chew* I think games with only two threads may run best in 2CU/2C mode, instead of 2CU/4C, where sharing is present but not needed, as there are only two demanding threads. And with two cores disabled, there should be a little more room for OC, as well.
I think you'd want a game with three threads game where the first thread is 100% set to a single CUQuote:
Originally Posted by dess
and the other two that are around 50-60% for the other threads both on the same CU.
set threw taskmanager or over drive.
Possibly but I'm using the other threads to offload other tasks.Quote:
Originally Posted by dess
Hi Chew,
It's a little bit OFF topic but have you been able to run four modules at 1866 MHz on BD?
@ 6.6.6.18? :dQuote:
Originally Posted by oliverda
Is this exclusive to the CHV, or does it work on the Sabertooth and M5A99X-EVO?
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,Quote:
Originally Posted by Oliverda
How's this for almost apples to apples :)
Sandybridge ( The only diff is this system has a 120gig boot/OS SSD higher ram but a tad looser)
http://www.xtremesystems.org/forums/...5&d=1318888545
BullDozer optimized NB and ram clocks now.
Attachment 121415
Nice, would be great once you have done testing to share your optimal bios setup :)
Wow nice comparison there Chew. So close between BD and SB. Keep it going
Brian
that is crazy!!!
I have a question:
when BD is running 1 thread per module, would the shared Fetch/Decoder works 4 instruction per cycle without switching in every clock cycle??
IMO, if BD still switching every clock cycle even one of the core is idle/disable, then the 20%~35% performance gain while BD is running on 1C/1M configuration is very impressive.
Imagine, if a BD design with full time Fetch/Decoder feeding to only 1 core like conventional design.
hehe here comes the sandy killer finally lol
turbo involved?
@haylui: I think the shared Frontend will now work fully for only 1 core. I guess its the frontend rather then the FPU thats limiting BD by sharing ressources...
No turbo, full clocks on both rigs.Quote:
Originally Posted by Oese
Is it possible to post a benchmark and settings for a lower oc at around 4 GHz?
Does the core deactivating is a Crosshair V Formula only feature, or is there a chance to find it on a cheaper board???
I would like to start my own tests, but i dont know.
To be an alternative for i5 or i3, it is not very cool, to take a >200 $ board and pair it with a 100 $ FX-4 CPU and oc it so much. The bill from the energy supplier is part of the package too.
So your smart nb oc and 4 GHz on 3C/3T or 2M/2T, could be an alternative in dirt 3 for example. A test would be very helpful against i7 with resolution and quality specs and with simulated FX-4 @ stock. The results would be very, very helpful to compare it to other reviews, with more games testet.
Thanks for the great testing.
Here some results from rage3d => FX@nb oc:
Battlefield
Min fps Avg fps
109,50% 108,9 %
Civilization 5
175% 110%
Crysis 2
100% 106%
Medal of Honor
114% 117%
Add the 3CU/3C performance gain of 10-20 % and in the end, FX IS a gamer CPU.
aha coolQuote:
Originally Posted by chew*
What graphics card are you using? And what settings?
Asrock boards have the "core leveling" feature available too atleast (dunno if it is enabled thou).
Other manufacturers can easily add it too, if they only want to.
1920x1080 maxxed out settings with a 6990 in each rig.Quote:
Originally Posted by boxleitnerb
Low quality settings might change things but why test at settings I won't use in real world ;)
Quote:
Originally Posted by Oese
do you think the frontend has very less effect on the performance difference?
but I think FPU really affected much since most of the desktop apps are FPU intensive
Is this 2cu/4core or 4cu/4core for the bulldozer test?Quote:
Originally Posted by chew*
Thats sandy bridge with HT disabled so 4 core.........Quote:
Originally Posted by EniGmA1987
What graphics card did you use and what resolution, AA/AF was applied? You know you are most likely GPU-limited unless you have a CF-rig.
