said:
I can see where you are getting at. And its a very valid point actually. Those ppl who already own an AM3+ motherboard with an FX or Athlons could really make use of a low power CPU, but then, since we have downclocked the CPU where its performance is really not more than 100$ worth...... it will eat the 4300 sales.....secondly......such a CPU will send the msg, that the AM3+ platform is only viable upto a 100$ CPU, anything more than that, and you are better off with an Intel CPU. I know, it really is the case now, but the new CPU will make this very apparent to the normal consumer and will highlight this fact. So, its better AMD doesnt do this........ or if they do have a new CPU like this, it better be near the price of the 4300. :) 


How would such a CPU send the message that FX is only good to the weakest CPU of the family? I don't see how that's any different from saying that the Intel line is only good up to the Pentiums.
a b à CPUs
October 29, 2012 6:54:58 AM

Raiddinn said:
All AMD has to do to get lower power chips is catch up in architecture.

32 nm technology can't compete in power with 22 nm technology.

As long as AMD stays far behind the curve in this, they will stay far behind the curve in power too.


All I asked for was a lower end model of the current FX line. That does'nt need an architecture tweak nor a die shrink.

32nm technology most certainly can compete with 22nm technology. The die process is just one factor in this. Proof of that is how even Intel's 45nm Nehalem products can beat some of AMD's Bulldozer 32nm products in both performance and power efficiency.
a b à CPUs
October 29, 2012 1:42:01 PM

Powering 4 cores at 32nm takes more juice than powering 2 cores at 22nm. It would be difficult for AMD to get comparable no matter how you slice it.

Does AMD really have some kind of interest in a processor that is:
Clocked slower than an i3
Less IPC than an i3
Still higher power anyway
Often has useless cores

?

It would be like reinventing the Core 2 Quad from back when dinosaurs roamed the earth.
a b à CPUs
October 29, 2012 2:27:28 PM

Raiddinn said:
Powering 4 cores at 32nm takes more juice than powering 2 cores at 22nm. It would be difficult for AMD to get comparable no matter how you slice it.

Does AMD really have some kind of interest in a processor that is:
Clocked slower than an i3
Less IPC than an i3
Still higher power anyway
Often has useless cores

?

It would be like reinventing the Core 2 Quad from back when dinosaurs roamed the earth.


The FX-4300 at 3.8GHz is a 95W TDP part with realistic power consumption that is only about 15W higher than the i7-3770K with both at stock according to AnandTech.



Dropping the frequency to around 3GHz should be enough to drop voltage enough to get it to a 65W TDP even if it has slightly inferior binning to the FX-4300. It could be priced right under $100 with the top Pentiums which it would beat in gaming performance according to Tech Report's chart.

It wouldn't even need to consume more power than the Ivy i3s at lesser CPU loads such as gaming.
October 29, 2012 2:28:03 PM

OP, have you seen this review from Hexus:
http://hexus.net/tech/reviews/cpu/47257-amd-a10-5700/?p...
http://hexus.net/tech/reviews/cpu/47257-amd-a10-5700/?p...

?

The A10-5700 is not an FX part, but it clocks at 3.4GHz and you can see from the page 6 discrete card power usage that it actually draws less power than an i3 Ivy. At those speeds it loses slightly to the i3 in performance but I think if an FX version for AM3 was released the L3 cache would have to be less power hungry than an igpu.

So, I would have to say that AMD have already proven they can do what you are proposing - they just haven't built one for the AM3 socket for some reason. IMO, probably the 5700s are requiring a fair bit of binning to achieve.
a b à CPUs
October 29, 2012 2:29:41 PM

designasaurus said:
OP, have you seen this review from Hexus:
http://hexus.net/tech/reviews/cpu/47257-amd-a10-5700/?p...
http://hexus.net/tech/reviews/cpu/47257-amd-a10-5700/?p...

?

The A10-5700 is not an FX part, but it clocks at 3.4GHz and you can see from the page 6 discrete card power usage that it actually draws less power than an i3 Ivy. At those speeds it loses slightly to the i3 in performance but I think if an FX version for AM3 was released the L3 cache would have to be less power hungry than an igpu.

So, I would have to say that AMD have already proven they can do what you are proposing - they just haven't built one for the AM3 socket for some reason. IMO, probably the 5700s are requiring a fair bit of binning to achieve.


I didn't see those reviews exactly (thanks for the links), but Trinity is part of what I was thinking about for this. Like you said, AMD has already proven that they can make such a CPU, I was just wondering if anyone else thought that they should.



