AMD CPU speculation... and expert conjecture

[toshevopc]

https://www.change.org/petitions/advanced-micro-devices-amd-amd-release-high-end-fx-processor-or-apu-with-steamroller-microarchitecture high end steamroller petition :O

 

[juanrga]

And more benchmarks that agree with early predictions

http://www.cpubenchmark.net/cpu.php?cpu=AMD+A10-7850K+APU+with+Radeon+R7+Graphics&id=2133

Early prediction:

I predict that the CPU of the top Kaveri APU will be about 26% faster than top Trinity APU and about 17% faster than top Richland APU. [...] I estimate that the Kaveri quad core APU will have a PassMark CPU score of about 6000 points.

http://www.brightsideofnews.com/news/2013/11/4/what-to-expect-from-kaveri-a-detailed-predictive-analysis.aspx

Kaveri average Passmark score of 5835 is the 97% of the value predicted of 6000 points. I failed by about 3%. I hate to be always wrong.

Richland scores 5009. Therefore Kaveri is a 16% better. I predicted 17%. I hate to be always wrong.

 

[blackkstar]

Juan, you are realizing you can just keep cherry picking benchmarks until you prove you are right, right?

A claim like "x is y% faster than z" is nearly impossible to prove, because you'd have to prove it in every single instance. Meaning if you really wanted to prove Kaveri was about 17% faster than top Richland, you'd need to run every piece of software.

If being that right is so important to you, just write your own review and cherry pick the benchmarks so they show a 17% increase over top Richland part.

 

[griptwister]

Hmmm... What's this?

http://hothardware.com/News/Kaveri-Software-Documentation-Shows-AMD-Considered-GDDR5-Might-Release-EightCore-Variant/

https://www.change.org/petitions/advanced-micro-devices-amd-amd-release-high-end-fx-processor-or-apu-with-steamroller-microarchitecture high end steamroller petition :O

Did everyone like... miss my post? I think we've got a 8 core steamroller chip coming of FM2+. And again, you know what, that would make me right because I said that they're canceling the FX name and possibly AM3+, Not the performance chips...

 

[juanrga]

I predicted 6000 points for Passmark and Kaveri hits 97% of that value. The error between prediction and measurement is of 3% for this benchmark.

I am comparing my predictions for each benchmark to values measured for the same benchmarks.

I predicted the Passmark score from taking the Richland Passmark score as base and multiplying it by the "y% faster" that I predicted. It is unsurprising that I am off reality by a ridiculous 1%.

I know that some people here is unhappy to the idea that my predictions were so accurate.

 

[jdwii]

Oh please he does not give up, Who cares what synthetic benchmarks say. All what matters is real programs and they show a very low improvement on the CPU side of things and smaller then expected improvement in the IGPU. Why can't people take facts for what they are?
Cherry picking benchmarks? The processor is easily 25% slower in real world applications compared to the I5 2500K and even worse at times when compared to a 8350fx.

Lets not also forget the times its actually slower then a A10 6800K.

 

[jdwii]

People still comparing PHII to BD? Guess they'll never learn.

Anyhow, a DT SR chip would be pretty decent as long as they can produce it on something that clocks high. BD's uarch absolutely needs high clock speeds, those 2 ALU's can only do so much when compared to design's that have 3~4 ALU's.

This is mostly AMD trying to tackle the enterprise sector which favors high core count chips, they won't ever be able to competitively compete here. Low power markets is where most of the chedder is and where they should focus their efforts. They just need to add a third ALU to each "core", would cost relatively small amounts of die space and provide the extra integer performance that everyone is complaining about.

Compared to BD phenom or the stars architecture is better even in today's benchmarks the majority of the time its safe to say Amd know's BD was a failure and since their focusing on 4 cores the module design is a failure and yes they know this now i hope. Amd will barley beat the star cores with the excavator design with PD Amd needs a 15% advantage in clock speed to beat a x6 phenom and now its 10% with steamroller and with excavator they will probably be around 5-10% faster in performance per clock compared to the stars design on average besides programs using the new instruction sets this is not in opinion this is a fact, again for the 100th time Amd would have been better with X86 performance if they sticked with the star's design and improved on it.

It's your opinion being regurgitated as fact.

