Very Fast, very STABLE! This thing is so much more stable than my i7 3770k, without losing the performance that everyone claims. For heavily multi-threaded programs/tasks and heavy multi-taskers, this is your new baby. The performance and power draw is stable throughout. I love it.
Even when running heavily multi-threaded and multi-tasking, this processor draws power evenly throughout. More than that, it stays stable and competent when switching between programs and applications; whereas the Intel does not. Well worth the purchase.
Windows 7 CPU rating is 7.8 as opposed to my 3770k (which gets a "measly" 7.7). Both processors score an equal 7.9 in Windows 8, if anyone is curious.
Performance per dollar (list price is currently $190) is significantly higher than anything the competition offers.
Raw data throughput is nearly 3X higher than the 3770k. This thing churns through data tables like there's no tomorrow.
Since I rate stability as #1 on my checklist for must-have's in electronics, this is, by far, the best processor I've owned since Thuban (and I've had all of Intel's offerings, before and since).
Due to the North Bridge still being a separate chip on the motherboard, the processor gets bottlenecked where it realistically shouldn't. When one stops to consider how much effort has been put into AMD's "APU" line and including a NB circuit on those CPU's, it really doesn't make much sense to not do it with their upper-level versions.
125W. Still. Granted, the 32nm process is significantly better than the 45nm predecessors, but Intel has already advanced to 22nm and 77W. Though, hilariously, it should be noted that this processor runs much cooler than my i7 3770k on both idle and load, and both processors are using the same exact heatsinks (ZALMAN CNPS9900ALED 120mm) and the Intel is in a case that provides better cooling.
Second Generation FX CPU based on the "Bulldozer" architecture. While there have been some improvements made, it's still not very good at single-threaded processes, which allows the competition to lead in this area. Not good for bad console ports; expect poor performance here.
Still stopped at 32GB of total system RAM. Really? (Insert expletive here)
Has a DDR3 1866MHz controller... if you don't mind Single Channel RAM. This would seem to be an egregious oversight. Almost criminal, really.
No PCI-e 3.0 support. Not that it really matters; the PCI-e 3.0 hasn't been fully adopted yet, but it does leave this processor behind the times by more than it should be.
Still no UEFI. Instead, you're still stuck with a 16-bit BIOS. D*** it AMD; get with it! This is another bottleneck, however slight, that this processor simply doesn't need.
While this CPU is a step in the right direction, AMD needs to do the following to surpass the competition:
Stronger Integer pipelines; this is hurting the performance.
Faster caches; about 4X faster than it currently is should do the trick.
256-bit AVX Extensions instead of the 128-bit that it currently posseses. Yes, the two 128's can be "unified" somewhat, but what this really means is that the 256-bit instructions are divided into 2 separate parts, processed, and then put back together. It should be obvious why this is bad.
Full adoption of UEFI. Reduce those bottlenecks, dammit. Why stick with a dated 16-bit firmware interface on a 64-bit platform when inexpensive alternatives exist? Laziness, maybe? I don't know. But the excuses are wearing a bit thin.
Still trails behind the competition (not by much, granted, but still) due to the NB circuit being a separate chip instead of being on-die, which bottlenecks the system. This is something that should be addressed by "Steamroller". This may be caused by the insane idea AMD has about keeping customers by making every darn chip "backwards compatible" with previously existing mobos. Bad AMD; no treats!
Still on 32nm. STILL. If Global Foundries can't keep up with the times and the pace set by the competition, then perhaps our good friends at AMD should find a new foundry to partner with and invest in. I hear Samsung is working on a 17nm micro-process - and has nearly perfected it; oh, wait, I know they are; I work for them. Perhaps a partnership is in order? <- This, I would love to see; the stench of Intel's impending doom would be thick in the air.
Marketing. The real reason Intel is surpassing AMD isn't performance (the difference isn't anywhere near as extreme as the corporate shills make it out to be), it's marketing. Every time I turn on the dang TV, I hear that stupid, annoying 5 tone theme of theirs. I guess the reason I find it annoying is this: I don't need some marketing shmuck telling me the product they're shilling for is the best because... well, because they say it is. But, apparently, the average joe will believe any garbage they see on TV. This being the case, it's long past time AMD let the average joe know: "Our stuff is good too... and at a fairer price!"