Quote:
Originally Posted by chew*
6990 is 2 6970's crossfire last I checked :)Quote:
Originally Posted by boxleitnerb
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
You sir are correct!Quote:
Originally Posted by chew*
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.
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:Quote:
Originally Posted by chew*
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.
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.
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 pineapplesQuote:
Originally Posted by boxleitnerb
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 ;)
No with 4 cores enabled you get better multithreaded performance versus 2 core 4 cluster.Quote:
Originally Posted by ohms
With 4 cores enabled though you lose oc headroom versus 2core 4 cluster.
@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
2 way :)Quote:
Originally Posted by chew*
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.
Comparing at resolutions and settings I will not run is not comparing.Quote:
Originally Posted by cold2010
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.
Can we stop with this GPU bottlenecking games :banana::banana::banana::banana:?Quote:
Originally Posted by cold2010
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 ;)
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 ;)
I asked because according to BIOS and Kernel Developer’s Guide BD's IMC is not capable for 1866 MHz with 4 modules.:down:Quote:
Originally Posted by chew*
Attachment 121436
Ahh yes that would be due to AMD's memory qualification engineer, I met him in person.Quote:
Originally Posted by Oliverda
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 :p:.
It seems they do 1866 ok for me.
^^ Ouch. :)
I think he meant if BD was in 4CU/4C or 2CU/4C mode.Quote:
Originally Posted by chew*
does any know if the FX-41XX/61XX models unlock to an 8 core?
just curious as to whether unlocking works or not.
thanks
Yep that was 4cu 4core since i already know it's pointless to run that game otherwise.Quote:
Originally Posted by dess
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.Quote:
Originally Posted by chew*
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.
I think this will be answered enough to know some of what I mean
http://www.abload.de/img/1xfxu.jpgQuote:
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
Lest we forget the test bed is different......diff AA no mention of AF, high quality is mentioned but there is an ultra setting......Quote:
Originally Posted by cold2010
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.
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.Quote:
Originally Posted by omninmo
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.........
http://www.xtremesystems.org/forums/...5&d=1318888545
Attachment 121446
What is the relationship of the AA and AF in what I mean? Of the strong cpu stock speed ؟Quote:
Originally Posted by chew*
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 ;)
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.
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?Quote:
Originally Posted by chew*
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...
how about a 2500k and a cheaper MB? you set HT off anyway..Quote:
Originally Posted by chew*
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?
So you're willing to play at 640x480? :rolleyes:Quote:
Originally Posted by Spectrobozo
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.
For me i always loved AMD for the price per performance relation and after seeing what you did i am sure that BD will give me the same. Now that they support SLI again i am happy to be back. As you indicated chew- in games the BD is on par with the SB and that is all i want. The BD performs better than my I7 setup on most of the test .
Thanks Chew and others here for all your tests and feedback -helped making a good choice
And older gen AMD isnt officially capable of DDR2000+ :rolleyes:Quote:
Originally Posted by Oliverda
Quote:
Originally Posted by dess
so for you the only options are 640x480 or 1920x1080?! :rolleyes:
and I thought this was a forum for enthusiasts, that care to explore the CPU performance in more ways, than saying "it's enough for my games", because the real word is far more complex than that, you have a lot of different games and situations within the game, and by lowering some settings you can more or less simulate a faster VGA on higher settings, or simply use it as a more valid CPU benchmark,
if you are willing to test OC and fine tuning gains I also think its valid to lower the settings and observe the gains more easily,
I'm not saying that testing on these settings have no use, what I'm saying is, how much time does it take to run one or 2 more settings since you already have all running? 3 minutes? so why not? its more information,
I'm not talking about 3d, I'm talking about using a 120Hz screen with vsync on,
and I'm just curious to see what happens when the GPU impact is lower and the CPU impact higher,
also when talking about price, don't forget the 2500k...
I'm also curious about 4m/8c vs 4m/4c, from what I see on the other pages only 2m/4c vs 4m/4c was tested,
anyway, thanks for the tests, but that's just my opinion.