BD uArch was designed to clock high, the K10 uArch can't achieve high clock rates due to design limitations. This is why the "clock for clock" or "at the same clock" argument fails. The last K10 uArch was Llano not Phenom II, so if you want to see a K10 APU then compare against that. It was a four core 32nm implementation of the stars uArch and had the same limitations. Cost, not clock speed, is what differentiates products from each other. AMD isn't "focusing on four core", four core is the max they can fit inside an APU and still have it at a reasonable cost and TDP for consumers. There is no admission to some imagined past grievances as you have stated. The original BD CPU wasn't much of an upgrade for the x4 or x6 crowd, but the 8350 was most definitely an upgrade. I happen to have a 980BE working in my backup system and it's no where near as good as my FX8350.

What you, and many others not clued in to how uArch's are made, don't understand is that the fictional imaginary stars design you have invented could never exist. There was no future 32/28nm "Phenom III" that was a scaled down design, it wouldn't of worked without a major redesign. As you have all just witnessed going to smaller nodes presents very real problems to design, chiefly thermal density and current leakage. Another thing your not realizing is that the BD uArch is far more efficient in utilizing processor resources then Phenom II is and while I can understand why they wanted to go with an 8x2 ALU design I still believe a 6x3 could of been better.

The only real weakness in the BD uArch that isn't present in the K10 is the way they implemented the SIMD FPU. When used in normal workloads it's fine but when synthetic benchmarks are ran on it it tends to perform poorly due to how it's a units are linked. I believe AMD went with this design because they eventually wanted to move most of the SIMD functions to the GPU component and the four 256-bit units present in the BD are sufficient for anything a consumer would throw at it (anything requiring more then four FPU's would be better one on a GPU anyway).

Anyhow I just got done reviewing historical information and 8350 crush's both the 980 and the 1100 overall, it beats it in single threaded and absolutely crush's them in multithreaded. This is primary due to the 980 and 1100 having their clocks being TDP limited, physics is a hard and cruel master. I don't mind people speculating and theory-crafting but the incessant "it's worse then phenom II" is horribly incorrect.

fx8350 vs 1100T
http://www.anandtech.com/bench/product/203?vs=697

fx8350 vs 980
http://www.anandtech.com/bench/product/362?vs=697

First do we want a benchmarking war? between a 8150fx and 1100T i will do so as well as a power consumption war(since the 8150fx used 20%+ more power on load)
Second you state that the stars architecture was limited becease of its high temps if we look back in time that was Global foundries fault since everything that came out of their factory at the time was crap since they just made their 32nm design.
You also state that they could not improve on the stars architecture since it would require a lot of work??? I never knew making a whole new design was easier?
I'll admit that a 8350fx when clocked 18% higher then a 1100t and while having 8 cores vs 6 cores is around 20% faster and yes indeed a better processor. But it took them 2 years added 58% more transistors and still had a 22% higher die size compared to the 1100t built on the 45nm processes.

So yes we have to wonder why.

 

[Cazalan]

Did everyone like... miss my post? I think we've got a 8 core steamroller chip coming of FM2+. And again, you know what, that would make me right because I said that they're canceling the FX name and possibly AM3+, Not the performance chips...

Didn't miss it, just don't think they're going to FAB it for various reasons (cost/demand/process) already mentioned in this thread. The architecture of course supports more cores.

 

[juanrga]

Hmmm... What's this?

http://hothardware.com/News/Kaveri-Software-Documentation-Shows-AMD-Considered-GDDR5-Might-Release-EightCore-Variant/

https://www.change.org/petitions/advanced-micro-devices-amd-amd-release-high-end-fx-processor-or-apu-with-steamroller-microarchitecture high end steamroller petition :O

Did everyone like... miss my post? I think we've got a 8 core steamroller chip coming of FM2+. And again, you know what, that would make me right because I said that they're canceling the FX name and possibly AM3+, Not the performance chips...

There is not 8 core steamroller chip coming to FM2+ and... your post was answered here

http://www.tomshardware.co.uk/forum/352312-28-steamroller-speculation-expert-conjecture/page-223#12487101

 

[palladin9479]

The only real weakness in the BD uArch that isn't present in the K10 is the way they implemented the SIMD FPU. When used in normal workloads it's fine but when synthetic benchmarks are ran on it it tends to perform poorly due to how it's a units are linked. I believe AMD went with this design because they eventually wanted to move most of the SIMD functions to the GPU component and the four 256-bit units present in the BD are sufficient for anything a consumer would throw at it (anything requiring more then four FPU's would be better one on a GPU anyway).

