Phenom II 965 seems faster in some games than I7 920

[Nater]

Why would intel want AMD at 1-2% market share? They wouldn't. AMD allows intel to point to regulators and say 'we have competition'. That doesn't work so well when you're competitor has 1% marketshare, but it does when they're at 15-20%. The actual percentage is less important than the market segmentation. AMD is only competitive with intel in extremely low margin, low cost, but high volume processors and in high margin, very low volume server chips. Intel dominates in the segments that make money. Desktop chips above $100, 1S and 2S workstations and severs. Intel makes a lot more money on processors sold than does AMD.

It's patently obvious from intel's current behavior that they; have AMD about where they want them and they don't want to push them any further due to government regulators. Intel could cut the prices for Clarksdale processors in half and take AMD out of the cheap desktop chip game over night if they wanted to. They could go back to a 2-year tick-tock cycle if they wanted to. That would allow them to be pushing out the 22nm Ivy Bridge chips a quarter or two after AMD launches 32nm Bulldozer, an architecture that will probably draw even with Nehalem and Westmere. But they won't because it would draw heat from the EU, Japanese, and the American trade regulation bodies.

The success of the K8 has a lot to do with AMD being smart, but it has even more to do with intel being stupid. Intel wanted to segment the market between the Pentium IV and the Itanium. It wanted desktops running the P4 and it wanted every server to be a mid-range or high-end box running Itaniums. They both had the same problems. The Pentium IV, while still x86 required a recompile to see the best performance. It also consumed way too much power. The Itanium and it's completely new IA-64 ISA didn't work well with the backend applications companies ran without the software vendors rewriting their code. Intel overstretched and AMD took advantage of it.

It wasn't AMD's success that allowed them to produce the best performing chips on the market, it was intel's failure. Intel actually had an architecture just as fast as the K8 in the Banias/Dothan but only used them for laptops. If they'd designed desktop/low-end server chips around that micro-architecture in 2003/4 instead of 2006 AMD wouldn't have gotten the Opteron off the ground.

Do you really think that intel will be that shortsighted again? I don't. I don't think that they particularly liked not having the fastest chips running an ISA that they invented. I don't think that they much cared for what it did to their business.

Yes, AMD spends a lot on R&D. Do they get much out of it? Not really. The only world-class semiconductor products that AMD makes are from the ATi division. While intel spends a lot less as a function of their total income, they certainly get more out of it. They have the best x86 chips, they have the best SSD controller on the market, they're the only company that seems to be making any progress on PCM (NAND's replacement), they have one of if not the best RAID ASICs.

Finally, semantics aside, intel and AMD do not trade the performance crown. People seem to think that AMD and intel trade off having the faster chip every couple of years. This is not true. 90% of the time, intel has the faster chips. They have more patents, they have more engineers, they have better fabs, they have more money. It's just that the x86 world order has gone back to it's normal state with intel a generation and node ahead and AMD selling cheap chips. Any lead AMD has in any particular market segment is because either intel is EOLing one product line and getting ready to introduce another (4S/8S) or it's a market that intel sees little value proposition in playing ($50 desktop chips).

[anubis44]

Why would intel want AMD at 1-2% market share? They wouldn't. AMD allows intel to point to regulators and say 'we have competition'. That doesn't work so well when you're competitor has 1% marketshare, but it does when they're at 15-20%. The actual percentage is less important than the market segmentation. AMD is only competitive with intel in extremely low margin, low cost, but high volume processors and in high margin, very low volume server chips. Intel dominates in the segments that make money. Desktop chips above $100, 1S and 2S workstations and severs. Intel makes a lot more money on processors sold than does AMD.

It's patently obvious from intel's current behavior that they; have AMD about where they want them and they don't want to push them any further due to government regulators. Intel could cut the prices for Clarksdale processors in half and take AMD out of the cheap desktop chip game over night if they wanted to. They could go back to a 2-year tick-tock cycle if they wanted to. That would allow them to be pushing out the 22nm Ivy Bridge chips a quarter or two after AMD launches 32nm Bulldozer, an architecture that will probably draw even with Nehalem and Westmere. But they won't because it would draw heat from the EU, Japanese, and the American trade regulation bodies.

Exactly. Intel would love to see AMD crushed, but that would mean anti-trust lawsuits, so like I said, the matter is out of Intel's hands.

