AMD FX "Bulldozer" Review - (4) !exclusive! Excuse for 1-Threaded Perf.

[boxleitnerb]

6990 is 2 6970's crossfire last I checked

Not to mention if you look further back in thread you can see that I posted results from same game with diff settings and the total system results are improving as I optimize the system AKA i'm cpu bottlenecked

Thanks. However, you can be bottlenecked by a component to a certain degree, it doesn't have to be complete. It could also mean, that you start to be seriously gpu bottlenecked at roughly 130fps and BD only reaches this mark with tweaking. To analyze this, you would have to do one of two things:

a) overclock sandy further. if the fps rise (almost) 1:1 with the clock increase, you are indeed cpu bound. If not, there is your limit, that now both cpus hit
b) overclock your 6990. if the fps rise, you are gpu bound.

Sorry, but so far it is not conclusive. Also, is it DX9 or DX11?

Here some results for comparison:

http://www.forum-3dcenter.org/vbulle...76&postcount=7
970@4000, gtx 570@800mhz sli: 153.4fps avg

His cpu is arguably slower than a sandy bridge at 4700mhz and he still has more fps, because his gpus are slightly faster (or scale better) than your 6990. So no cpu bottleneck.

[maltrabob]

Chew*, thanks for posting your impression and experience. Is there any chance measuring the power consumption of your BD and compare it to that of i7 when gaming? What makes me confused is a difference in measurements people are getting from their BD processors. Thank you.

[chew*]

Thanks. However, you can be bottlenecked by a component to a certain degree, it doesn't have to be complete. It could also mean, that you start to be seriously gpu bottlenecked at roughly 130fps and BD only reaches this mark with tweaking. To analyze this, you would have to do one of two things:

a) overclock sandy further. if the fps rise (almost) 1:1 with the clock increase, you are indeed cpu bound. If not, there is your limit, that now both cpus hit
b) overclock your 6990. if the fps rise, you are gpu bound.

Sorry, but so far it is not conclusive. Also, is it DX9 or DX11?

Here some results for comparison:

http://www.forum-3dcenter.org/vbulle...76&postcount=7
970@4000, gtx 570@800mhz sli: 153.4fps avg

His cpu is arguably slower than a sandy bridge at 4700mhz and he still has more fps, because his gpus are slightly faster (or scale better) than your 6990. So no cpu bottleneck.

Thats all fine and well however the point is at at around the same speed on 2 almost identical systems with identical settings the scores and gaming experience and results are almost identical. aka this is apples to apples, not grapefruits to pineapples

I'm also quite aware if I ran at 1280x1024 results would be different but this aint the 90's and not indicative of the real world


Put simply do you think that guy with 570's hitting 150fps on a monitor capped at 60hz is getting any better gaming experience than myself on these 2 rigs?

It's 64 bit win 7 if you haven't noticed the ram quantity, and the game and OS supports DX 11..........

The goal here is not E-peen fps, just a direct comparison at real world settings that most users will end up with for 24/7

btw your link = inconclusive Sli? 2 way? 3 way? not enough details imo

[chew*]

You sir are correct!



So is the general consensus 4 cores disabled to get better single thread IPC performance? But with 4 cores disabled you less overclocking headway?

Sorry for my ignorance but I'm not very good when it comes to all the CPU technical jargon.

No with 4 cores enabled you get better multithreaded performance versus 2 core 4 cluster.

With 4 cores enabled though you lose oc headroom versus 2core 4 cluster.

[cold2010]

@chew
Yes, in the real world do not play in this one, but when you compare how teams should be gaming performance
CPU VS CPU, you can do this with 1680x1050/HQ/without AA And AF? and For more performance
2600K is suffering because of the GPU Limited With these settings

[boxleitnerb]

Thats all fine and well however the point is at at around the same speed on 2 almost identical systems with identical settings the scores and gaming experience and results are almost identical. aka this is apples to apples, not grapefruits to pineapples

I'm also quite aware if I ran at 1280x1024 results would be different but this aint the 90's and not indicative of the real world


Put simply do you think that guy with 570's hitting 150fps on a monitor capped at 60hz is getting any better gaming experience than myself on these 2 rigs?