Here's the issue for that: That setup only work on two conditions:

1: No dGPU is present in the system
OR
2: An AMD dGPU is present in the system

Thanks to WDDM, the OS can only handle one graphical driver at a time. And using the iGPU is currently handled GPU driver side. HSA, as far as SOFTWARE IMPLEMENTATION goes, is no different then using OpenCL with a dGPU, with the iGPU as the target instead.

So why should I bother with HSA, which only works on a single product line [APUs], when I can use OpenCL and get basically the same thing working across every GPU in existence?

If I were AMD, I would be pushing both MSFT and the Linux kernel developers to allow multiple graphic driver installations for this very reason.

Never said it was a particularly good idea, just I see where they were going. Ultimately 99% of what consumers do isn't going to need more then four 256-bit SIMD FPU's, you and I both know that. Any anything that would need would be better served on a dGPU, and I'm not talking HSA. CUDA, OpenCL, and GPGPU don't require an iGPU and all are much better at doing SIMD work then any conceivable CPU implementation. Two of those standards are even vender agnostic and used heavily in rendering programs. I'm far more concerned about integer performance as that's what consumer software use's the most. AMD's done spectacular work with just 2 ALU's, gotta hand it to them for squeezing that much work out of them, but they need to give up the ghost and add a 3rd ALU, possibly convert one of the two AGU's into an ALU since it's very rare you'll use more then 1 AGU at any time. That would easily add another 10~15% single thread performance, provided the scheduler doesn't jam up, which it shouldn't with the SR implementation.

 

[juanrga]

Oh please he does not give up, Who cares what synthetic benchmarks say. All what matters is real programs and they show a very low improvement on the CPU side of things and smaller then expected improvement in the IGPU. Why can't people take facts for what they are?
Cherry picking benchmarks? The processor is easily 25% slower in real world applications compared to the I5 2500K and even worse at times when compared to a 8350fx.

Lets not also forget the times its actually slower then a A10 6800K.

First, the excuse was that the predicted numbers were fake.

When the numbers finally agreed with measurements, then the excuse was changed to "cherry picking".

When it is emphasized the agreement of predictions for benchmark "A" with measurements of "A", "B" with "B"... now the excuse is being changed to who cares what synthetic benchmarks say.

The facts that some are real-world benchmarks or that synthetic and application benchmark don't disagree seems irrelevant.

Which will be the next excuse?

 

[palladin9479]

Spoiler

People still comparing PHII to BD? Guess they'll never learn.

Anyhow, a DT SR chip would be pretty decent as long as they can produce it on something that clocks high. BD's uarch absolutely needs high clock speeds, those 2 ALU's can only do so much when compared to design's that have 3~4 ALU's.

This is mostly AMD trying to tackle the enterprise sector which favors high core count chips, they won't ever be able to competitively compete here. Low power markets is where most of the chedder is and where they should focus their efforts. They just need to add a third ALU to each "core", would cost relatively small amounts of die space and provide the extra integer performance that everyone is complaining about.

Compared to BD phenom or the stars architecture is better even in today's benchmarks the majority of the time its safe to say Amd know's BD was a failure and since their focusing on 4 cores the module design is a failure and yes they know this now i hope. Amd will barley beat the star cores with the excavator design with PD Amd needs a 15% advantage in clock speed to beat a x6 phenom and now its 10% with steamroller and with excavator they will probably be around 5-10% faster in performance per clock compared to the stars design on average besides programs using the new instruction sets this is not in opinion this is a fact, again for the 100th time Amd would have been better with X86 performance if they sticked with the star's design and improved on it.

It's your opinion being regurgitated as fact.

BD uArch was designed to clock high, the K10 uArch can't achieve high clock rates due to design limitations. This is why the "clock for clock" or "at the same clock" argument fails. The last K10 uArch was Llano not Phenom II, so if you want to see a K10 APU then compare against that. It was a four core 32nm implementation of the stars uArch and had the same limitations. Cost, not clock speed, is what differentiates products from each other. AMD isn't "focusing on four core", four core is the max they can fit inside an APU and still have it at a reasonable cost and TDP for consumers. There is no admission to some imagined past grievances as you have stated. The original BD CPU wasn't much of an upgrade for the x4 or x6 crowd, but the 8350 was most definitely an upgrade. I happen to have a 980BE working in my backup system and it's no where near as good as my FX8350.