Pulled from your link, we can see that such a CPU as I suggested could use less power than the i3-3225 for gaming. Anand's link shows us that when you try some work that can load this CPU's cores, it'd consume more power than the i3s, but games that don't load up all cores wouldn't necessarily even make this CPU use more power than i3s. Piledriver really is something, huh? Architecturally, it seems to be able to be more efficient than Ivy for gaming despite a significant process technology disadvantage!
a b à CPUs
October 29, 2012 3:19:15 PM

Let me get this straight...

We are comparing an APU minus the on processor graphics with no L3 cache that is powering a discrete video card against an i3 powering the same discrete video card, is that right?

Said a different way, we are talking about a Piledriver version of Athlon 2 x4?

I just don't see how it will deliver similar performance at similar power as compared to an i3.

Maybe if all 4 cores are being used in cases where HT can't be used, but that is its only chance that I can see.

a b à CPUs
October 29, 2012 3:36:17 PM

Raiddinn said:
Let me get this straight...

We are comparing an APU minus the on processor graphics with no L3 cache that is powering a discrete video card against an i3 powering the same discrete video card, is that right?

Said a different way, we are talking about a Piledriver version of Athlon 2 x4?

I just don't see how it will deliver similar performance at similar power as compared to an i3.

Maybe if all 4 cores are being used in cases where HT can't be used, but that is its only chance that I can see.


Look at the second picture of my first post. This proposed CPU would perform well under the i3s in gaming performance, but it would beat the similarly priced Pentiums in gaming performance quite significantly. For reference, that chart is not in FPS, it's in perceivable FPS that has stutter accounted for, so it's as realistic of a measurement of gaming CPU performance as you can get.

My proposed CPU is basically an FX-4300 with the frequency and voltage dropped to fit in a 65W TDP, so it'd be around 3GHz for a base clock. We can see from the chart that it'd be right between the Pentium G2120 and the i3-3325 in gaming CPU performance because it'd perform about as well as an A8-5600K (my proposed CPU has a lower frequency, but it also has an L3 cache).

According to the hexus charts, it'd consume a little less power than the i3-3225 for most games, but Anand shows us that with well-threaded workloads unlike the lightly threaded gaming workloads, it'd use around as much power as the i5-3570K, probably a little less. We can clearly see from many sites how it'd idle at lower power consumption than the i3-3225.

Basically, this:
Gaming performance: ~halfway between the i3-3225 and the Pentium G2120
Gaming power consumption: ~ Pentium G2120 and i3-3225
Fully threaded power consumption (probably including BF3 MP): ~i5-3570K
Idle power consumption: Lower than Pentium G2120
Price point: Around $85-90

I think that it'd be a great addition to AMD's line as competition for the top Pentiums. Do you disagree?
October 29, 2012 3:47:33 PM

blazorthon said:

*snip*

Pulled from your link, we can see that such a CPU as I suggested could use less power than the i3-3225 for gaming. Anand's link shows us that when you try some work that can load this CPU's cores, it'd consume more power than the i3s, but games that don't load up all cores wouldn't necessarily even make this CPU use more power than i3s. Piledriver really is something, huh? Architecturally, it seems to be able to be more efficient than Ivy for gaming despite a significant process technology disadvantage!


I'd be careful about using the word "efficiency" here. If you check the actual gaming benchmarks (from the same review), you will see that although the A10-5700 can beat the i3 Ivy for power usage, it also chalks up some losses in CPU-bound games. Now, my opinion is that once you're above 60fps a loss or a win isn't worth much, but I know many people feel differently. Therefore, it should be pointed out that, strictly speaking, the i3 provides more performance for the same energy usage, which would make it more efficient.

AMD's Trinity (and therefore Piledriver) has the potential to match Intel's power usage, but at a performance deficit. Personally, I would be fine with that provided the cpu provided a "good enough" performance, which the benchmarks show it does (again, a reminder that my standards are that the cpu must not tank the gpu below 60fps average). AMD would obviously have to sell their chip at an attractively lower price to make up for the relative performance loss, but since it would be good enough, I would be satisfied. It would make a chip like the i3 obsolete, really, because if your argument is that "good enough" performance is not good enough, then an i3 is not a large enough jump. You have to get 4 real cores and run an i5 for the real performance jump.
a b à CPUs
October 29, 2012 3:51:50 PM

designasaurus said:
I'd be careful about using the word "efficiency" here. If you check the actual gaming benchmarks (from the same review), you will see that although the A10-5700 can beat the i3 Ivy for power usage, it also chalks up some losses in CPU-bound games. Now, my opinion is that once you're above 60fps a loss or a win isn't worth much, but I know many people feel differently. Therefore, it should be pointed out that, strictly speaking, the i3 provides more performance for the same energy usage, which would make it more efficient.

