More MHz for the Masses
AMD had a small refresh of the Bulldozer series and we dig in.
AMD has had a rough time of it lately when it comes to CPUs. Early last year when we saw the performance of the low power Bobcat architecture, we thought 2011 would be a breakout year for AMD. Bulldozer was on the horizon and it promised performance a step above what Intel could offer. This harkened back to the heady days of the original Athlon and Athlon 64 where AMD held a performance advantage over all of Intel’s parts. On the graphics side AMD had just released the 6000 series of chips, all of which came close in performance to NVIDIA’s Fermi architecture, but had a decided advantage in terms of die size and power consumption. Then the doubts started to roll in around the April timeframe. Whispers hinted that Bulldozer was delayed, and not only was it delayed it was not meeting performance expectations.
The introduction of the first Llano products did not help things. The “improved” CPU performance was less than expected, even though the GPU portion was class leading. The manufacturing issues we saw with Llano did not bode well for AMD or the upcoming Bulldozer products. GLOBALFOUNDRIES was simply not able to achieve good yields on these new 32 nm products. Then of course the hammer struck. Bulldozer was released, well behind schedule, and with performance that barely rose above that of the previous Phenom II series of chips. The top end FX-8150 was competitive with the previous Phenom II X6 1100T, but it paled in comparison to the Intel i7 2600 which was right around the same price range.
Perhaps the biggest disappointment for many users was the launch of the FX-6100. This chip originally sparked a lot of interest as it was a 3 module/6 core product with a 95 watt TDP for under $200 that promised to be something of a budget enthusiast’s dream. When performance figures came out, the chip did not hold up to the lofty expectations held by users. At a stock 3.3 GHz, it was simply outclassed by products such as the i5 2500 and 2500K variants, which were again in the same price range. Even though the processor had a maximum Turbo Core of 3.9 GHz, it just could not keep up in most applications with the smaller and less power hungry Intel Sandy Bridge based products.
Hope springs eternal though, and when news leaked out that AMD would be introducing refreshes for the FX-4000 and FX-6000 chips people were expecting there to be an improvement in thermals and perhaps even a minor revision update to the design. The chatter turned to potential per clock improvements, better overclocking headroom, and a couple of products under $200 that could hold their own against Intel’s mighty Sandy Bridge. Then the actual details came out. These chips were of the same revision as the previous parts, and the speed increase they saw was due to the TDP ceiling being raised from 95 watts up to 125 watts. Still, there was hope that these products could more than hold its own against the i3 and i5 in the same price range. Does the FX-6200 succeed? I guess we are about to find out.
The FX-6200
Ryan and I have covered the Bulldozer architecture in previous articles, so I will not go over the finer details here. I will discuss some of the basics of the chip and the architecture.
The FX-6200 is based off of the same Bulldozer revision as the previous FX series of parts that was released last October. There may have been some minor changes along the way, but they would have more to do with manufacturing rather than any kind of base silicon or extreme metal layer change. The product is still built by GLOBALFOUNDRIES on their 32 nm HKMG SOI process. The chip is a native 4 module/8 core product, but one of the modules has been fused off and is unavailable for unlocking. This leaves 3 modules/6 cores for the processor to work with. Each module features 2 MB of L2 cache to be shared between the two integer units and the single FPU/MMX/SSE/AVX unit, for a total of 6 MB of L2 for the entire CPU. The L3 features the full 8 MB of cache that is available on fully enabled CPUs.
The full meal deal of a retail/boxed processor. The 3 year warranty and heatsink/fan are likely worth it to most users.
The base clock is now set at 3.8 GHz, which is faster than the 3.6 GHz stock clock for the FX-8150. To achieve this clock AMD raised the TDP to 125 watts, which is up from the 95 watts of the FX-6100. The max Turbo Core speed of the FX-6200 is 4.1 GHz. AMD did raise the Northbridge/L3 cache speed on the FX-6200, and it runs at 2.2 GHz rather than the 2.0 GHz that the FX-6100 was set to. The memory controller is heavily revised from the previous Phenom II generation of parts, and it can handle official speeds up to DDR-3 1866 MHz.