What you, and many others not clued in to how uArch's are made, don't understand is that the fictional imaginary stars design you have invented could never exist. There was no future 32/28nm "Phenom III" that was a scaled down design, it wouldn't of worked without a major redesign. As you have all just witnessed going to smaller nodes presents very real problems to design, chiefly thermal density and current leakage. Another thing your not realizing is that the BD uArch is far more efficient in utilizing processor resources then Phenom II is and while I can understand why they wanted to go with an 8x2 ALU design I still believe a 6x3 could of been better.

The only real weakness in the BD uArch that isn't present in the K10 is the way they implemented the SIMD FPU. When used in normal workloads it's fine but when synthetic benchmarks are ran on it it tends to perform poorly due to how it's a units are linked. I believe AMD went with this design because they eventually wanted to move most of the SIMD functions to the GPU component and the four 256-bit units present in the BD are sufficient for anything a consumer would throw at it (anything requiring more then four FPU's would be better one on a GPU anyway).

Anyhow I just got done reviewing historical information and 8350 crush's both the 980 and the 1100 overall, it beats it in single threaded and absolutely crush's them in multithreaded. This is primary due to the 980 and 1100 having their clocks being TDP limited, physics is a hard and cruel master. I don't mind people speculating and theory-crafting but the incessant "it's worse then phenom II" is horribly incorrect.

fx8350 vs 1100T
http://www.anandtech.com/bench/product/203?vs=697

fx8350 vs 980
http://www.anandtech.com/bench/product/362?vs=697

First do we want a benchmarking war? between a 8150fx and 1100T i will do so as well as a power consumption war(since the 8150fx used 20%+ more power on load)
Second you state that the stars architecture was limited becease of its high temps if we look back in time that was Global foundries fault since everything that came out of their factory at the time was crap since they just made their 32nm design.
You also state that they could not improve on the stars architecture since it would require a lot of work??? I never knew making a whole new design was easier?
I'll admit that a 8350fx when clocked 18% higher then a 1100t and while having 8 cores vs 6 cores is around 20% faster and yes indeed a better processor. But it took them 2 years added 58% more transistors and still had a 22% higher die size compared to the 1100t built on the 45nm processes.

So yes we have to wonder why.

And you would lose, every time. And while I'm positive you could cherry pick a BM where someone handicapped one CPU over the other, it would be exposed for the dishonestly that it is. You keep trying to fall back on "clock for clock" bullsh!t which won't fly as you'd have to handicap one of the CPU's in the process or do something REALLY dishonest like compare an fx4 to an i7.

BD had better performance the K10, it required higher clocks to do so but amazingly it had those higher clocks designed in. PD was a fairly significantly improvement on BD in the efficiency field, it got higher clocks as a result of less power usage. SR's current implementation has the same CPU performance as PD (fx4) but with lower clocks due to material being used, it'll be interesting to see if they make one without the iGPU so as to free it from the TDP constraints that imposes. There would of never been a 4.5ghz K10, it can't clock that high. You can try it yourself if you want, the A8-3870K was a Phenom II made at 32nm. Get a decent board for it and a dGPU. Turn the iGPU's clock rate to the absolute lowest inside BIOS, you won't be using it anyway and the less power it use's in standby the better. Now raise the multiplier on the 3870K and see how high you can get a 32nm quad core K10 to go, benchmark and compare it against the BD/PD CPU's. And before attempt to complain about the lack of L3, Llano had 1MB of full speed L2 cache per core instead of 6MB reduced speed L3 cache (the L3 cache runs at the same speed as the IMC in AMD CPU's btw). See how far you get with it, that's your "upgraded" stars uArch btw, the one you keep harping about.

Another thing I find funny is your keep jumping between the 980 and the 1100T as though they were the same CPU, they aren't. The 980 had better single thread performance due to higher clocks via more TDP room while the 1100T had better multithread performance from it's additional two cores but lower clocks which resulted in lower single thread performance.

The fx8350 has much better multithreaded performance then the 1100T, and that's before overclocking. It also has better singlethreaded performance then the 980, again before overclocking. The 8150 had better multithreaded performance then the 980 and about the same single threaded (+/- a few percent based on workload), that resulted in about the same multithreaded performance as the 1100T, which is pretty amazing when you realize it's 16 ALU vs 18. Overall the fx8150 was the 980 and 1100T combined which is why many of us who had those CPU's didn't have a reason to upgrade. PD was the upgrade to the fx8150 and thus is better then either the 980 or the 1100T, which is why I upgraded and moved my 980 to the secondary system.