AMD's Trinity (and therefore Piledriver) has the potential to match Intel's power usage, but at a performance deficit. Personally, I would be fine with that provided the cpu provided a "good enough" performance, which the benchmarks show it does (again, a reminder that my standards are that the cpu must not tank the gpu below 60fps average). AMD would obviously have to sell their chip at an attractively lower price to make up for the relative performance loss, but since it would be good enough, I would be satisfied. It would make a chip like the i3 obsolete, really, because if your argument is that "good enough" performance is not good enough, then an i3 is not a large enough jump. You have to get 4 real cores and run an i5 for the real performance jump.


Ignoring Hyper-Threading, Ivy really isn't more efficient in this situation. Compare my proposed CPU to say the Pentiums that consume about the same amount of power as the i3s and we can see how the loss of Hyper-Threading cuts realistic gaming performance so much that efficiency takes a nose dive on the Pentiums. Remember, this CPU would compete with the Pentiums, not the i3s, so it doesn't need to best the i3s, it simply needs to best the Pentiums and that it can do.
October 29, 2012 3:56:17 PM

blazorthon said:
Ignoring Hyper-Threading, Ivy really isn't more efficient in this situation. Compare my proposed CPU to say the Pentiums that consume about the same amount of power as the i3s and we can see how the loss of Hyper-Threading cuts realistic gaming performance so much that efficiency takes a nose dive on the Pentiums. Remember, this CPU would compete with the Pentiums, not the i3s, so it doesn't need to best the i3s, it simply needs to best the Pentiums and that it can do.



Regarding the theoretical cpu in your above post (I did not see it until I posted myself), I would say it sounds like a winner for budget builds. Cpuworld has listed specs for some salvaged igpu-less Trinity parts that have 65W ratings, so such a part may come to fruition. Based on the availability of similiar salvaged Llano Athlons, though, I wouldn't expect them to be easy to find. Also they would be lower performing than a purpose-built new model such as you are proposing.
October 29, 2012 3:58:56 PM

::sigh::

I miss AMD... where did you go...? to the Stars apparently... or that you never really left at all...

Dry
a b à CPUs
October 29, 2012 4:01:02 PM

Yeah, I doubt that AMD will really give this market a serious push too :( 

Looking at the numbers, it seems like AMD could really crush Intel at this price point, but like you said, AMD's not interested and the closest that they're likely to get to it are hardly available FM2 Athlons that wouldn't even perform as well.
a b à CPUs
October 29, 2012 4:19:06 PM

I don't think there is a huge market for things that perform worse than an i3.

Even if there was, I don't know that people would really want 3570k level gaming power usage with performance that is worse than an i3. Correct me if you are wrong, but that seems to be suggested by these lines:

Gaming performance: ~halfway between the i3-3225 and the Pentium G2120
Fully threaded power consumption (probably including BF3 MP): ~i5-3570K
a b à CPUs
October 29, 2012 4:25:41 PM

Raiddinn said:
I don't think there is a huge market for things that perform worse than an i3.

Even if there was, I don't know that people would really want 3570k level gaming power usage with performance that is worse than an i3. Correct me if you are wrong, but that seems to be suggested by these lines:

Gaming performance: ~halfway between the i3-3225 and the Pentium G2120
Fully threaded power consumption (probably including BF3 MP): ~i5-3570K


For the majority of games, this CPU would consume about as much power as its Pentium G2120 competition while beating that Pentium in performance. Only in a game such as BF3 MP that can use pretty much as many cores as you can throw at it in a consumer system would this CPU use more power and even then, it's only a little more. The FX would out-perform the Pentium so much in BF3 MP that it'd still have an efficiency win.

Basically, in any task that can use enough threads to get this FX CPU using more power than the Pentiums and i3s, this FX would outperform the Pentium competition so much that it wouldn't really matter because it'd finish any task so much quicker and then go to idle that it'd end up being more power efficient. It'd have comparable power efficiency to the i3s in such well-threaded tasks because of its lower idle power consumption being so much lower that it can make up for the higher load power consumption.
a b à CPUs
October 29, 2012 4:52:05 PM

Maybe they thought they would just be cannibalizing the market for their own 4300s?

Either that or they just didn't think there was a huge market for a toned down 4300?

Maybe they just don't have the money to support developing even more platforms especially considering they are giving up competing with Intel in the desktop market as it is.
a b à CPUs
October 29, 2012 5:05:55 PM

Raiddinn said:
Maybe they thought they would just be cannibalizing the market for their own 4300s?

Either that or they just didn't think there was a huge market for a toned down 4300?

Maybe they just don't have the money to support developing even more platforms especially considering they are giving up competing with Intel in the desktop market as it is.


