AMD CPU speculation... and expert conjecture

[blackkstar]

I really want to know what's happening at GloFo. Some of you are talking of 28nm bulk throughout 2015 which would, IMO, be catastrophic.

I wonder where this leaves SOI. AMD was the primary customer for SOI and it is almost like AMD is abandoning it because GloFo can't deliver.

However it will be interesting to see how SR APU clock on bulk. I have a feeling that we will never see 5ghz+ APUs again if AMD abandons SOI. I have a feeling SR APU on 28nm will have a very hard time breaking 4.5ghz on any sort of cooling without massive volts.

EDIT: however I'm thinking that this might be a step backwards. Even if SR is 20% faster IPC that's the same as a 4.44ghz Piledriver.

 

[de5_Roy]

I highly doubt kaveri will be 40-45% over a 980be in single threaded performance. wishfull thinking though

possible, with a little benchmark engineering from trusty(!) ol' amd marketing dept. (that c.a.l.f. oh-so-blindly trust).
1) run both a10 7850k and 980b.e. at default settings with identical specs. a single threaded benchmark will push the 7850k to use it's maximum single core turbo clockrate and 980 be will fall behind because it has no turbo. it's more likely for 7850k to hit it's max single core turbo in a controlled lab than outside. then nonchalantly report test rig specs.
2) you can sweeten the pot with a bench that runs new instruction sets e.g. avx, avx2, fma3, fma4, xop ( or sse 5!!) and so on. poor phenom ii 980 be will be toast. it'll be even better (worse for ph ii) in multithreaded benches.
3) run optimized bios for the 7850k on a good mobo so that it has no trouble hitting turbo clocks. <borat voice> ph ii 980 b.e. has no turbo, everyone know turbo is for girls.
4) never do a clock-for-clock test for single thread (e.g. 7850k, 980 b.e. both locked at 3.5 or 3.8 ghz with power savings and turbo disabled) because that's the one will reveal real per core improvement. if such a test is done, try to dismiss as academic and worthless.
5) compile the benchmark for each cpu individually so that 7850k's uses new instructions. this one is a bit of a stretch.. but my limited knowledge says it's possible to pull off.
6) use max. validated ram citing default configuration. 7850k in that case, would be running 1.5v ddr3 2133 vs phenom's..uh.. ddr3 1600...? i forgot the specs. then run a ram-intensive compression utility for lulz.
7) report igpu accelerated bench results. hah!

I wonder where this leaves SOI. AMD was the primary customer for SOI and it is almost like AMD is abandoning it because GloFo can't deliver.

it may be the other way around. glofo may be reluctant to serve the sole low-profitting, high margin, high-performance-demanding s.o.i customer instead of the likes of mediatek and rockchip.

 

[etayorius]

According to cpu boss 6800 is only 3% better than a 980 in single core test but my oc 750 = 112, a 20% improvement would put it right around a i3 2120 1.34-1.38.
http://cpuboss.com/cpus/AMD-Phenom-II-X4-Black-980-vs-AMD-A10-6800K

Did you checked on the Benchmarks of AMD A10 6800k vs 980 on the same link you posted? it`s almost on the bottom part of the page:

"Better cinebench r11.5 score 4.37 vs 3.58 More than 20% better cinebench r11.5 score"

"Better PassMark (Single core) score 1,568 vs 1,329 Around 20% better PassMark (Single core) score"

https://www.dropbox.com/s/eewrea2pluzkd07/6800k%20vs%20980.jpg

I think it`s on spot with my own claims about Richland being ahead of 980 by 20%, the same is being reported on my own test with PerformanceTest:

https://www.dropbox.com/s/e7luobxhgidmuos/PTst.jpg


That`s 3 different Benchmarks stating the 6800k is more than 20% faster than the 980BE.


I wish there was a Benchmark with few Games with the 6800K against FX-4100, FX-4350, 980BE and FX-8350, we would actually see the real gaming performance.