The success of the K8 has a lot to do with AMD being smart, but it has even more to do with intel being stupid. Intel wanted to segment the market between the Pentium IV and the Itanium. It wanted desktops running the P4 and it wanted every server to be a mid-range or high-end box running Itaniums. They both had the same problems. The Pentium IV, while still x86 required a recompile to see the best performance. It also consumed way too much power. The Itanium and it's completely new IA-64 ISA didn't work well with the backend applications companies ran without the software vendors rewriting their code. Intel overstretched and AMD took advantage of it.

It wasn't AMD's success that allowed them to produce the best performing chips on the market, it was intel's failure. Intel actually had an architecture just as fast as the K8 in the Banias/Dothan but only used them for laptops. If they'd designed desktop/low-end server chips around that micro-architecture in 2003/4 instead of 2006 AMD wouldn't have gotten the Opteron off the ground.

So, AMD didn't actually design the K8? Intel just screwed up and mystically, the K8 came into existence? I'm not sure what you're saying here. Clearly the situation was a combination of Intel making a blunder AND AMD producing an excellent product. AMD had to produce an excellent product in order to take the lead. As far as Banias/Dothan, well, they're really Pentium III chips in their essence - it's amazing Intel ever thought they could improve on the Pentium III with a much longer pipeline. 20 stage vs. 12? What were they thinking?

Do you really think that intel will be that shortsighted again? I don't. I don't think that they particularly liked not having the fastest chips running an ISA that they invented. I don't think that they much cared for what it did to their business.

I think you're ascribing too much agency to Intel. It's not merely a question of what they WANT. This isn't Star Wars, and Intel is not able to use the dark side of the force to crush its enemies, this is real life. In real life, large corporations can fail. You would have been laughed out of the room if you'd predicted GM's downfall in 1960, and I'm sure you can't imagine Intel ever failing either, but the fact is, it can happen, and it isn't all up to whether they like it or not.

Yes, AMD spends a lot on R&D. Do they get much out of it? Not really. The only world-class semiconductor products that AMD makes are from the ATi division. While intel spends a lot less as a function of their total income, they certainly get more out of it. They have the best x86 chips, they have the best SSD controller on the market, they're the only company that seems to be making any progress on PCM (NAND's replacement), they have one of if not the best RAID ASICs.

O.K., I'm not even going to bother with how ridiculous your sweeping and uninformed assessment of AMD's R&D is. You hit on something important, however, and that is AMD's possession of ATI. Intel still doesn't have an answer to the Radeon 5000 series architecture or NVidia's Fermi. AMD has already told everybody they're on a trajectory to integrate their x86 and Radeon circuitry into single processor dies, and establishing themselves as the leader of over-all integer/floating point performance. I don't see Intel making up the lack of ground they've just lost mucking around with Larabee. It's like the Itanium all over again.

Finally, semantics aside, intel and AMD do not trade the performance crown. People seem to think that AMD and intel trade off having the faster chip every couple of years. This is not true. 90% of the time, intel has the faster chips. They have more patents, they have more engineers, they have better fabs, they have more money. It's just that the x86 world order has gone back to it's normal state with intel a generation and node ahead and AMD selling cheap chips. Any lead AMD has in any particular market segment is because either intel is EOLing one product line and getting ready to introduce another (4S/8S) or it's a market that intel sees little value proposition in playing ($50 desktop chips).

I've already made the case for AMD's trajectory, I'm not going to bother typing it again.

[Nater]

The big threat to intel isn't AMD, or any other microprocessor vendor. It's the move from PCs to mobile devices. Mobile devices are not a good fit for x86, but perhaps intel's enormous engineering talent can make it work. Even so, it's putting 10 lbs into a 5 lb bag.

That intel doesn't have a competitor to the HD 5000 series or Fermi is of little consequence right now. About 70% of PCs sold ship without a GPU, the figure for servers is probably like 99%+.

AMD's Fusion processor is no threat to intel in terms of performance. Why? Because it's a very low-end GPU first of all, second of all, intel will have very similar functionality in their native 32nm generation (Sandy Bridge) and will have already had experience in shipping processors with integrated graphics functions.