I'm thinking that the first point is the main thing. The second doesn't make much sense unless they're being stupid given how well Intel's Pentiums sell (although they have been known to make similar mistakes in the past, so I admit that it's entirely possible) and I don't get what you're saying with the third. This'd be on the same platform, AM3+, and it'd just be a lower-clocked version of a CPU that they already have, so I don't see why it'd need much R&D.

I agree that AMD probably won't do this; I'm just thinking that it'd be a good addition to their lineup if they did and wanted to ask if others thought so too.
a c 480 à CPUs
a c 122 À AMD
October 29, 2012 5:18:33 PM

AMD's APUs are designed to be power efficient. Their FX CPUs are designed for performance and while PileDriver uses less power than Bulldozer, it is still a bit high compared to Intel CPUs.

The only way AMD will release CPUs that are slower than the FX-4300 is if they have a lot of "defective" FX CPUs that cannot run at 3.8GHz. These "defective" CPUs can either be simply trashed, or they can "de-tune" them so that they run at a lower clock speed and thus consume less power... and naturally sell them for a lower price. Doing so is better than simply trashing them, but the profit margin on these chips could be very small. It costs the same amount of money to produce a FX-4300 and a de-tuned "FX-3900" that runs at 3.0GHz.

If AMD were to release the "FX-3900", it might cannibalize some sales of the FX-4300 which would reduce AMD's potential net profits. At this moment in time AMD cannot afford to release products the might cannibalize the sale of higher end models which carries a higher profit margin. How much does it actually cost to produce a FX-4300? I don't know.
a b à CPUs
October 29, 2012 5:25:45 PM

jaguarskx said:
AMD's APUs are designed to be power efficient. Their FX CPUs are designed for performance and while PileDriver uses less power than Bulldozer, it is still a bit high compared to Intel CPUs.

The only way AMD will release CPUs that are slower than the FX-4300 is if they have a lot of "defective" FX CPUs that cannot run at 3.8GHz. These "defective" CPUs can either be simply trashed, or they can "de-tune" them so that they run at a lower clock speed and thus consume less power... and naturally sell them for a lower price. Doing so is better than simply trashing them, but the profit margin on these chips could be very small. It costs the same amount of money to produce a FX-4300 and a de-tuned "FX-3900" that runs at 3.0GHz.

If AMD were to release the "FX-3900", it might cannibalize some sales of the FX-4300 which would reduce AMD's potential net profits. At this moment in time AMD cannot afford to release products the might cannibalize the sale of higher end models which carries a higher profit margin. How much does it actually cost to produce a FX-4300? I don't know.


I've already shown how it should consume as much power as I claim, so I won't go into that again.

There's that again, eating into FX-4300 sales while having poor profit margins. Also, IDK how much it cost, but unless they changed from Bulldozer, the FX-4300 uses the same die as the eight and six core models, so it probably already has poor profit margins, but they're undoubtedly, as you said, better than trashing the dies that don't pass as eight or even six core models.
a b à CPUs
October 29, 2012 6:54:39 PM

Maybe platform isn't the right word to use in this context, but what I mean to say is that downclocking a 4300 to 3.0 or whatever and then selling it isn't an essentially free activity.

It has to be marketable to distributors, essentially AMD would have to convince stores to take it. That means AMD at a minimum has to prove that it is worth shelf space. Then there is the problem of having to market it at the corporate level. AMD isn't like Apple in that people will buy something purely because it exists.

Not only that, but AMD has XXX amount of production capacity for these chips and they have historically had some catastrophig problems ramping up production in the not too recent past. It isn't at all clear that large quantities of these things can just be willed into existence.

Making these would likely mean that certain parts of the manufacturing process have to be redesigned and it could very easily eat into the number of 4300 chips they are capable of producing. Essentially, they could be looking at not selling a higher priced thing in order to instead sell a lower priced one.

That could work, but I would think they would only want to do it with defective chips and afaik most chip defects can't be fixed by just scaling back in clock rate unlike disabling a flawed module or two.

Additionally, AMD as a company doesn't need to encourage people to use low end processors. It benefits them to push as many cores as possible onto people which in turn gets people demanding support for additional cores. Since AMD is in the business of providing the most cores per dollar, they have a strong incentive to sell 8000s whenever possible. Trying to market yet more quads doesn't align with any long term objective that will help the company itself to do well in general.

Lastly, AMD already has plenty of chips in the market that are below the power level of the 4300. There are still plenty of Phenoms, Athlons, etc that can be sold at costs under what the 4300 sits at and still provide a reasonable performance for the money spent. Mostly these are just prior made and marketed chips that sellers still haven't been able to work through. The sellers want to be rid of these chips and marketing a product at the same power level will just make it harder for them to do that. So there is a potential rejection by distributers and retailers in that vein as well.
!