It's in single threaded applications (or very poorly multi-threaded ones) wherein Intel is the clear winner. It makes sense in the here-and-now to design a processor for single-threading, and it's exactly this that has left AMD struggling; they've looked so far forward, that the FX processors are ahead of their time in a very real way. Most programs these days lightly multi-thread if they do it at all. Even those that do heavily multi-thread are usually poorly coded. In the very few programs/applications that are programmed to multi-thread efficiently, this processor is a monster whose raw power and performance are unmatched by anything the competition offers short of $1K processors.
Currently, I'm not anticipating AMD's next release. If they were to read this review and incorporate my suggestions as changes for "Steamroller" however, I'd be positively drooling for the next offering. As it is, I'm merely cautiously hopeful.
Depending on the application, this sucker is quick. But, one of the few places where it really shines is crappy console ports. And, as any fan of the recent Transformers game knows, all some "developers" are "good" at is making bad games and then porting them over to great hardware, thus making them worse. This processor makes up for some of that garbage code/support. Some, but not all.
The place where this processor truly shines, however, is 3D rendering software. In this regard, it's without peer in its price range. Dont' get me wrong, the FX 8350 is certainly no slouch, but that extra 12% - 27% less render time per frame can make all of the difference while you're doing the render walk. Also, at the time of writing this, the FX 8350 lists for $190 as opposed to the $330 price tag this bad boy carries; clearly the 8350 is priced far more competitively, given it's performance. Which, of course, means that the only thing the competition (AMD) offers in this price range is server CPU's - which DO NOT count, since we're talking desktops here. Meaning that it's only really competing with it's older brother, "Sandy Bridge" 2600k and - ha ha - it beats that CPU quite soundly.
256-bit AVX extensions. Now, I'll grant you, there are very few video/file encoders out there than can really make use of this, but the ones there are give this processor an incredible edge in encoding performance. Again, you have to have proper programming to make this work worth a damn though. Anybody got a few thousand dollars to spare for the proprietary programs that can do it? No? Me neither.
Some of the North Bridge functions have been shifted to the CPU die, reducing the standard PC bottleneck. YAY! ...sort of. (I'll explain later).
Any decent programmer can tell you that Simultaneous Multi-threading (what Intel refers to as "Hyperthreading") is a risky proposal at best. Serious Engineers can tell you that utilizing one worker to do two tasks simultaneously sounds great, but in reality presents a whole new range of problems. In processors, this means data intersection. In the original i7 series and the "Sandy Bridge" processors that followed, data intersection was a huge problem. It got better in "Sandy Bridge", certainly less intrusive and detrimental than it's predecessor. In "Ivy Bridge", this has been mitigated even further, but still remains an issue. Data Intersection, in layman's terms, results in data loss and/or corruption. Usually corruption is the big issue, resulting in loss for both total data processed and in time due to the potential of the data needing to be reprocessed. This, of course, leads to the next downside...
... This is "Ivy Bridge" 's biggest issue: Stability. Part of this - I'll say 50/50 - is due to the inherent issues with "Hyperthreading". But, as I've stated before, stability of a platform is #1 in my book. With absolutely no excuses regarding "performance". While I can't speak for anyone else, I'm not willing to sacrifice overall system stability for meager performance in programs I don't regularly use; at least not for my workstation or daily use desktop.
No apparent North Bridge circuit on the CPU die. While I can't be 100% certain, not without having actual blueprints of the circuits, that this is the case; it certainly appears to be so. What is clearly visible are "extra" clock/cache schedulers, which appear to have been re-purposed to use as an apparent NB. This would seem to be a good idea, in theory, since all a NB does is schedule cycles and assign tasks. But what it really results in, especially in laptops which have the bad habit of running H-series chipsets (which use, by design, the Intel HD series integrated GPU's), is the Intel chip defaulting to its onboard GPU; which is so severely underpowered that it causes complete system instability. BAD INTEL! No Treats!