It's pretty ludicrous to claim that AMD will 'have the highest performing CPU for floating point operations'. That would require a program to both use 100% of available CPU power and 100% of the GPU as well. The problem is that both AMD's and nVidia's GPUs are extremely difficult to program. Especially AMD's.

Larrabee was supposed to be the first really programable vector coprocessor you could buy for a PC. Unfortunately for everyone, intel aimed way too high. The chip, with a high-end load-out of 100 processing cores would have needed something like 1.5TB/s of memory bandwidth. That's 6.5 times the bandwidth that nVidia will be offering on the GTX 480 (assuming they use 4.8Gb/s ICs) and about ten times the bandwidth available on the HD 5870. It just wasn't going to work with current technology.

That sucks, because Larrabee would have not only really brought about a super scalar/vector hybrid processing architecture in the PC, but it would have also forced AMD and nVidia to actually make their GPUs fully programable. With intel pushing, GPU compute could have gone to places that it won't with just nVidia pushing it. Certainly it would have forced the change sooner rather than later.

As for the K8, yep AMD designed it. Still, it wouldn't have been nearly the success had intel decided to use a Banias/Dothan style architecture for their desktop and low-end/mid-range server processors instead of NetBurst. No doubt about it. Why? Because everyone knows that the said architecture was just as fast as K8. When intel updated it for use in desktops and servers, AMD's marketshare started falling about as fast as it's stock price.

[nukefault]

This discussion makes me sad. Let's wrap it up:

1. Intel will probably not destroy AMD soon, and AMD is probably not going to dominate Intel anytime soon. Just like the earth will probably not crash into the sun anytime soon. Intel needs AMD for competition, AMD is very competitive at the low end but has miserable marketing and a massive manpower deficiency and tiny budgets and a huge market that basically lives as Intel's permanent finger puppet. All this means that the status quo will continue until there's an upset probably from a source not related directly to the current competition.

2. Anubis has spent several pages arguing that one place AMD makes sense is when you're already invested in the platform. It's a fair point that AMD's mobos are a great long-term value whereas Intel tends to deep-six its sockets on a regular basis. It's also a fair point that it isn't hard for AMD to seem competitive when they have a $100-300 mobo (and in some cases RAM) pricing advantage. Less flaming and more constructive discussion would make it a lot easier to note these two reasonable and opposing points in one civil post rather than 50 uncivilized ones.

3. I'm only 27. I've been alive for less time than some of you people have been looking at punch card porn (admit it - it exists, you liked it, and the first time you saw ASCII porn you dumped your girlfriends and never looked back) and let I'm perfectly capable of looking at the current market trends in a reasonable, analytical fashion. I hope we can all put aside our abacus lunchboxes and get back to the current simple discussion. Speaking of which...

4. Yes, there are price points and workloads for which AMD makes sense. Not many, but they are out there. Examples:

4a. Budget gaming. X2 250 and X3 435 are both significant improvements over the E6300. You can make an argument for the 255 as well.

4b. Budget multicore via ACC. If your workload is multicore-friendly, you'd have a hard time finding better budget-segment value than a $100 X2 555 that turns into an X4 940 with a simple BIOS option and an X4 965 with a moderate OC. Clarkdale/Lynnfield is inarguably a better architecture, and AMD is inarguably competing on price... but that doesn't mean they aren't competing.

Right now the 555 is $100. It turns into a 965 with minimal hassle. Anyone wanna suggest an Intel CPU that can outperform an X4 965 for $100 in an SMT-friendly workload? Didn't think so.

I should note that AMD no longer has an argument for platform costs even in the budget segment. H55 boards can be had for $85 before bundle discounts. You can spin that to an AMD advantage by comparing it to DDR2 785G boards, but of course there are plenty of cheap DDR2 LGA775 boards. For all intents and purposes, mobo pricing is balanced these days.

So, to summarize: Intel has the upper hand in the market and in architecture and they will, barring unforeseen catastrophic events, be keeping both advantages for a good long time. Intel is inarguably a better option at most price points and for most workloads, but not all. There are a few select cases where AMD makes sense even in a new build. We're all OCers here so it makes sense to compare overclocked results. This generally puts Intel in an even better light since their chips are ridiculously OCable, but it does validate AMD's budget ace-in-the-hole with ACC. $100 for a 3.2GHz quad is simply an unbeatable deal if you need multithreading on a tight budget.