Now for what's really happening. Your suffering from unmet expectations, you set your bar too high and when it wasn't met you felt anger and resentment. Behavioral scientists have found that unmet expectations are perceived as hostile actions, similar to theft btw, by humans and we react accordingly.

 

[esrever]

First do we want a benchmarking war? between a 8150fx and 1100T i will do so as well as a power consumption war(since the 8150fx used 20%+ more power on load)
Second you state that the stars architecture was limited becease of its high temps if we look back in time that was Global foundries fault since everything that came out of their factory at the time was crap since they just made their 32nm design.
You also state that they could not improve on the stars architecture since it would require a lot of work??? I never knew making a whole new design was easier?
I'll admit that a 8350fx when clocked 18% higher then a 1100t and while having 8 cores vs 6 cores is around 20% faster and yes indeed a better processor. But it took them 2 years added 58% more transistors and still had a 22% higher die size compared to the 1100t built on the 45nm processes.

So yes we have to wonder why.

That only shows how shitty the early 32nm process at glofo is. Star cores stopped scailing with frequency at about 4ghz. I tested this with my cpu, 4.2 and 4.0 was pretty much the same performance. I don't think it was throttling either. Phenom II was already a frequency scaled design from phenom I, there really wasn't much headroom there. It took years for them to even get to the 1100T considering. The power consumption is also very high on the 1100T, pretty much on par with an FX8150 if you want the same multithreaded integer performance.

The only reason the FX chips are so large is because of the massive L3, which is probably more useful for server space. Llano cores are larger than half a trinity module.

 

[palladin9479]

Thuban is actually a bigger die then Bulldozer and about the same energy consumption.

https://en.wikipedia.org/wiki/Phenom_II

Die size: 346 mm²
TDP: 125W

https://en.wikipedia.org/wiki/List_of_AMD_FX_microprocessors
FX8150
Die size: 319 mm2
TDP: 125W

Many people confuse it with Deneb (980BE) which was a four core design from the start, which again has the same TDP as BD.

Die size: 258 mm²
TDP: 125W

What makes BD "power hungry" is overclocking it. Where as K10 slams into a wall and simply won't go much higher, BD and PD both can clock significantly higher then their TDP allows for but ends up with massive energy consumption. BD is about 4.4~4.5Ghz and PD is 4.8ghz when it hits that massive curve, K10's wall is 4Ghz. I know these numbers intimately as I built a custom WC kit for my 980 that I used on my fx8350.

 

[noob2222]

Rofl. Juan is still going? Funny how wrong I was to him for saying kaveri = 20% ipc - 11% clock speed for roughly 9% over richland. Still thinks he has something to prove. Lol.

 

[colinp]

Anyone heard of Danth's law?

"If you have to insist you’ve won an internet argument, you’ve probably lost badly"

 

[Reynod]

Spoiler

People still comparing PHII to BD? Guess they'll never learn.

Anyhow, a DT SR chip would be pretty decent as long as they can produce it on something that clocks high. BD's uarch absolutely needs high clock speeds, those 2 ALU's can only do so much when compared to design's that have 3~4 ALU's.

This is mostly AMD trying to tackle the enterprise sector which favors high core count chips, they won't ever be able to competitively compete here. Low power markets is where most of the chedder is and where they should focus their efforts. They just need to add a third ALU to each "core", would cost relatively small amounts of die space and provide the extra integer performance that everyone is complaining about.

Compared to BD phenom or the stars architecture is better even in today's benchmarks the majority of the time its safe to say Amd know's BD was a failure and since their focusing on 4 cores the module design is a failure and yes they know this now i hope. Amd will barley beat the star cores with the excavator design with PD Amd needs a 15% advantage in clock speed to beat a x6 phenom and now its 10% with steamroller and with excavator they will probably be around 5-10% faster in performance per clock compared to the stars design on average besides programs using the new instruction sets this is not in opinion this is a fact, again for the 100th time Amd would have been better with X86 performance if they sticked with the star's design and improved on it.

It's your opinion being regurgitated as fact.