 

[etayorius]

First the discussion was single core, second go back and look again, you mixed up the cinebench scores. the 980 beats the 6800

No, look again:

https://www.dropbox.com/s/a2o87dlvcnd8pnk/2.jpg

I don`t know about you... but i clearly see the 6800k having more Single core Performance on both Cinebench and PassMark.

 

[noob2222]

Go back a few pages. I posted some with the a10 5700. Closest I could find. Compared to the 4300 the a10 is up to 40% slower. Certain games need the l3 cache.

http://m.techspot.com/review/670-metro-last-light-performance/page6.html

 

[etayorius]

Go back a few pages. I posted some with the a10 5700. Closest I could find. Compared to the 4300 the a10 is up to 40% slower. Certain games need the l3 cache.

I don`t doubt your claims, but the key word is "Certain Games".

 

[etayorius]

Woah! slow down, where the hell did i mentioned 20% in Multicore? i been talking all along IPC and Single Threaded i been all over that, i don`t want to get into the he said that you said that i said like you all go on with juanrga... but seriously, i did NEVER mentioned Multi Core performance at all, you will not even find Multicore Word in any of my recent replies.

 

[jdwii]

http://www.anandtech.com/bench/product/362?vs=675

Phenom is even clocked 100mhz lower yet its still a good 10% better in performance.

 

[jdwii]

Phenom II 980 vs 8350fx

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

Amd 1100T Vs 8350fx

6-10% faster in single thread and clocked 21% higher.

 

[jdwii]

"i been talking all along IPC and Single Threaded i been all over that" ^

 

[etayorius]

According to cpu boss 6800 is only 3% better than a 980 in single core test but my oc 750 = 112, a 20% improvement would put it right around a i3 2120 1.34-1.38.
http://cpuboss.com/cpus/AMD-Phenom-II-X4-Black-980-vs-AMD-A10-6800K

Did you checked on the Benchmarks of AMD A10 6800k vs 980 on the same link you posted? it`s almost on the bottom part of the page:

"Better cinebench r11.5 score 4.37 vs 3.58 More than 20% better cinebench r11.5 score"

"Better PassMark (Single core) score 1,568 vs 1,329 Around 20% better PassMark (Single core) score"

https://www.dropbox.com/s/eewrea2pluzkd07/6800k%20vs%20980.jpg

I think it`s on spot with my own claims about Richland being ahead of 980 by 20%, the same is being reported on my own test with PerformanceTest:

https://www.dropbox.com/s/e7luobxhgidmuos/PTst.jpg


That`s 3 different Benchmarks stating the 6800k is more than 20% faster than the 980BE.


I wish there was a Benchmark with few Games with the 6800K against FX-4100, FX-4350, 980BE and FX-8350, we would actually see the real gaming performance.

Oh ok, you´re right, i thought that was part of the Cinebench Single Threaded performance, my mistake... indeed the Phenom has stronger Cores compared to Piledriver which uses weaker cores, but your claim about Richland being only 3% faster than 980BE is Wrong too, you missed this:

Better PassMark (Single core) score 1,568 vs 1,329 Around 20% better PassMark (Single core) score

And that´s 20% faster IPC in favor of Richland, we were both wrong in a way.

 

[etayorius]

Phenom II 980 vs 8350fx

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

Amd 1100T Vs 8350fx

6-10% faster in single thread and clocked 21% higher.

I already mentioned this in one of my previous comments, i think Anandtech is not allowing Turbo/Boost in any of their CPUs benchmark, i don´t think we should dismiss Turbo/Boost performance, after all is part of the maximum performance of any given CPU, lets try CPUBoss instead which does uses full Max speed of benchmarked CPUs:

Again the FX-8350 beats the 980 in single core performance and remember Richland is Faster than the FX-8350 Single Thread and IPC

http://cpuboss.com/cpus/AMD-Phenom-II-X4-Black-980-vs-AMD-FX-8350

It seems without the Boost and both CPUs at Stock, the FX-8350 and the 980BE have almost the same overall IPC/Single Threaded performance, but when the Boost is active the FX just pulls ahead by a noticeable margin.

