AMD FX-8150 Review: From Bulldozer To Zambezi To FX

[cangelini]

[citation][nom]gerhardb[/nom]I would really like to know why an Intel 980x or 990x are not in this list?[/citation]

Would have loved to included all processors. However, AMD gave me about a week and a half with this chip. Picked the five most interesting competitors (in my mind), ran all six through completely fresh Win7 installs using updated versions of the apps, and then wrote 12,000 words. Time simply didn't allow for the $1,000 parts that aren't in the same price league anyway.

Best,
Chris

 

[chumly]

TL;DR, Sandy Bridge is better. Waaaa waaaa waaaaaaaaaaa.

 

[hyteck9]

wait wait wait... Seriously... this can't be right. I have both AMD and intel CPU's in my boxes so no bias here but let's just think about this a moment. If you read between the lines this article is basically saying AMD took a new design, developed it, polished all the manufacturing processes to be able to mass-produce it, brought it to market thinking it was the new Value/Efficiency hotness and NEVER did any comparison tests along the way? I'm going to guess AMD is not so stupid and this early test is stumbling on a BIOS issue. if AMD actually "IS" that stupid... they deserve to go bankrupt by close of business.

 

[Guest]

@halcyon and srgess

and yet for such a spectacular failure it managed to absolutely wipe the floor with the i2500 and even the i2600 in some bench, makes for an interesting point and i'll admit it loses out to more benches then it wins but then that's an awful funny chip, it's either got some interesting potential or just plain schizo

and if you hadn't noticed there's an awful lot of AMD fans stating their disappointment but i guess you missed all 11 pages of it

 

[_Pez_]

[citation][nom]torque79[/nom]I am so pissed off I waited for this load of crap before buying a new pc. What a HUGE disappointment. It's unfortunate that gaming companies are very slow developing heavily multi-threaded software, so AMD's approach of more cores just does not work. I would love to see more effort from software companies to have open-ended multicore processing rather than limited to only using one or two extra cores.[/citation]
same here pissed off and dissappointment

 

[busuan]

IMO, AMD should take Apple's approach: alternating optimization of hardware AND software.
This excellent review pointed out FX's great weakness, its lack of specific optimization from software developers. While AMD is trying to surpass Intel in CPU design, its suffering and embarrassment will continue because few software in real world would show FX's supposedly superior architecture. Hadn't they learned anything from the slow death of the beautiful Alpha21x64, as well as R10k???
Now, take a look at Apple's approach with A4/5 and iOS. While iOS is always nifty, A4/5 is hardly the best among its peers. But Apple's combination of the two is purely magical; they optimized their chip to precisely match iOS, and they went back to change iOS if their chip evolved a few more new tricks.
CPU and OS are inseparable. AMD must realize they cannot beat Intel in a Wintel game; Intel and Microsoft set the rules, and they had a head start. But AMD could try changing the rules. Go to Microsoft and all other developers. Ask specifically what they want, then revise your design. Show them your superior CPU moves, beg (or bribe) them to put optimized code in their products. As long as a few popularly-adopted software products, such as Excel, Photoshop, Final Cut, AutoCAD, 3dmax and iTune(!) could show clear benefits from AMD chips, AMD may not win by this approach, but at least, will not lose.
Because I don't want to see lousy CPUs from Intel either.

 

[Guest]

i always envisage (maybe incorrectly) that the XOP instruction set would outsource it's number crunching to the GPU side of the APU, was the XOP test done on a machine with the IGP as the primary display device?

I thought the idea behind AMD's next gen chips were that there was no way in hell they going to beat intel core for core performance, intel is way ahead in that game, i thought the idea was to leverage GPGPU to boost the CPU performance, in a number crunching game a GPGPU should have an advantage over the CPU

 

[danwat1234]

I hope AMD Trinity + Piledriver will bring a great improvement in the mobile sector! Piledriver is Bulldozer version "1.1"

 

[TeraMedia]

@m8x776, I think you're onto something about the pipeline length.

Unlike many of you, I'm not disappointed. I'm frustrated - for AMD. Because I too want competition. I have CPUs from both in various computers, so I'm not a brand loyalist.

The fact that AMD's new chip smokes the 2600K in a couple of tests means that parts of the architecture have the potential to scream. Simply put, if it didn't have the potential to be faster, it never would be. But parts of this chip clearly aren't up to par. And given the low IPC performance, it appears as if something related to prefetch, branch prediction, cache latency or similar isn't working properly. The problem doesn't show up in bandwidth tests (in terms of instruction throughput, not memory), but test for latency (e.g. conditional branches) and performance tanks. Perhaps that part of the architecture can be changed in a future update? Otherwise, AMD needs to start over again, and they might not survive that effort this time.