BD uArch was designed to clock high, the K10 uArch can't achieve high clock rates due to design limitations. This is why the "clock for clock" or "at the same clock" argument fails. The last K10 uArch was Llano not Phenom II, so if you want to see a K10 APU then compare against that. It was a four core 32nm implementation of the stars uArch and had the same limitations. Cost, not clock speed, is what differentiates products from each other. AMD isn't "focusing on four core", four core is the max they can fit inside an APU and still have it at a reasonable cost and TDP for consumers. There is no admission to some imagined past grievances as you have stated. The original BD CPU wasn't much of an upgrade for the x4 or x6 crowd, but the 8350 was most definitely an upgrade. I happen to have a 980BE working in my backup system and it's no where near as good as my FX8350.

What you, and many others not clued in to how uArch's are made, don't understand is that the fictional imaginary stars design you have invented could never exist. There was no future 32/28nm "Phenom III" that was a scaled down design, it wouldn't of worked without a major redesign. As you have all just witnessed going to smaller nodes presents very real problems to design, chiefly thermal density and current leakage. Another thing your not realizing is that the BD uArch is far more efficient in utilizing processor resources then Phenom II is and while I can understand why they wanted to go with an 8x2 ALU design I still believe a 6x3 could of been better.

The only real weakness in the BD uArch that isn't present in the K10 is the way they implemented the SIMD FPU. When used in normal workloads it's fine but when synthetic benchmarks are ran on it it tends to perform poorly due to how it's a units are linked. I believe AMD went with this design because they eventually wanted to move most of the SIMD functions to the GPU component and the four 256-bit units present in the BD are sufficient for anything a consumer would throw at it (anything requiring more then four FPU's would be better one on a GPU anyway).

Anyhow I just got done reviewing historical information and 8350 crush's both the 980 and the 1100 overall, it beats it in single threaded and absolutely crush's them in multithreaded. This is primary due to the 980 and 1100 having their clocks being TDP limited, physics is a hard and cruel master. I don't mind people speculating and theory-crafting but the incessant "it's worse then phenom II" is horribly incorrect.

fx8350 vs 1100T
http://www.anandtech.com/bench/product/203?vs=697

fx8350 vs 980
http://www.anandtech.com/bench/product/362?vs=697

First do we want a benchmarking war? between a 8150fx and 1100T i will do so as well as a power consumption war(since the 8150fx used 20%+ more power on load)
Second you state that the stars architecture was limited becease of its high temps if we look back in time that was Global foundries fault since everything that came out of their factory at the time was crap since they just made their 32nm design.
You also state that they could not improve on the stars architecture since it would require a lot of work??? I never knew making a whole new design was easier?
I'll admit that a 8350fx when clocked 18% higher then a 1100t and while having 8 cores vs 6 cores is around 20% faster and yes indeed a better processor. But it took them 2 years added 58% more transistors and still had a 22% higher die size compared to the 1100t built on the 45nm processes.

So yes we have to wonder why.

And you would lose, every time. And while I'm positive you could cherry pick a BM where someone handicapped one CPU over the other, it would be exposed for the dishonestly that it is. You keep trying to fall back on "clock for clock" bullsh!t which won't fly as you'd have to handicap one of the CPU's in the process or do something REALLY dishonest like compare an fx4 to an i7.

BD had better performance the K10, it required higher clocks to do so but amazingly it had those higher clocks designed in. PD was a fairly significantly improvement on BD in the efficiency field, it got higher clocks as a result of less power usage. SR's current implementation has the same CPU performance as PD (fx4) but with lower clocks due to material being used, it'll be interesting to see if they make one without the iGPU so as to free it from the TDP constraints that imposes. There would of never been a 4.5ghz K10, it can't clock that high. You can try it yourself if you want, the A8-3870K was a Phenom II made at 32nm. Get a decent board for it and a dGPU. Turn the iGPU's clock rate to the absolute lowest inside BIOS, you won't be using it anyway and the less power it use's in standby the better. Now raise the multiplier on the 3870K and see how high you can get a 32nm quad core K10 to go, benchmark and compare it against the BD/PD CPU's. And before attempt to complain about the lack of L3, Llano had 1MB of full speed L2 cache per core instead of 6MB reduced speed L3 cache (the L3 cache runs at the same speed as the IMC in AMD CPU's btw). See how far you get with it, that's your "upgraded" stars uArch btw, the one you keep harping about.

Another thing I find funny is your keep jumping between the 980 and the 1100T as though they were the same CPU, they aren't. The 980 had better single thread performance due to higher clocks via more TDP room while the 1100T had better multithread performance from it's additional two cores but lower clocks which resulted in lower single thread performance.