"Better PassMark (Single core) score 1,525 vs 1,329 Around 15% better PassMark (Single core) score" (In favor of the FX-8350)

And here is a Gaming Bench against FX-8350 and PhenomII 980BE, which gives the FX a lead against Phenom:

http://media.bestofmicro.com/Y/Q/357650/original/skyrim-1680.png

 

[etayorius]

Woah! slow down, where the hell did i mentioned 20% in Multicore? i been talking all along IPC and Single Threaded i been all over that, i don`t want to get into the he said that you said that i said like you all go on with juanrga... but seriously, i did NEVER mentioned Multi Core performance at all, you will not even find Multicore Word in any of my recent replies.

I put it in bold just so you would not miss it,.

Yup, you were right all along in that regard, check my previous comment, still i think the faster IPC/Single Threaded Performance in Favor or Piledriver against PhenomII persist.

I have no problem declaring i was wrong, but i was wrong in a good way haha i seriously thought it was part of the Richland Single Core advantage, but turned out to be the Multi Core Bechmark in Favor of PhenomII, which is not unexpected as we all know Phenom uses Beffier True Cores.

In Resume:

The FX-8350 is 15% IPC faster than the 980BE
FX-4350 is about 5% IPC Faster than the FX-8350
6800k is about 7% IPC Faster than the FX-4350.
Kaveri 8750K Should be 20% IPC Faster than the 6800k
Kaveri 8750K Should pull a wholly 40-45% IPC compared to Phenom II 980BE, it does not sounds as crazy now huh?

The Zambezi FX-8150 was a total fail in IPC, but it was indeed faster in it`s Max Boost Speed than the 980BE, turns out Applications were mostly not forcing the Turbo/Boost on Zambezi FX CPUs, and due the higher Frec but lower IPC the 980BE pulled ahead in most Applications that were designed to use Single Threaded Performance.

http://cpuboss.com/cpus/AMD-Phenom-II-X4-Black-980-vs-AMD-FX-8150

As you all can see, the FX-8150 could pull faster Max IPC compared to 980BE, but it was useless because somehow Turbo/Boost was not kicking into Games and Single Threaded Applications, i even tested several PerformanceTest FX-8150 CPUs against my own and others 980BE CPUs and the FX-8150 was always winning by a small margin when the Boost was kicking in, in any case Fail for AMD for not doing supporting the Turbo/Boost in the rightway.

 

[blackkstar]

I wonder where this leaves SOI. AMD was the primary customer for SOI and it is almost like AMD is abandoning it because GloFo can't deliver.

it may be the other way around. glofo may be reluctant to serve the sole low-profitting, high margin, high-performance-demanding s.o.i customer instead of the likes of mediatek and rockchip.

I wonder if there is a big list of customers GloFo serves.

You do make a point, AMD had wafer supply agreements and they had that huge issue with tons of Llano stock not getting sold. That could have been a red flag for GloFo that they won't be making enough money off of AMD going SOI and relying on AMD to be their primary SOI customer.

That's extremely discouraging. I don't have high hopes at all for AMD to release parts that overclock like their parts previously did or to release a large high end chip if they are stuck on bulk.

I'm kind of holding out because I would think AMD might bring those products back if they get to SOI later and they haven't denied that, but if AMD goes bulk forever then I'm not even going to sit here and try and justify AMD continuing with big cores and not letting ARM take over. To me, availability of SOI on smaller nodes is the single largest factor of AMD bringing in FX class chips, not ARM or anything else.

 

[noob2222]

^^. that could also go bad for GF. AMD would have no reason to stay.