And the CEO doesn't need to be "shot" over this - Dirk already was.

 

[dante01010]

I think the applications needs to be rewritten before make the benchmarks, and some tests with Linux and BSD would be very welcome

 

[simpleshadows]

Being a longtime AMD/ATI diehard myself, I understand the inbred desire to want to see greater performance figures when going up against the incumbent Intel. That said, it is quite evident that the majority of people commenting here are devilishly impatient and cannot take the same long view that AMD is themselves taking.
Is your overclocked PII really losing you that many FPS in the games you play? Would those of you with decent systems not gaming at 2560x1600 even notice the difference?
Software, as the article states, is the primary driver of scalability here. While I agree that AMD certainly has some work to do (and why did they use the FX moniker?), let us not put the blame squarely on their shoulders. Instead let us ask ourselves why mainstream applications are not being coded to keep up with the multi-core architectures that dominate today's landscape.
Sure, there is no question of Intel's supremacy in the upper echelons of the price tier. 2600k? upcoming 3960X? Yup. For the rest of you interested in neither the braggadocio of having more FPS than your eyes can measure nor wanting to sell a kidney to get great performance in a balanced system, stick to your guns.

 

[amk-aka-Phantom]

[citation][nom]nerrawg[/nom]AMD and Intel, along with mainstream PC builders would be fools not take us seriously. Sure, sales to enthusiasts are low volume total, but critically:Who do family and friends go to when they are going to buy a new PC, laptop or phone: US! Enthusiast influence isn't about their purchase power at all, its about are market influence. There is not a family member or friend who buys a PC without asking us for advice first, because we are free (and for the most part unbiased). Also most normal members of the public don't understand all the techno jargon, but have been done over by PC world and the likes so many times that they know they need to come to us first before they buy. Bottomline: We enthusiasts might only buy a products each, but each of us influences the purchases of 10s-100s of consumers. Only the most foolish of marketing divisions would ignore this.[/citation]

THIS! + over 9000!!! Blast the people who don't understand that! The only profit I gave to Intel was my i7-2600, but in the last two months I've configured about 5 Intel builds and 10 more are waiting in the next two weeks (new library). So enthusiasts rule the high-end purchases.

I hope Bulldozer finds its place on the server, but I'm keeping my i7-2600 (not like I was gonna change it anyhow, lol, but now I've at least preserved the bragging rights)

 

[amk-aka-Phantom]

[citation][nom]simpleshadows[/nom]Is your overclocked PII really losing you that many FPS in the games you play?[/citation]

LOLWUT? Pentium 2 not losing us many frames per sec-

Oh. You meant Phenom II. :lol: Use X4/X6 after PII next time.

 

[dragonsqrrl]

[citation][nom]larkspur[/nom]Intel no longer has anyone pushing them to make more powerful chips. Now they can slow down and just milk what they have. They could delay Ivy and just rebadge Sandy Bridge for 3 more years and AMD won't have caught up. Intel's chips are going to get more expensive methinks. Arrrrrrrrrghh! What happened to you AMD?[/citation]
Intel hasn't had anyone pushing them to do so for the past 5 years, and yet significant performance and efficiency improvements still come every generation, they're still ticking and tocking annually, and they're still using the same price points they've used in they're consumer lineup since Core2.

 

[BlackHawk91]

That was a great article Chris. I always felt that an "octo-core" at that price tag would be "disappointing" somehow, but not that disappointing. Now I just lost two months of gaming because of waiting for bulldozer. I should have bought that 2500k since my system died.

 

[BaronMatrix]

I have to go with AMDZone's Abinstein on this. AMD needs to work harder to optimize SW for its arch. I looked at multiple cases where the test were run over and over with no improvements, which means the SW is not optimizing using LWP.
It also looks like the way the INT scheduler works its in some cases causing increased latency. This could be due to thrashing as BD's L2 is now exclusive so if we're having to go back to L2 rather than using prefetched instrs, there will be higher latency.

AMDs XOP and AVX implementations are top notch so we need the SW to catch up. If we take 10% for SW optimizations, 10% for Win8 optimizations and 10% for Piledriver, the BD in Q3 of next year will be 30% faster than this one.