That said, I don't see much value in any of the AMD X4 line. Intel competes with higher clocks, higher OC, and better speed per-clock vs the Athlon X4 and low-end Phenom X4. It doesn't always win with that strategy (as a quick check of Anandtech Bench will confirm) but it usually does and often by very convincing margins. And of course, the higher-end Phenoms have to deal with Lynnfield where they can compete by being a bit cheaper... but cheapness doesn't hold as much weight in the $200 segment as it does in the $100 segment. Lynnfield doesn't have the OC power of Clarkdale, but it's still more power efficient and faster in most workloads.

There's a market for the 965... but it's very small and frequently limited to people who have $180 but can't come up with the $195 for an i5 750, and for some reason don't care that they'll make up the $15 in energy costs.

The rest of that market is going to be citing some very specific workloads in Anandtech Bench.

I'll freely admit I'm a mild AMD fan, and I have an AMD laptop. It's a Gateway NV53 and I bought it because it's $100-150 cheaper than an i3 with comparable onboard graphics so I can run the occasional Dawn of War. I happily sacrificed a bit of battery life and a bit of CPU performance - the M300 is a nice boost over the QL-65 but the i3 still smokes it. If I had to make the purchasing decision again, the $100-150 price advantage and comparable light gaming experience would still put me in AMD's corner... but I fully recognize that my market segment is tiny, and if an i3 machine with similar specs was within $50 I'd snatch it right up.

Apologies for a longish post but I hope this cleared things up a bit and we can go back to being nice now

[Colossus]

I do not believe in comparing a 2 core cpu to a 4 core Intel. Even if you can do so simply in the BIOS.

The Intel I7 920 is only $200 so well worth it

http://www.microcenter.com/single_pr...uct_id=0302727

[RealBeast]

. . . The Intel I7 920 is only $200 so well worth it http://www.microcenter.com/single_pr...uct_id=0302727

Wow, wish I was near one of their stores, that's $90 cheaper than newegg.

[nukefault]

I do not believe in comparing a 2 core cpu to a 4 core Intel. Even if you can do so simply in the BIOS.

We all know how to OC. The discussion is about best value for the money and areas where buying AMD makes sense. It makes sense to buy AMD if you need budget quad-core and you don't mind doing some light tweaking. The value of what is essentially a a $100 X4 965 is unbeatable in the budget quad segment... ACC + OC is a small price to pay and everyone notes Intel's OC abilities as a positive so it makes sense from that perspective too.

[Colossus]

Well lets compare apples to apples. The X4 965 is $185, while the I7 920 is $199. Hmmm hard choice for a $14 difference! NOT!

[nukefault]

Actually the 965 is $180 at Newegg. Details.

Anyway, like I said, we're comparing overclocked value. I mentioned the 965 because the 555 can easily attain that level of performance and it makes for easy benchmark comparisons.

The OC ability of the Intel chips has been widely cited as a benefit in this discussion, yes? ACC unlocking falls in the same category as OC in terms of maximizing the value of your purchase.

The point is that you can get the performance of a $180 965 for $100. That's a steal of a deal. Of course, we'd compare to an OC'd chip on the Intel side, but an OC'd Lynnfield/Nehalem isn't price-competitive and an OC'd Clarkdale isn't performance-competitive. Given that we're going to OC, Intel cannot match AMD at the ~$100 level.

Also, the i7 920 is crippled by platform costs. Cheapest X58 board on Newegg is $160. H55/785g boards are $85ish. Go ahead and add another $75 to the cost of that 920.

So, to compare platform costs:

i7 920: $200 i3 530: $125
X58: $160 H55: $85
TOTAL: $360 TOTAL: $210

965: $180 555: $100
785g: $85 785g: $85
TOTAL: $265 TOTAL: $185

Since you can get 965 performance out of a 555, it essentially means you can get a 965 for half the platform cost of an i7 920. That premium will be worth it to some people, but not to everyone.

To be fair, I should be mentioning OC'd values for Intel chips too, but in this case it doesn't matter. A highly OC'd Clarkdale is still not competitive with 965-level performance in multithreaded tasks (and Clarkdale is still $25 more expensive than the 555 too). Lynnfield at any clock speed is not cost competitive with the 555+ACC. That means there is a market segment that the 555 combo will appeal to. Make sense?