Speaking of the GPU, the "HD 4000" on this thing is f**king atrocious. It's horrifying, really. When I look at the performances measurable by on-die GPU's, AMD isn't just in the lead; it's stomping the competition so completely that it astounds me people even buy anything Intel makes. Most users I run across need faster/stronger GPU's than they do processors... especially in mobile and ultra-mobile platforms. And the onboard graphics processing unit for this thing is dismal. Compared to the 7000 series already onboard the new Trinity GPU's (which aren't really 7000 series, but rather a more powerful version of the 6000 series cores), the HD 4000 is garbage. It simply doesn't live up to the standard that AMD set. It's laughable, really. Usually Intel sets the trend, but just like 64-bit processing, AMD is clearly in the lead here. It gives me the giggles.
Price. Good God, is this thing expensive! A chip cost of $330 doesn't seem like much, but considering the mobos for this bad boy are equally expensive (considering what the competition offers) that it's hard to justify the added expense.
If you want a separate rendering machine so that your daily use desktop/workstation isn't being hogged up by the rendering process consuming every last resource it can (Mental Ray is especially greedy), this is definitely your new render machine. As I've stated before, the extra 18% - 27% over what AMD's offerings are capable of can really reduce the overall render times.
H.264 Encoding on this thing is fast, true, but I don't trust it. I don't trust it because I've had too many projects that have had to be re-rendered due to the aforementioned "Cons" this processor carries. Render out dozens of videos or 3D projects per month and you'll soon be repeating some stuff too. Most of the time, the additional speed can make up for this, but I question the validity of such an argument if I have to keep re-rendering projects due to quality loss per-frame. Thankfully, this is approximately 0.001% which means it's relatively rare. "Rare" of course, being a completely relative term. Since the FX "Vishera" processors do not experience this at all, even "rarely" may be enough that many find the trade-off to be unacceptable. Since I now intend to use my i7 solely for rendering purposes, it means a slight annoyance for me. For others who either can't afford to - or have no desire to - own several purposed computers (or "Servers"; which is what they are, really) "rarely" may prove to be far too often.
Overall, this is the best Intel chip I've owned in years - stability issues aside - but it's still not as exciting to own or use as one might think. The GPU is complete garbage compared to what AMD's APU's offer, in terms of performance and, as a whole it's just... well, mediocre. It just doesn't live up to the hype. Not that any Intel processor ever really has. At least not since the Pentium 4. With stability issues resulting from a combination of data intersection due to hyperthreading and what appears to be an improper NB circuit, I found it hard to stay excited or even mildly satisfied with my ownership of this processor. In point of fact: I removed this thing from daily usage just as soon as I could afford to. Because of the unwieldy expense of purchasing this hunk o' junk, I had to wait an incredible length of time to build back up enough funds to buy an AMD rig - which, thankfully, has been proven to be infinitely more stable - and thus my Intel has been relegated to an unceremonious and anonymous position on my computer parts rack. Shelved until I can afford the various items I need to turn it into a render server. It may make brief forays in public as a LAN box, but will do so without bearing any Intel stickers. It's been my experience that anytime an Intel-fanboy witnesses system instability in an Intel platform, they automatically assume that you've done something wrong and begin to try to convince you to allow them to fiddle around with your system to "improve" it. Only by getting on their systems and then reproducing the error, then showing them an endless stream of white papers and allowing them the time to reproduce this for themselves do they ever really understand that this is just how "Hyperthreaded" processors work. You would think that this would lead to a conversion on their part, but more often than not it ends with the other person resenting you instead. All in all? Not worth the effort. Maybe I'm just not enough of an AMD fanboy to feel the need to run around convincing folks to switch. It's funnier to listen to them bitch endlessly about the things I've known about since Intel launched the i7 920.
Haswell is supposed to be better than this current gen, so we'll wait and see what Intel does with it. Just like my cautious hopefulness for AMD's next chip, I'm cautiously hopeful for Haswell too. Though, my hopes for Haswell are significantly different. My hopes with Haswell is that it will actually, finally, either live up to the hype - thus warranting the extraneous cost - or be more competitively priced. I'm sick of just a CPU and mobo costing $500+. That's just not acceptable; not when that much will buy nearly an entire AMD system.