The fx8350 has much better multithreaded performance then the 1100T, and that's before overclocking. It also has better singlethreaded performance then the 980, again before overclocking. The 8150 had better multithreaded performance then the 980 and about the same single threaded (+/- a few percent based on workload), that resulted in about the same multithreaded performance as the 1100T, which is pretty amazing when you realize it's 16 ALU vs 18. Overall the fx8150 was the 980 and 1100T combined which is why many of us who had those CPU's didn't have a reason to upgrade. PD was the upgrade to the fx8150 and thus is better then either the 980 or the 1100T, which is why I upgraded and moved my 980 to the secondary system.

Now for what's really happening. Your suffering from unmet expectations, you set your bar too high and when it wasn't met you felt anger and resentment. Behavioral scientists have found that unmet expectations are perceived as hostile actions, similar to theft btw, by humans and we react accordingly.

As resident forum psychologist, part-time jester and sometime devil's advocate I feel it appropriate to mention the term your obliquely referring to is in fact "Cognitive Dissonance" and you have hit the nail on the head Palladin.


People its just a microprocessor ... nobody need die of shame or frustration ... its a good chip ... its just not earth shattering.


Juanrga can talk it up all he likes but it isn't going into one of my gaming rigs fullstop.

I think its quite good and should I raise enough money when I sell the wife in April I might get one myself ... for a general purpose desktop and light gaming rig ... light gaming rig means you let other people play on it..

 

[de5_Roy]

As resident forum psychologist, part-time jester and sometime devil's advocate I feel it appropriate to mention the term your obliquely referring to is in fact "Cognitive Dissonance" and you have hit the nail on the head Palladin.

i've heard of it. i think one of my cousin's sister's friend's mother-in-law has it.

People its just a microprocessor ... nobody need die of shame or frustration ... its a good chip ... its just not earth shattering.

i agree. i mean, what's the point in putting your emotion on a lifeless piece of silicon(s)? as for 4-6 core phenoms, they've aged very well imo. amd has gone a different way since then.

Juanrga can talk it up all he likes but it isn't going into one of my gaming rigs fullstop.

he isn't. what he's really doing is simply trying to pursuade others to unconditionally believe him every time he posts something/anything, without argument or asking for proof. all of his claims, data, arguments, harrassments, trolling are for that simple reason. there's no brand or hardware or uarch loyalty involved in it.
the problem is that most people ask for proof or information, that's why he's repeatedly trying expecting different(i.e. preferred) outcome.

 

[Reynod]

Is your cousin's sister's friend's mother-in-law making $7000 a month ??

<deadlinky>

 

[juanrga]

Rofl. Juan is still going? Funny how wrong I was to him for saying kaveri = 20% ipc - 11% clock speed for roughly 9% over richland. Still thinks he has something to prove. Lol.

I recall you saying that Kaveri would be 37% slower than Bulldozer and more recently posting about if AMD had finally improved over old (pre-Buldozer) architecture or not.

Juanrga can talk it up all he likes but it isn't going into one of my gaming rigs fullstop.

I think its quite good and should I raise enough money when I sell the wife in April I might get one myself ... for a general purpose desktop and light gaming rig ... light gaming rig means you let other people play on it..

I am discussing benchmarks and how they agree/disagree with predictions. Nobody is saying you what you have to buy. Or do you see some Euro or Dollar symbol?

he isn't. what he's really doing is simply trying to pursuade others to unconditionally believe him every time he posts something/anything, without argument or asking for proof. all of his claims, data, arguments, harrassments, trolling are for that simple reason. there's no brand or hardware or uarch loyalty involved in it.
the problem is that most people ask for proof or information, that's why he's repeatedly trying expecting different(i.e. preferred) outcome.

I am merely mentioning benchmarks of Steamroller/Kaveri when they are appearing online and then comparing them to predictions made in this same thread. This must disagree to people who claimed that the numbers were fake or to people who did failed predictions and is now trying to hide them. It must be also disagree to people like you who either post off-topic stuff or personal attacks as this recent one.

 

[juanrga]

A new score has been submitted to Passmark

http://www.cpubenchmark.net/cpu.php?cpu=AMD+A10-7850K+APU+with+Radeon+R7+Graphics&id=2133

Now the average is 5865. This is now a 98% of the predicted value. The error between prediction and measurement has reduced to 2%. It seems like if someone submitted some bad chip score initially and that all the news submission are increasing the average. I expect the score to continue increasing in next months with more submissions of normal chips.