The retail package comes with a pretty hefty heatsink and fan combination that should be able to handle the extra thermal load that the 125 watt CPU provides. The CPU has a standard 3 year warranty for the boxed version, while a tray chip has a 30 day warranty. The boxed product retails for $169 US, but can often be found cheaper with instant rebates and sales.
The FX processors are all unlocked so it makes them very easy to overclock. Most retail motherboards cover all of the settings needed to increase the multipliers on the CPU and overclock the chip. Success of any overclock is not guaranteed and the choice of motherboard will also directly affect the ability of the chip to reach higher speeds.
Thanks Ryan for doing this i
Thanks Ryan for doing this i bet it wasn’t that easy to get because i don’t thank amd sent these out for people to review. thanks for the review just wish there was at least one intel cpu i the line up. but good job and i know it would lose to probably a i3 too.
Ryan didn’t write this…
Ryan didn’t write this…
He didn’t say Ryan wrote
He didn’t say Ryan wrote this, he said Ryan did this, as in making the decision to review this CPU.
Great article, Josh!
Great article, Josh! Thanks.
Its sad to see that AMD just can’t seem to catch a break on the processor side. I was really hoping that AMD would be able to push back with this part.(I like to root for the little guy)
I also have to agree that it is pathetic that they cant beat out previous generations, even with a decent overclock. Those were the strangest results to me. If the new processors can’t beat out previous generations, maybe don’t release it.
If Trinity still can’t beat out the X6 I’m going to suggest that they just give up on processors. Or re-release the Phenoms on the smaller process.
Thanks Nil!
Yeah, I wonder
Thanks Nil!
Yeah, I wonder what it would take to redo the Phenom IIs at 32 nm. Probably not really worth the effort since Bulldozer is in full production, and is finally yielding good enough to meet demand. Plus Piledriver is almost out.
On the positive side (at least we hope it will be positive), Trinity will be displayed to journalists next week, and I would imagine (though I don’t know the timeline) that we will see initial reviews within a month or so. Definitely before Computex. The rumor is that desktop Trinity will be released after that time on the new(ish) FM2 platform.
Vishera is a big question mark as well. I have no clue on that particular timeline… but sooner would be better…
I think it maybe worth it to
I think it maybe worth it to redo the Phenoms. Judging by benchmarks by everyone, the Phenom architecture looks to be much more promising than what Bulldozer has to provide, and possibly more flexible. Also a die shrink would provide improved thermals, possibly more OC headroom, ect. I think like Intel, AMD needs a kind of fallback architecture similar to the Pentiums. You know something tried and true, so to speak. It maybe worth it.
I appreciate that AMD is trying to rethink everything we know about processors, and innovate again with new ideas like Bulldozer.
On the other hand if the new parts you were betting on aren’t up to par, maybe they should think about re-releasing successful architectures, ALONG with the new ones.
I can’t help but think of Bulldozer as a beta type architecture, maybe even alpha.
I have not heard of Vishera, is that another entirely new architecture or based off Bulldozer or Piledriver?
Vishera is the AM3+ version
Vishera is the AM3+ version of Piledriver. It will give up the integrated graphics portion and likely add L3 cache (Trinity does not have a L3).
The 6100 and 6200 are six
The 6100 and 6200 are six cores like the top phenom II’s but they were not made to replace them and debuted at much lower cost that the Phenom six cores. They are one of the budget bulldozers (along with the FX 4XXX). Both do however now have higher passmark scores, and they are GREAT buys for the price.