 

[halcyon]

This must be the month for big disappointing news. First the iPhone, now this.

 

[amk-aka-Phantom]

Hey, AMD fanboys who flame Intel over changing sockets frequently... isn't i7-920 still doing good compared to your new shiny BD?

Hmm, maybe AMD needs to stop supporting a 3-year-old socket and just make a new one - that ought to fix their lagging behind.

[citation][nom]dragonsqrrl[/nom]Intel hasn't had anyone pushing them to do so for the past 5 years, and yet significant performance and efficiency improvements still come every generation, they're still ticking and tocking annually, and they're still using the same price points they've used in they're consumer lineup since Core2.[/citation]

Indeed. It's quite funny, in my country prices don't change for like 1.5-3 years after the hardware gets released... so Core 2 Duo's price is same as Core i3. The choice is obvious, though

[citation][nom]dante01010[/nom]I think the applications needs to be rewritten before make the benchmarks, and some tests with Linux and BSD would be very welcome[/citation]

Yeah, rewrite the applications... in other words, we're AGAIN doing it all wrong, right? How about releasing a product that can actually kick ass with CURRENT software? Sandy Bridge and Bloomfield can!

Linux benchmarks? True-blue Linux fanboy uses Atoms and C2Ds/Celerons and whines about how other CPUs are a waste of electricity; adn when his distro doesn't work on a current-gen hardware, it's branded "fancy" and dismissed. Why do you need benchmarks on Linux?

 

[amk-aka-Phantom]

[citation][nom]halcyon[/nom]This must be the month for big disappointing news. First the iPhone, now this.[/citation]

You forgot multiplayer in ME3 and iCloud replacing iTunes (just now). Well, at least Jobs...

*gets mauled by angry Apple fanboys*

 

[kaitheus]

Why don't they setup 2 different PCs (intel/Nvidia) and a (AMD) setup to thoroughly test the FX chips as we all know several of the tests and benchmarks are either Intel or Nvidia biased, so why not ? Frankly I'm ganna wait till the FX8170 releases to really care about benchmarking, till then my X4 955BE at 3.8GHz does just fine with everything I throw at it.

 

[that_aznpride101]

[citation][nom]amk-aka-phantom[/nom]You forgot multiplayer in ME3 and iCloud replacing iTunes (just now). Well, at least Jobs...*gets mauled by angry Apple fanboys*[/citation]

You're a scumbag for saying that it's good news that Steve Jobs died from pancreatic cancer. I really don't like Apple products either, but even I wouldn't go so low as you just did.

 

[amk-aka-Phantom]

You're a scumbag for saying that it's good news that Steve Jobs died from pancreatic cancer. I really don't like Apple products either, but even I wouldn't go so low as you just did.

... all I meant to say is that Jobs movie will be released soon, as the news say. That's good, isn't it?

 

[clonazepam]

First, I want to say I should have gone straight to Tom's site this morning. Instead, I opened gmail, and had a tigerdirect email advertising the new BD line-up, their prices, and deals. As soon as I saw the price, I knew already what to expect in the benches. Was a buzz-kill. I would have liked to have discovered the BD article first and at least had that initial excitement for a little while.

[citation][nom]maddy143ded[/nom]TOM's please give us bench for the FX4100 and FX6100 series too.......8150 is way too disappointing . I am thankfull that I waited to upgrade my AM2+ mobo. now I will use until IVY Bridge is here.....may have to look into the FX4100 for a HTPC though.I predict better performance then Core i3 SB. series.If AMD is going to give such disappointing stuff after having us wait for nearly 3 years, then Its begging us to go over to INTEL[/citation]

I'm excited to see benchmark comparisons for the FX4100 the most. I had this idea in my head that I wanted to take the least expensive BD part for a new HTPC build. The problem I now face is that I am currently using a Q6600/GTX 570 as a HTPC but I'd like to sport some newer better cpu+motherboard to make better use of the SSD, and extra all-around storage performance. Please include stock speed comparisons of any of the new BD pieces compared to the C2D or C2Q generation. I know the Q6600 is still out in the wild and still quite popular.

[citation][nom]cangelini[/nom]Soon as we have access to the processors, you can count on it![/citation]

Please get a hold of that other team that did the generational clock for clock, core for core comparison testing and have them bench the leaders from each generation. I really would love to see the BD vs socket 775 products.