Although their review of Kaveri was rather bad, I am rather sympathetic to this

http://www.extremetech.com/computing/174980-its-time-for-amd-to-take-a-page-from-intel-and-dump-steamroller

Finally Steamroller is more efficient than Kabini (Jaguar) but I agree that big cores are not the future. In particular the article agrees with my early prediction that near future is towards lower clocked designs. In fact, I expect Excavator to introduce another clock regression, but this is better left for the Excavator thread... haha

 

[de5_Roy]

http://www.techpowerup.com/img/14-01-21/89b.jpg
due to my newbieness, i completely overlooked this little interesting detail before. look kids, no southbridge!! this is the first motherboard where i see no southbridge/pch/fch/chipset. am1 may have the first socketed socs (kabini apus). afaik, all intel's dt solution uses 2 chips and bga.

Is your cousin's sister's friend's mother-in-law making $7000 a month ??

<deadlinky>

in this tough economy, she just stays at home and earns $213fhg987 per hour with her keyboard.

I am merely mentioning benchmarks of Steamroller/Kaveri when they are appearing online and then comparing them to predictions made in this same thread. This must disagree to people who claimed that the numbers were fake or to people who did failed predictions and is now trying to hide them. It must be also disagree to people like you who either post off-topic stuff or personal attacks as this recent one.

aww, do you really think we're blind? even after being told to stop linking to your website, you found(!) a workaround(!!) by linking to the same stuff on bsn website (which links to...wait..for it... here it comes...it's just that obvious... your website). imo, that's called 'gaming the system', not discussion. that's long before we argue the validity and credibility of your uh... whatever you call it.
i can assure you neither this post and the post before where i mentioned you is off-topic. in the previous post, i was stating how you use stamroller rumors (for example, among many others) to incite certain events for personal satisfaction rather than healthy discussion or argument. if you take that as a personal attack, i can't help you there. i have noticed that any time any one (even mods, lol) argues with you, or presents a valid counter-argument, you cry "personal attack". as for off-topic, i've posted since this thread started, and can safely claim that i post on-topic almost all the time. when i don't, i don't deviate too far, even if i am writing against amd's favor/bias.

 

[de5_Roy]

microcenter adds a10 7850k and 7700k
http://www.microcenter.com/product/427565/Kaveri_Black_Edition_4GHz $180
http://www.microcenter.com/product/427566/A10_7700K_38_Ghz_Black_Edition_Boxed_processor $160
not much cheaper than newegg (both have bf4 coupon), but closer to amd's bulk price for 1000pc. assuming the stock cooler costs from $5-10, mc prices seem cheaper. otoh, fx8350 has in-cart price of $180 and fx6300 $110. at newegg, those are $200 and $120 resp. while fx6350 is $140.
kaveri adopters should wait out this period, until price comes down to sane levels for apus (hsa software aren't even widespread). i applaud the early adopters.

 

[-Fran-]

Well, I was looking around and a G3220+R7250 GDDR5 is around the same price, so I'm really wondering now about trades from both approaches.

The MoBo with a decent SPU (sound asdadasd unit) is VERY expensive for socket 1150, but I won't be needing expensive RAM for the Pentium. The exercise is quite cool, so AMD better put pressure to Intel there.

And by the way, how does the G3220 compares to Kaveri? Anyone has tested that?

Cheers!

 

[logainofhades]

Well, I was looking around and a G3220+R7250 GDDR5 is around the same price, so I'm really wondering now about trades from both approaches.

The MoBo with a decent SPU (sound asdadasd unit) is VERY expensive for socket 1150, but I won't be needing expensive RAM for the Pentium. The exercise is quite cool, so AMD better put pressure to Intel there.

And by the way, how does the G3220 compares to Kaveri? Anyone has tested that?

Cheers!

You can get the G3220 with an HD 7790 for less than what a 7850k is going for on newegg right now.

PCPartPicker part list / Price breakdown by merchant / Benchmarks

CPU: Intel Pentium G3220 3.0GHz Dual-Core Processor ($64.97 @ OutletPC)
Video Card: Asus Radeon HD 7790 1GB Video Card ($114.99 @ Newegg)
Total: $179.96
(Prices include shipping, taxes, and discounts when available.)
(Generated by PCPartPicker 2014-01-21 11:54 EST-0500)