It's 64 bit win 7 if you haven't noticed the ram quantity, and the game and OS supports DX 11..........

The goal here is not E-peen fps, just a direct comparison at real world settings that most users will end up with for 24/7

btw your link = inconclusive Sli? 2 way? 3 way? not enough details imo

2 way

I just wanted to analyze your results because on another forum someone was under the impression that with those tweaks BD would have the same IPC as SB, which still is not true. Gameplay experience is the same, of course. However, at a given price point I would always chose the cpu with more reserves, no matter if I need it right now or not. After all, it's all about BD here and BD is a cpu, not a gpu. I cannot stand gpu-limited cpu benchmarks, because they just don't make sense to me. And who is to say that with a 580 SLI you wouldn't have squeezed out more fps from BD in this benchmark? I'm a scientist guy, and as such I am interested in theoretical potential without any shackles. There are always more intensive games and more powerful graphics cards on the horizon and I also want to know how my component will perform then.

[chew*]

@chew
Yes, in the real world do not play in this one, but when you compare how teams should be gaming performance
CPU VS CPU, you can do this with 1680x1050/HQ/without AA And AF? and For more performance
2600K is suffering because of the GPU Limited With these settings

Comparing at resolutions and settings I will not run is not comparing.

I only have 1 concern, will I noticeably lose performance in the games I play at the settings I use with almost identical hardware and realistic clock configurations.

If yes then BD is not for me.

If no then I have no issues using it.

[muziqaz]

@chew
Yes, in the real world do not play in this one, but when you compare how teams should be gaming performance
CPU VS CPU, you can do this with 1680x1050/HQ/without AA And AF? and For more performance
2600K is suffering because of the GPU Limited With these settings

Can we stop with this GPU bottlenecking games ?
He has 6990 for crying out loud, and that game is AMD GPU friendly, that means it likes xfire. I didn't get any bottlenecks with single 6970 and thuban.
Besides, you are not playing games on your CPU alone. there are other system components as well.
Also, isn't it when you apply higher resolutions and settings the workloads increase as well not only on GPU, but on CPU and interconnects as well?

But still, it doesn't change the fact that you loose half the co..(sorry, compute units) to be able to play games at higher fps

[chew*]

Ok so price point since you mentioned it.

We will ditch the ram, GPU, PSU,HD all out of the equasion since it's all the same for the most part save the SSD in the sandy rig and the 16gig matched 2133 set is a tad pricy but no sense crying over spilt milk.

The cpu and mainboard are the leftovers.

2600k newegg $314
http://www.newegg.com/Product/Produc...name=Core%20i7

M4E B3 $309
http://www.newegg.com/Product/Produc...82E16813131700

BD 8150 $279
http://www.newegg.com/Product/Produc...82E16819103960

Ch5 $229
http://www.newegg.com/Product/Produc...82E16813131735

So a person in my position who wants a secondary rig without sacrificing gaming performance at real world settings and already has a sandybridge rig can save $100 bucks.........

If i need to run some super pi for daily use i can use the sandy rig but usually that gets boring after 1 or 2 runs then i go back to actually using my pc

[Oliverda]

I haven't tried with this set since it's mismatched, honestly high speed 16g sets are not as fast as a set that can do 7-8-7 1600....if i went 9-11-9 i'm sure 1866 would be fine though,

I asked because according to BIOS and Kernel Developer’s Guide BD's IMC is not capable for 1866 MHz with 4 modules.

bd_imc.png

[chew*]

I asked because according to BIOS and Kernel Developer’s Guide BD's IMC is not capable for 1866 MHz with 4 modules.

bd_imc.png

Ahh yes that would be due to AMD's memory qualification engineer, I met him in person.