[citation][nom]Seabound07[/nom]Lets see how this plays out, to be honest, i should've waited for this, and PCI-E 3.0, but core i7-2600k will be alright til next year, i hope. im glad AMD got this out before the end of the year, any later and they would've had to release it when Southern islands is scheduled to come out... that'd be a good combo CPU+GPU for xmas.[/citation]

This chip doesn't appear to support pcie-3.0 unless I read something wrong. The i7-2600k will be plenty for a few years depending on what you are asking of it.

[citation][nom]mapesdhs[/nom]Including the 920 in the review provides an interesting baseline. AMD isfortunate CA didn't use an 870 instead as that would make the 8150 lookeven worse for those tests where the 920 is ahead (re the 870's muchhigher Turbo levels). Must admit, I had been expecting 8-core BD to be amatch for 6-core SB; I was not expecting it to end up being pegged as a'match' for just the 4-core SB. That's a real shame. My main app isHandbrake; I though the 8150 would be well ahead of the 2600K, so thisis a big letdown.homeboy2 is spot on to say this is bad news for Intel fans aswell. Themarket needs competition; BD isn't good enough to make Intel rethink itspricing, except perhaps at the low end when the 4100 is reviewed. Withouta doubt, SB-E is going to remain being priced at a premium level.Chris, speaking of the 920, why is it so fast for the 7-zip test? Thegraph shows it significantly outperforming the 2500K (51s vs. 62s). Oris it the 920's HT which somehow gives it such a large gain over the2500K? Though I would have thought the 920's HT effect would be swampedby the 2500K's much higher clock, etc. Strange. And has there been anyresponse from AMD to your review? Just curious.Thanks for the article! 8)Ian.[/citation]

I really expected to see a decent 'win' for BD with Handbrake. I had this theory AMD's multi-core technology would see the most benefits in a program like Handbrake, and other software that's freeware/shareware/open-source/GPL, and that would have been possible if not for BD's shiny new arch. I think we will see measurable improvements as revisions, bios updates, software patches/upgrades come down the pike.

[citation][nom]CaedenV[/nom]Intel has always been much larger than AMD, but that did not stop AMD from killing intel through the P4/PD years. They are simply grasping at straws, and not able to innovate as well as intel lately on performance (and surprisingly loosing in the TDP area as well in spite of all their low power tech in their awesome mobile lineup). That is even more sad for AMD. Much faster ram and still not able to keep up...Must have killed you reviewing something new that barely keeps up with it's own predecessor, but good job all the same.[/citation]

I think Intel shot itself in the foot in the P4/PD years and AMD happened to be in a great position to capitalize. I used to be in retail, and back in those days, I recommended AMD-based systems all day long. I have a laptop from 2005 with the AMD Athlon 64 4000+, 2GB RAM, and for what I need from a laptop, I really can't complain about its performance, though, clearly the battery life is lacking considerably, but that was a different time.

I still would like to invest in a BD-based HTPC paired with a 570 for some gaming, but I really need to see how that 4100 stacks up. I definitely would love a comparison between it and the Q6600 at relatively same speeds. Regardless of that performance comparison, the overall gains from an aging nVidia 750i chipset for a new 990FX (or i3+p67) could make it worth it. Not in a hurry at all, so I can wait for those to come out, for BD to mature, and all related softwares before I finalize a decision.

Thanks CA!

 

[_Pez_]

LONG LONG TIME WAITING MANY MANY DELAYS FOR AN EPIC FAIL !!!

 

[dragonsqrrl]

Bulldozer in a desktop environment is a mixed bag at best. At best, it's able to compete with Intel's Sandy Bridge architecture in certain thread optimized content creation and productivity applications. This in itself is an impressive accomplishment, but it's only able to achieve this at what expense? Abysmal single and lightly threaded performance, higher idle and load power consumption then the competition, twice the transistor count of a quad-core Sandy Bridge (and remember SB also includes an integrated GPU), increased stock clocks (attempt to compensate for decreased IPC) coupled with negligible improvements in overclock-ability ...

I use productivity and content creation software on a daily basis, and in this market Bulldozer certainly could be a viable alternative to Sandy Bridge, but far too much was sacrificed simply to achieve performance parity with Sandy Bridge in certain workstation applications. I think most people could agree that performing worse then your predecessor in any area, much less a wide variety of common applications is simply unacceptable.

Unfortunately, a workstation environment is the only area where the FX-8150's $245 price is justifiable. But as a general use desktop processor, running both gaming and workstation applications (like I do), it should at most be priced the same as the i5-2500K.