Hey Josh, great review. I
Hey Josh, great review. I think the Euler 3D tests show quite clearly that in certain workloads the Bulldozer modules are simply choked when both cores/module are used. It probably is the scheduler and the FPU that probably are the main bottlenecks, though the high cache latency can’t help. It seems the long integer pipe and the high clocks aren’t the performance hindrance one may have thought (P4 style)
I’m not very sure the automatic design has anything to do with the lack of performance. Say the routing is 20% larger – this would reflect in die surface – so less profits and more power consumption, but the performance hit would be indirect at best. 20% less performance – this is debatable – I really doubt it means 20% less performance per clock. It probably means some level of common mode or crosstalk that the automatic routing can’t detect but an engineer could given enough time – and this probably is translated in 20% less frequency headroom due to parasitic coupling or the chip hits the power ceiling at lower clocks. I just can’t see how an under-performing FPU can be blamed on automatic routing tools unless their internal CPU simulators are rubbish (which can be blamed on being under-staffed).
So to conclude my little analysis I would say that even using manual routing the CPU would still have flopped due to starved cores, undersized schedulers and FPU, these are all design decisions. They may have reached 4GHz@95W on all cores, but performance per clock couldn’t have been much different.
Can’t wait for the Piledriver core, hope it gives the same boost as Deneb over Barcelona.
Good points. I was just
Good points. I was just sorta going off that conversation another site had with an ex-AMDer who spoke up about the use of automation at AMD for processor design. I guess I was thinking more along the lines of “20% slower” not so much in clockspeed, but in IPC terms. I know that seems somewhat nebulous (and it really is), but when looking at these results… this is really pretty extreme. The FX series are disappointing. Seemingly a poor use of die space and transistor counts to achieve performance that is similar (in pretty much all ways) to the previous generation.
The one thing that I can’t really get over is the fact that the Bobcat architecture came out so good. It really was a tremendous product for AMD in 2011. For Bulldozer to come out this way… that just seems really odd.
IT seems that AMD is where
IT seems that AMD is where Intel was with its P4.
Bulldozer underperforming and overpowered (Watts) just as was the P4. Then Intel looked back at the architecture that worked the, p3, and the rest is history.
Now if AMD can look at their Bobcat architecture and see what it can do on the destkop side.. (ok wishing)… but it seems that history is repeating it self. I personaly would hope that AMD looked back at the Bobcat and not the Phenom II.
Still as you mentioned the chance is next to none since their production is in full swing with the Bulldozer and soon to be Piledriver. Even if then cam get a Sandy Bridge performance out of the Piledriver, I would then stick in the AMD camp.
However some “leaked” Trinity benchmarks are not that promissing. Still as Trinity is L3 deprived, there is still hope that Piledriver AM3+ CPU’s can still bring something useful to the consumer table.
In the end, you and the entire PC Per team is amazing. Can’t wait for the next Podcast (and here is to wishing that it gets a tad longer) … Don’t go there Josh 😛
Yes, it certainly is
Yes, it certainly is interesting to see that AMD is at nearly the same place as Intel was with the jump from Northwood to Prescott. Double the complexity, no real improvement in performance, thermals were not real good, and they were harder to make. It took a while for Intel to fix those things, I’m hoping it won’t take AMD nearly as long.
Longer podcasts… you mean 1.5 hours of our drivel isn’t enough? Yes, I went there…
No… not long enough …
No… not long enough … lol….
I’m looking forward more so
I’m looking forward more so for “steamroller” for the AM3+ platform. Not official yet but possible.
A lot of AMD forums talk
A lot of AMD forums talk about heat issues with these processors. One thing missing in the above review is its performance with a higer-end cooler installed. Some of the forum participants were able to significantly improve the performance and mitigate some of the increased power consumption via improved cooling of the CPU versus the stock cooler.
I just wish you guys wouldn’t
I just wish you guys wouldn’t trash AMD when they come out with parts that are comparable to Intel at the same price point. I know we all want AMD to really outperform Intel, but then we can’t go badmouthing their products when they don’t excel, but “merely” deliver.
I don’t buy AMD because they are better than Intel, though I wish they were. I buy AMD because I know I’m getting a decent part and it specifically isn’t an Intel part. My dislike of Intel goes back to the Pentium 90 I owned that couldn’t even do math, we were later to find out. At the time I was fresh into college and my dad paid over $2000 for the total unit when these days thanks the AMD competing with Intel I can make a decent high end desktop at half that price.