3D Vision generally requires 120fps 2d performance if you hope to have a fluid 60fps 3D experience. Sure you can play at lower framerates but sub 60fps with 3D is the suck as far as I'm concerned.Quote:
Originally Posted by dess
And that's quite a stretch to assume you'd need to play at 640x480. There is a reason I'm running my system only at 1680x1050... Spectrobozo poses a very valid question.
At 1920x1200 with full IQ (ie gpu limited) this 580 performed roughly the same on both my 920 and 2600k when it came to sufficient 60hz monitor performance. Now when you bring 1680x1050 and the 120hz performance ceiling into the equation ( of which is less gpu limited ) the 2600k pulls ahead by a fair margin ( I also often run games with little to no AA as to have consistently high performance )
Most reviews tend to do both extremes ( low IQ and resolution or high res and IQ ) and not a middle ground. Techpowerup and HWC are two of the better outfits where this is concerned so its easier to see the full picture. I bought a 580 for 1680x1050 because its capable of a much larger performance gap at this resolution relative to the other single gpu options available.
All this said, I'd like to see more middle ground results comparing various architectures clock for clock. I want to believe the reason why we don't see more of this is 60hz is still the defacto standard and people don't seem to care as much how things perform at lower resolutions ( the whole " More than 60 frames per second is a waste argument " )
EDIT:
^ This 100%Quote:
Originally Posted by Spectrobozo
Comparisons are never straight-forward, what with software acting as erratic as it does. Blender in Linux works so much better on AMD hardware than in Windows (refer to review by Johan at AT), it makes you think how many things are borked in Windows. So everything is relative. If it works for you good. If it doesn't, it doesn't mean it is not good, but merely that it doesn't work for you. Buy what you must and stop pushing your opinion on others.
All the Intel shills quibbling over a difference of 10 odd fps (yes, this is slightly exaggerated) at resolutions nobody plays at, is starting to be a buzz-kill in an AMD thread. You want to buy Intel, then do so. You want to talk about what's screwed up in BD, then do so. However, please don't troll someone who's actually doing a whole lot of favours for all of us here at XS. There were also twerps trolling Movieman, and now i see some people hassling chew*. I'm dead certain that there's a reason why AMD called him for their world record run. He's being modest, but that doesn't stop me from praising him.
Well Bulldozer wont be pulling ahead of SandyBridge in games, and 2500k or 2600k users would do a mistake to change for a BD setup. Thats not the question right now.
This thread is not about AMD vs Intel or what is better for gaming, it is about exploring the BD, its bottlenecks, and tweaking capabilities. Still more of an analysis then of a claim "its just as fast as Sandy". If I say such, its more like a joke, noone here is of the opinion BD is actually the better processor. But the more interesting one.
Me and I think most People here are interested in the tech mostly... Seeing BD performing kinda comparable under some tweaking gives an insight on where it is limited in stock settings.. Nothing more nothing less.
Ah well, and its giving hopes to people that wanna have one anyway, that they gonna be able to do some gaming actually and dont need a second rig for that :) - and what they can do about tweaking it...
Like it or not....BD already released now:)
Anyone can buy and test it...then give their opinion good or bad with real facts not only based on reviews
Let BD be as is...AMD's new chip in good or bad ways depending on your own experience with it:)
Why are you even posting your results here then? If you only bench these two systems to see if they can play dirt over 60Hz why post it here?Quote:
Originally Posted by chew*
I thought you posted here because you wanted to share about the performance difference between BD and SB when tweaking the BD. If you hide that performance difference with GPU limitations because you don't care about the performance difference why bother post it? Don't you want to know how large differences you can make with tweaking? That there is no real world difference in one GPU limited game doesn't mean that it won't be differences in other games. It's not apples to apples if one apple is a pineapple viewed from another angle. You can make the pineapple look like an apple how much you like but it will never be an apple.