 

[gamerk316]

Win32: http://www.headline-benchmark.com/results/41a7eed7-7dbf-45d3-8d03-d0ef407b00f6
Win64: http://www.headline-benchmark.com/results/7a13d4da-64e0-420e-9047-cd8064c8d234

Significant jump in performance there. Passed Yuka in two out of three benchmarks, and you'd expect a 2700k to beat a 2600k head to head 😛

 

[jdwii]

"Richland is Faster than the FX-8350 Single Thread and IPC"
No its not richland has the SAME design as trinity(which is Piledriver) for the CPU it has a clock speed increase,

 

[jdwii]

"The FX-8350 is 15% IPC faster than the 980BE
FX-4350 is about 5% IPC Faster than the FX-8350
6800k is about 7% IPC Faster than the FX-4350.
Kaveri 8750K Should be 20% IPC Faster than the 6800k
Kaveri 8750K Should pull a wholly 40-45% IPC compared to Phenom II 980BE, it does not sounds as crazy now huh?"

Are you getting single core performance confused with performance per clock?

 

[Ags1]

Win32: http://www.headline-benchmark.com/results/41a7eed7-7dbf-45d3-8d03-d0ef407b00f6
Win64: http://www.headline-benchmark.com/results/7a13d4da-64e0-420e-9047-cd8064c8d234

Significant jump in performance there. Passed Yuka in two out of three benchmarks, and you'd expect a 2700k to beat a 2600k head to head 😛

The results are by default compared to my sorry little laptop (circa 2008). You can compare to any other system though... here are the Gamerk2 v Yuka results:

http://www.headline-benchmark.com/results/7a13d4da-64e0-420e-9047-cd8064c8d234/083b462f-9ef2-46f6-b1bb-a3c83c1f9cce

 

[de5_Roy]

I wonder if there is a big list of customers GloFo serves.

glofo's website says 150, worldwide. didn't find any list (didn't check hard enough). i know that arm, mediatek, rockchip are among glofo's clients.
here's a little blast from the past:
http://www.xbitlabs.com/news/other/display/20090719105611_Globalfoundries_to_Reveal_Names_of_Additional_Customers_in_30_Days.html

Spoiler

You do make a point, AMD had wafer supply agreements and they had that huge issue with tons of Llano stock not getting sold. That could have been a red flag for GloFo that they won't be making enough money off of AMD going SOI and relying on AMD to be their primary SOI customer.

That's extremely discouraging. I don't have high hopes at all for AMD to release parts that overclock like their parts previously did or to release a large high end chip if they are stuck on bulk.

I'm kind of holding out because I would think AMD might bring those products back if they get to SOI later and they haven't denied that, but if AMD goes bulk forever then I'm not even going to sit here and try and justify AMD continuing with big cores and not letting ARM take over. To me, availability of SOI on smaller nodes is the single largest factor of AMD bringing in FX class chips, not ARM or anything else.

i dunno much about manufacturing... but isn't f.d.-s.o.i. the preferable process for future nodes? or at least that's what i've been seeing in this thread. i don't know if bulk substrate can undercut f.d.s.o.i.
if sr cores are minor upgrade like pd was over bd, forced by glofo's low yields, amd might decide against making an sr cpu. since we know that carrizo will use excavator core, imo it makes more sense if amd skips SR for cpus and makes cpus with EXC cores instead. a smaller node might even become available by then.
ofc, there's always a chance that some other foundry makes better offers to amd.. or bulk processes become as good as s.o.i. through optimizations and so on.

 

[palladin9479]

http://www.headline-benchmark.com/results/403bbefa-261b-4637-90de-a43fe7c8fc00/403bbefa-261b-4637-90de-a43fe7c8fc00

That's what I got. I want to know what gamer did to his setup to get that seeing as its not even being detected right. The software detected my CPU right, memory wrong though(16GB) and I think it's only using one of my 780 Hydro's. Does having SLI enabled degrade it's performance?

 

[noob2222]

Interesting article posted today on BF4 and Mantle.

http://www.tomshardware.com/reviews/johan-andersson-battlefield-4-interview,3688.html

One thing caught my attention

Q: Do we have a solution for consistent multi-player testing?