To put it kindly he isn't very open minded. I had mentioned about running thuban stable at 2100 and he told me basically that thats not jedec and it's not possible .

It seems they do 1866 ok for me.

[dess]

^^ Ouch.

Thats sandy bridge with HT disabled so 4 core.........

I think he meant if BD was in 4CU/4C or 2CU/4C mode.

[radaja]

does any know if the FX-41XX/61XX models unlock to an 8 core?
just curious as to whether unlocking works or not.
thanks

[chew*]

^^ Ouch.


I think he meant if BD was in 4CU/4C or 2CU/4C mode.

Yep that was 4cu 4core since i already know it's pointless to run that game otherwise.

[boxleitnerb]

Ok so price point since you mentioned it.

We will ditch the ram, GPU, PSU,HD all out of the equasion since it's all the same for the most part save the SSD in the sandy rig and the 16gig matched 2133 set is a tad pricy but no sense crying over spilt milk.

The cpu and mainboard are the leftovers.

2600k newegg $314
http://www.newegg.com/Product/Produc...name=Core%20i7

M4E B3 $309
http://www.newegg.com/Product/Produc...82E16813131700

BD 8150 $279
http://www.newegg.com/Product/Produc...82E16819103960

Ch5 $229
http://www.newegg.com/Product/Produc...82E16813131735

So a person in my position who wants a secondary rig without sacrificing gaming performance at real world settings and already has a sandybridge rig can save $100 bucks.........

If i need to run some super pi for daily use i can use the sandy rig but usually that gets boring after 1 or 2 runs then i go back to actually using my pc

No one gets a 2600K for gaming. The cheaper 2500K is the way to go. And you can find similarly priced mainboards for both of them. Who pays more than 150 bucks for a mainboard anyway.
What do you want with super pi? I told you, I see the benefit of a faster cpu in

• feeding CF/SLI
• feeding future graphics cards
• running cpu intensive games (which dirt 3 is not) like Arma 2, Starcraft 2, GTA 4, BFBC 2/3 Multiplayer, strategy and simulations and so forth
• avoiding fps drops below 60 or 120fps (vsync and vsync+3d)

My point is: If you can get more performance for the same or a very similar price, it makes no sense to choose the slower product.

[cold2010]

I think this will be answered enough to know some of what I mean

On to the numbers! Our testbed consisted of a Core i7-2600 processor, running at its stock 3.4GHz clock speed

http://www.pcworld.com/article/22163...ly_please.html

2600K does not need to work in 4.7 in order to give the frame 123, with HD 6990, can get this in 3.4
The test page in the past is because the GPU Limited 2600K Gets on 131 frame in 3.4 with HD 6990

[chew*]

I think this will be answered enough to know some of what I mean





http://www.pcworld.com/article/22163...ly_please.html

2600K does not need to work in 4.7 in order to give the frame 123, with HD 6990, can get this in 3.4
The test page in the past is because the GPU Limited 2600K Gets on 131 frame in 3.4 with HD 6990

Lest we forget the test bed is different......diff AA no mention of AF, high quality is mentioned but there is an ultra setting......

Pre release 11.4 reviewers drivers for 6990 launch

And a stock 6990 with a stock cpu is getting 10fps more than my setup and your not questioning that but your questioning me...........

Seriously, enough with the pineapples melons and pumpkins compares.

[omninmo]

Chew, do you think there is a significant difference between running with 4 cores disabled in bios and running all 8 of them but using something lile the thg task manager i posted a while back to force thread affinity to 4 different cu? Perhaps even more performance to be had by allowing os and other background tasks to be offloaded to other ccores? Or the performance benefit from 4c/4cu only happens when the remaining clusters are hard disabled?

Thank you for your time!

[chew*]

Chew, do you think there is a significant difference between running with 4 cores disabled in bios and running all 8 of them but using something lile the thg task manager i posted a while back to force thread affinity to 4 different cu? Perhaps even more performance to be had by allowing os and other background tasks to be offloaded to other ccores? Or the performance benefit from 4c/4cu only happens when the remaining clusters are hard disabled?