Just as a PIII 800MHz and an 990X isn't apples to apples only because they score the same when you play Quake III at high resolutions with an Geforce 2 MX.
Please, just run a bench in low resolutions to see how far behind BD actually is with these tweaks. It's very relevant to many of us.
Full ack. At least run without AA/AF because they don't have anything to do with the CPU. Btw, how do I give thanks for a post? I didn't find any button in this forum.
Lower left, just above "Reply to Thread".
No, but for 120fps in many recent games you may need to lower the settings as much.Quote:
Originally Posted by Spectrobozo
Here you can find such tests as well with several games that showed there is no gain running games with less than 5 threads in 4CU/8C "mode". It could even impact performance negatively, in some cases. A special game with 8 threads of course liked it.Quote:
I'm also curious about 4m/8c vs 4m/4c, from what I see on the other pages only 2m/4c vs 4m/4c was tested,
For fluid 60fps, yes. But, I had such a kit a few years back and it wasn't as much bad experience with lower framerates, like 30fps. (And I'm not such a guy that can't tell the difference between 30fps and 60fps, in fact I clearly distinct 60fps from 120, as well, on a 120Hz display.)Quote:
Originally Posted by Chickenfeed
Of course, it all depends on the game, how much one needs to lower settings. There and now many very demanding games, GPU wise.Quote:
And that's quite a stretch to assume you'd need to play at 640x480. There is a reason I'm running my system only at 1680x1050... Spectrobozo poses a very valid question.
Yes, it all depend on the workload. I just hate when people make general claims that "BD sucks!".Quote:
Originally Posted by Dumo
Is it Firefox that swallows it, or bound to a certain rank?Quote:
Originally Posted by DedEmbryonicCe1
I think you need to get past 100 post before can thank anyone.Quote:
Originally Posted by boxleitnerb
If an application were to change affinity at a rate equal to the time slice of a thread from odd to even cores, could this reduce the heat enough to get a higher average Turbo? Is it module temperature or core temperature that is used decide to Turbo or not?
Here are some 2CU/2C tests with a 4100, with only affinity adjustments. It seems the thing is effective enough (+9-11%, compared to 1CU/2C) even without disabling the cores in the BIOS! :yepp:
Btw, regarding the searching for the cause of CMT not scaling in many times as well as it "should", part of the "effect" could be the so called L1D invalidation issue, which has been mentioned in another thread here (in case someone missed it). Although, AMD is claiming it accounts really for a slowdown of up to only 3%, but what if they're underestimating it? Anyway, there is a patch underway that will prevent it from happening, so at least some of the performance could be regained then, without disabling cores, benefitting also programs with 8 threads (for which the 4CU/4C mode wasn't a solution).
What motherboards have the option to disable cores? My Gigabyte 990XA doesn't appear to have the option.:mad:
Look the post one above yours. :)Quote:
Originally Posted by Baam
You can do that with a little program that were linked here a few days back.
BD is 4 clusters off to.......2600K is more effec than 2500K and faster in games which is why i bought it.Quote:
Originally Posted by Spectrobozo
HT off is really not pertinant here since this game loads 4 cores at about 75%.
Feel free to buy a 2500K and mobo for me just for this testing that you want to see ;) I already had a 2500K and sold it after doing a similar compare.
Like i said i already know what happens if i drop res, and it's a moot point since I won't actually game like that......
I'm in no way saying that BD will beat 2600K but it is an alternative..........so is thuban or deneb, I have in no way told people they should sell off there current sandy,thuban or deneb rig becasue BD is the greatest........I have merely pointed out it's an alternative.
As far as other compares with cards I have 3 580's but these systems are 24/7 rigs not bench rigs and i'm not changing my config out just for a compare + all my 580's have tri slot coolers so 3 way is not possible.
2x6990 would not be hard but I've already run that config in my sandy rig and it has alot to be desired and is pointless.