Rooting for the underdog means you balance your expectations while understanding that they are an underdog and if they can get a decent part out then you congratulate them for keeping the big dog’s nose to the grindstone. While Bulldozer is not stellar, it does compete more or less in it’s price point. Bombing their products like this could cost them the place they hold in the market at the moment…
Excuse my negativity please, I am really hoping AMD stays in the market and worried that they may end up doing something else.
That is so true Prestaeus,
That is so true Prestaeus, only because if you install and new computer AMD or Intel parts your are going to have a speedy machine. That is light-years ahead of last gen. If it plays the games you want it to play or encode faster than your last rig, then you did alright! Modern CPUs are great at multitasking so you can do more than one thing at a time. Ive seen AMD products that were playing HD podcast, and encoding Blurays. Also while surfing the web, just as good as my i5. So I don’t see any issues with today AMDs products…
Don’t get my wrong, I am
Don’t get my wrong, I am rooting for AMD as well. Not because I am somehow in love with the company, but rather I feel we need a decent company that can stand (mostly) toe to toe with Intel. I remember back in the mid 90s that Intel’s roadmap was amazingly conservative. They were planning in the 1998/99 timeframe of having a Katmai type processor running at 400 MHz max. AMD came along with the Athlon and suddenly we were pushing 800 MHz to 1 GHz in that time frame with much more aggressive parts (and pricing) from Intel.
My analysis and final determination of the product is based on the following factors. The FX-6200 is being offered at around the same price as the 1100T was when it was EOL. So, consumers are getting lower performance for the price as compared to last gen products. The die size of Bulldozer is approximately 315 mm square, which is not far from the 346 mm square of the 1100T. This is for a processor that is based on the 32 nm HKMG/SOI process, and not the older 45 nm that the 1100T is on. A full node die shrink is around 45%, so on 32 nm the 1100T would take up approximately 190 mm square. Bulldozer is obviously much bigger and more complex than that, but performs worse. Finally, the only time the FX-6200 really overcomes the 1100T is when overclocked to 4.5 GHz. At that point the FX product pulls 104 more watts at the wall than the 1100T it barely outpaces.
I want AMD to succeed. I do not expect them to pull a rabbit out of their hat with each generation of product, but I do at least expect them to be relatively competitive with what Intel has. The Phenom II was pretty comparable to what Intel had with the Core 2 Quad series and the later Nehalem based processors. They were very competitive (if not at an advantage to Intel) in price/performance. But Intel moved on with Sandy Bridge, and now the upcoming Ivy Bridge. AMD has frankly been stagnant in what it offers. So, while at $169 the FX-6200 is not a bad processor, when we compare it to the i3 and i5 processors in that price range, it just does not stack up in performance or power consumption. Also, going back to die size, the FX-6200 and FX-4170 dwarf the smaller, cooler, and more efficient Intel products. That is not good for AMD, and it is certainly not good for customers.
So, my point was not to be mean to AMD, or unjustly punish them. My point is that the truth of how it performs and what kind of tradeoffs AMD had to make to get these to market needed to be spotlighted. Yes, we now have more power in our computers that we barely tap into most of the time, but that does not give either AMD or Intel the right to just sell us any old dog. All technology moves forward, and in this case AMD essentially took a step backwards.
I am thinking that this is not news to AMD. The reason why we have seen so few FX based reviews past the original FX-8150 is because AMD just stopped sampling CPUs for review. We had to purchase this CPU ourselves to get the review out. AMD is not pushing AM3+ products, and they are saving their budget and energy for the upcoming Trinity. After next year, there will not be a standalone CPU type socket for AMD, it will all go to APU type products from here on out. AM3+ will get a Piledriver based update, but then it will be phased out next year.
AMD is in trouble, but its troubles are far more than little old me giving the FX-6200 an honest review.
but they arent as good at
but they arent as good at doing it as the i5 – they’re using more power to do it.