Thank you for your time!

Needs to be hard disabled.

So back on track now........

No scaling with cpu eh?

Seems rather clear to me that at a certain point minimum frame rate continues to scale.........

[cold2010]

Lest we forget the test bed is different......diff AA no mention of AF, high quality is mentioned but there is an ultra setting......

Pre release 11.4 reviewers drivers for 6990 launch

And a stock 6990 with a stock cpu is getting 10fps more than my setup and your not questioning that but your questioning me...........

Seriously, enough with the pineapples melons and pumpkins compares.

What is the relationship of the AA and AF in what I mean? Of the strong cpu stock speed ؟
I mean that 2600k, Can get same frame here at the stock 3.4GHz without need to work in 4.7
This test was carried out with yourself and you will know the answer

[boxleitnerb]

Minimum fps may scale, because as I said before, you can be bottlenecked only to a certain degree. Meaning, there may be scenes in the rather long benchmark, where the cpu does indeed play a temporary role. But in the grand scheme of things, this benchmark is gpu bottlenecked with your hardware and settings. Run something else that is more cpu intensive and the difference between BD and SB will be there again.

[dess]

Needs to be hard disabled.

Why could be this? I mean, what's the difference between disable in BIOS and preventing threads execution on the second cores? Perhaps it's what haylui suggested, so that the decoder supplies the first core in every cycle in the former case?

[omninmo]

Having to actually hard disable the cluster through bios is such a buzzkill
i was rather hoping that setting affinity would suffice...

Ita weird though, this sort of clashes a bit with your finding that for example games with 2 heavy threads andd 2 light threads wouldn't suffer much of a performance hit.. oh well, thats life for ya..
What really aggravates me atm is that here in portugal i can get a 1100t for 155€, a 2500k for 175€ but the very cheapest fx8120 available is priced @ 209€!!! Even if i could forgive the performance, this is just too much! Im gonna wait till xmas, see how prices evolve and how the radeon 7k series stand in terms of price performancebl before i plunge for an upgrade, i suppose...

[Spectrobozo]

Ok so price point since you mentioned it.

We will ditch the ram, GPU, PSU,HD all out of the equasion since it's all the same for the most part save the SSD in the sandy rig and the 16gig matched 2133 set is a tad pricy but no sense crying over spilt milk.

The cpu and mainboard are the leftovers.

2600k newegg $314
http://www.newegg.com/Product/Produc...name=Core%20i7

M4E B3 $309
http://www.newegg.com/Product/Produc...82E16813131700

BD 8150 $279
http://www.newegg.com/Product/Produc...82E16819103960

Ch5 $229
http://www.newegg.com/Product/Produc...82E16813131735

So a person in my position who wants a secondary rig without sacrificing gaming performance at real world settings and already has a sandybridge rig can save $100 bucks.........

If i need to run some super pi for daily use i can use the sandy rig but usually that gets boring after 1 or 2 runs then i go back to actually using my pc

how about a 2500k and a cheaper MB? you set HT off anyway..
how about someone that have a faster VGA (2x, 3x 580), or will buy a new VGA next year? or play more games? or lower the details for higher framerate (one example, 120fps for 120hz, is it possible with these CPUs with lower settings to achieve a minimum of 120?)
if you want to test at those settings OK, but you could lower the resolution to and anti aliasing to see what happens, it's a valid and easy thing to do, also you could lower the i7 clock to analyze the impact, and test the FX with the 8c enabled... why not?

[dess]

or lower the details for higher framerate (one example, 120fps for 120hz, is it possible with these CPUs with lower settings to achieve a minimum of 120?)

So you're willing to play at 640x480?

ps. you don't need 120 fps for 120 Hz, if your goal is 3D. The GPU will render to two framebuffers at any rate and it's the two buffers that will be alternated at 120 Hz.