Johan: Unfortunately not. We do have some internal tools that sort of records parts of the network stream that we use and then play that back and so you sort of play back multi-player footage, which is quite good. But it's not really packaged together in a sense that we can use it in the retail game unfortunately. It's also not fully true either because it's skipping quite a few things, actually, that you're doing in a full multi-player scenario there. Multi-player, unfortunately, as I say it’s rather chaotic, but it is the truer thing to see the real performance when you especially when it comes to CPU performance. GPU performance you can see pretty well with the single-player. But CPU performance is primarily multi-player. We don’t really have a great solution there unfortunately.

I think one of the best things one can do for benchmarking, but it's very tiresome for you guys, but that is to essentially play multi-player on the same level and play it similar ways.

pretty much sais the same as what others have. Single player testing shows almost nothing of the cpu's potential in BF4.

He had some interesting insight on Mantle as well,

I would like to see that, first of all, that we prove out that Mantle actually works and that we deliver great improvements with it and we have a good trajectory with that.

 

[palladin9479]

And what can I do with integer performance inside my daily life, Juan?

Cheers!

EDIT: This question reads a bit trollish, but trust me it ain't that the case. I really want to know what we use on a daily basis that is biased towards integer performance.

Actually ... everything your PC does is "integer performance". Integer is just another way of indicating that its' the ALU doing the work. The ALU is responsible for regular ADD, SUB, MUL, DIV work along with the logical operations like all the compares, jumps and moves. Though it's the MMU that does the data memory access and date control it's the AGU that generates the physical address from the virtual one your code is working on. The *vast* majority of code that runs on a PC is integer code, there is relatively little SIMD FPU code. The thing is, doing SIMD work with integers is very inefficient and utilizing a vector language instead of scalar can be a few hundred to a few thousand percent increase in performance. The old "games need FPU performance" is something left over from the 90's when 8087 was used for geometry setup and software rendering (think Doom). For awhile it was MMX and then some SSE but by then games had started to move geometry and lighting calculations to the GPU thanks to NVidia's push for it.

Ultimately games are FAR more integer limited then SIMD limited do the raw amount of logic they have to use. Getting programmers to get away from doing all the work in one main thread was hard enough as is.

 

[Ags1]

http://www.headline-benchmark.com/results/403bbefa-261b-4637-90de-a43fe7c8fc00/403bbefa-261b-4637-90de-a43fe7c8fc00

That's what I got. I want to know what gamer did to his setup to get that seeing as its not even being detected right. The software detected my CPU right, memory wrong though(16GB) and I think it's only using one of my 780 Hydro's. Does having SLI enabled degrade it's performance?

Thanks! I also would expect stronger results from the 780 v 770. SLI will not work automatically for a typical Java application. You have to force AFR and in any case the app must use full screen (which I don't). See: http://www.geeks3d.com/20101209/tips-how-to-enable-sli-and-crossfire-support-for-opengl-applications/

I am also interested in why gamerk's DPU is ID'd wrong - it's a blocker for me.

 

[gamerk316]

http://www.headline-benchmark.com/results/403bbefa-261b-4637-90de-a43fe7c8fc00/403bbefa-261b-4637-90de-a43fe7c8fc00

That's what I got. I want to know what gamer did to his setup to get that seeing as its not even being detected right. The software detected my CPU right, memory wrong though(16GB) and I think it's only using one of my 780 Hydro's. Does having SLI enabled degrade it's performance?

My CPU/RAM wasn't detected at all, but it looks like it performed right as far as the benchmarking went. (2600k & 8GB 1333 DDR3) Kinda surprised Yuka's 2700k gets detected and my 2600k didn't though. More confusing, other people's 2600k got detected right... *shrugs* And I know its not OC'd right now (it was at one point, but had a few cooling issues and haven't bothered to re-OC yet), so that probably isn't it either.

I'll toy around tonight and see if I can track down what went on there.