Josh one day I will meet you
Josh one day I will meet you irl and I will rub your bald head, and not in a gay way either… sorry
I will be on the lookout for
I will be on the lookout for random head rubbers (not in a gay way).
Excellent article, Josh! I
Excellent article, Josh! I had high hopes for this CPU, so I am a little disappointed with it’s performance (I really thought the higher clock speed would give it an edge).
I had originally planned on making a budget computer with this proc at the center of it. Now I need to re-think my plan of action. I wanted to use AMD (not b/c I’m a fanboy, but because I fear AMD won’t be making CPUs much longer if they don’t get some support from the computer community). I may still make the 6200 the proc, but use it as a learning experience for OCing (something I’ve never done).
Love the podcast – keep up the great work.
The FX-6200 is not a terrible
The FX-6200 is not a terrible choice, and if you want to learn how to overclock it is a lot of fun that way.
I agree with you, AMD needs support and people buying their products. I’m hoping Vishera is a bigger leap forward for the company than what I am fearing will actually happen.
I meant to ask
I meant to ask previously…how were the temps, especially when you OC’d? Is the beefier CPU cooler sufficient or should I go with an after market cooler?
Thanks!
You know, even when
You know, even when overclocked the temps never really got above 50C. I use an older, but still really beefy, 80 mm unit from Thermalright. XP-80 I think? Regular temps were high 30s to mid 40s otherwise.
I would imagine the regular CPU cooler they include will be sufficient. If you really want to push the raggedy edge, then obviously something more expensive is going to be needed.
I got to see an engineering
I got to see an engineering sample of Trinity APU. The one I saw will represent the desktop A10-5800K model.
Compared to the A8-3870K, the Piledriver-based Trinity needs to be 800Mhz faster in order to just edge out against the Llano based APU.
Clock-for-clock, the K10.5-based Llano APU is better.
It appears AMD didn’t really address the fundamental issues with the Bulldozer design in Piledriver. What they did was implement “clock mesh” in order to reduce the power footprint, so that they can aggressively clock up the speeds. (To compensate for the poor clock-for-clock performance!) …AMD’s Bulldozer is analogus to Intel’s Pentium 4 in this regard.
On the iGPU side…
Trinity’s IGP performance = Llano’s IGP + discrete GPU in hybrid Crossfire.
The IGP side is much better. Its a pity the CPU side is still underwhelming.
I don’t expect AMD to fix things until at least 2014 or 2015. (More likely the latter.)
I’m going to invest in Intel’s Ivy Bridge (haven’t upgraded since 2005!), and slap on a mid-range Geforce or Radeon GPU. The Intel IGP will be used if the video card dies! 🙂
Forgot to add…
I checked
Forgot to add…
I checked AMD’s financials. The only thing keeping this company alive is the success of their APU lines. OEMs can’t get enough of them!
Bulldozer (FX or Opteron versions) didn’t make any serious money for AMD.
Well, at least in the case of
Well, at least in the case of Trinity vs. Llano, IPC is not the primary goal I think. The push is obviously better performance per watt. So, if they have improved Piledriver to a decent degree over Bulldozer, and have been able to decrease TDP per clock, then it could still end up being decent.
I saw some benchmarks and it looked like the Trinity product clocked at 3.8 GHz (dual module) was around 15% faster than the 2.9 GHz Llano. So, a 45% increase in clockspeed nets 15% performance increase. At first glance that does seem unappealing. Then again, when we see that 15% base CPU performance increase will likely come with a 35% to 50% increase in graphics performance, then things start looking a little better. Plus all of this is within the same 95w to 100w TDP.
If those numbers hold up, then it will be a nice improvement for AMD on both the CPU and GPU side of the APU as a whole. Sure, we could wish for a greater improvement in IPC then what we are getting, but the design allows for TDPs down to 17w. That is going to be very good for notebooks and ultra-thins.
Very good article Josh and it
Very good article Josh and it is indeed dissapointing that the FX series of CPU’s cannot match or surpass the Phenom II’s in your testing. There does however appear to be plenty of headroom for overclocking which can’t all be bad. Vishera series not due until Q3 2012.
Thanks for reading. This was
Thanks for reading. This was one of those articles that seemed to flow out pretty easily. The results were fascinating for me when I was running all of the tests. I finally was able to grasp where some of the bottlenecks were for Bulldozer and what some of its issues are.
The CPU can clock quite nicely. It was dead simple for me to go to 4.5 GHz with a small voltage increase. I do have good cooling on it, and I probably should jump it up a few notches to see where it peters out. Those are some seriously leaky transistors though.
I think we need to step back
I think we need to step back and really look at Bulldozer. Intel made a real mess with AVX and FMAC jumping back and forth from three to four operands. Also, I just bought an Opteron 6234 (12 core 2.4GHz) and I had to turn off AVX because it isn’t supported by HyperV. If it had come out sooner it would be even worse.
But when you use AVX or XOP, you get MUCH better numbers so it just looks like Intel played fast and loose so AMD would have an incomplete ISA for Win7. It may even take until PD and Win8 to get “real” numbers where the OS is optimized for the CPU.
Bulldozer has things like core parking for Win8 and Win7 doesn’t understand the module because it’s NOT HyperThreading.
I’m really PO’d that no one is using FMAC yet because Intel changed it several times. It’s a shame when a company acts like they do.
Obviously I will be very
Obviously I will be very interested to see the performance of Bulldozer on a fully optimized software platform, but I still think that it has too many bottlenecks in the design overall to really be that good of a performer in most apps. Sure, AVX and FMAC support would show a nice jump, but how nice as compared to Ivy Bridge running the same software? I honestly do not have an answer for you there.
I just purchased this
I just purchased this processor and it really is a good processor currently OCed at 4.18 ghz i have this dumb acrylc clear case i thought would be cool just a pain but regardless to air flow is good but with the stock CPU heatsink trying to OC anymore was giving me heat problems so went out yesterday got this nice Heatsink installed everything running cool when i get a chance tonight i will see what this bad boy can do my goal is 4.5-4.7ghz maybe MAYBE try 5.0ghz see if its stable. For the price this Processor is amazing i spent 160$ on it and i can promise you it would keep up with the intel i5 2500k thats prices around 220 for the price its a beast. Runs everything perfect games are clean New betas running amazing Counter-Strike:GO runs perfect same with Diablo 3 BF3 and while i play those games im streaming live, downloading torrents, encoding blurays, web surfing. doing it all and this processor still has alot to give. This is my first AMD processor ever always been an Intel guy nvidia GPU and Intel CPU were always my setups. Saw this deal and couldnt look away and im happy I didnt. Benchmarks yes they do show performace this and that but get youre hands on it and really test it out and you will fall inlove
This is a good cheap
This is a good cheap alternative now against a i5.
Get a 6100 for 120$ 7.3 WEI, compare at 7.5 I5 2500k.
Over clock @ 4.6Ghz on air and you have an i5 performance for cheap. Spend the difference on a better gfx card and stronger case.
I have a 4170 with a 7970x
I have a 4170 with a 7970x sli and have had absolutely no issues with gaming at ultra with dual 52” led tvs. I run @5.2GHZ no issues. H80 Cooler (Runs ok at 6ghz passmark score beats all ivy bridge i5s and i7 2700k at 6ghz)
WEI 7.9 cpu, 7.9 Ram OC 2133, GFX 7.9, HDD 5.9
The high GHZ rating will allow better performance than I7 stock in games. This cpu is easier to overclock and somewhere to start for beginners. However if you spend the money to buy an i5 then a good heat stink odviously that would be better. (just a waste of $)
Of course some hopped to see
Of course some hopped to see a much better CPU from AMD, but also some users didn’t care very much. Improvements came at last.
I think this is a good step from FX-6100. It achieves higher frequencies but the power consumption at stock speed remains in some limits.
I look behind and I see that Phenom was distilled for about 3 years until the chip Phenom II became very competitive.
I like electronics, and if they pay me, I would like to help AMD with this issue and innovative ideas.