[Colossus]

You still cannot compare them. I do not rely on a fluke of a BIOS hack to change it. Which can change anytime.

So comparing your numbers above the price difference is less then $100 for a better setup. I920 + Board over 965 + Board.

If you need to think about it, you are a lost cause.

[Learux]

Colosus, with all do respect. Obviously looking at your signature you are a power user and want bleeding edge hardware. When I put my rig together the savings for building a AMD versus intel were about$125.

For me that was about 25% of total cost. Most intensive thing I do is game and the cheaper 955 was more then suffice.

I would have loved to have an i7, cost couldn't justify it.

To each its own, no matter how you look at it $100 for a lot of people is a lot of money.

Now lets go shoot at each other in BF:BC.

[nukefault]

You still cannot compare them. I do not rely on a fluke of a BIOS hack to change it. Which can change anytime.

Fluke of a BIOS hack? You mean like overclocking in general? Note that I specified that this comparison is for SOME PEOPLE who MAXIMIZE VALUE by OVERCLOCKING. My argument isn't that everyone should buy the chip, but that it makes sense for some tweakers looking to maximize value in a cheap machine.

And it can change any time? What are you smoking? It's a BIOS setting, just like overclocking.

So comparing your numbers above the price difference is less then $100 for a better setup. I920 + Board over 965 + Board.

Are you ignoring my words? I'm comparing tweaked value. 555 = 965 performance with a simple, reliable tweak. Ditto for a 920 that can run at 950+ speed with OC.

Also, your argument ignores maximizing value. 920+X58 = 200+160 = $360. 965 (to humor your non-argument) + 785G = 180 + 85 = 265. $95 is enough to jump from a 5770 to a 5850, which will make a helluva lot more difference for gamers than a 920 upgrade could ever hope to. Knock off another $80 by going with the tweaked 555 and you're almost at a 5870. I think any reasonable person can at this point see the potential value of tweaking a 555 rather than buying a 920.

If you need to think about it, you are a lost cause.

I don't need to think about it. You're clearly not responding to my argument that a tweaked 555 makes sense over a 920 SOME OF THE TIME for SOME PEOPLE. If I was a ****, I might tell you you're a lost cause or something. Instead, here I am like a civilized person trying to explain the very simple concept that some people like to maximize their available budgets through tweaking, which you seem very resistant to acknowledging.

My argument: "If you don't mind tweaking, this is a good value."
Your argument: "I don't like tweaking."
My argument: "Good for you. Some people do like tweaking, and they will save money for doing it. Have a cookie. The rest of us will save $180 for near-equivalent gaming performance."

[Colossus]

The argument of bios changes is the same argument back in the day about a bios flash on an ATI card that brought what was it a 9600 to a 9800 in performance?

Sorry not the same.

So the less then $100 difference for a quality chip and board over AMD anytime of the week. Video card only makes a difference depending on what resolution you can game at. So you are incorrect about that

Heck if you want to talk about performance get a I920 and spend the extra $100 or less. Overclock to an easy 4GHz and totally blow over the AMD. But we dont want to compare overclock to overclock =) Which takes us back to square one. 965 vs I920.

[Learux]

Don't take this too serious but your avatar says it all

Bottom line is still $125, your argument about the quality chip and board holds no ground and you know it.

[nukefault]

The argument of bios changes is the same argument back in the day about a bios flash on an ATI card that brought what was it a 9600 to a 9800 in performance?

Sorry not the same.

So the less then $100 difference for a quality chip and board over AMD anytime of the week. Video card only makes a difference depending on what resolution you can game at. So you are incorrect about that

Heck if you want to talk about performance get a I920 and spend the extra $100 or less. Overclock to an easy 4GHz and totally blow over the AMD. But we dont want to compare overclock to overclock =) Which takes us back to square one. 965 vs I920.

Really? You aren't seriously grasping at this straw are you? Tell me you forgot the /sarcasm tag.

The goal of budget gaming is to find a CPU fast enough to keep the graphics card happy for as little money as possible, so as to maximize leftover funds for a good gfx card and monitor. You know this. I can get a few extra fps here and there in... uhh, what games care about the CPU that much? A few hardcore flight sims and maybe SupCom 1?

Or I can jump from a 5770 to a 5870. Go ahead: Tell me I'm better off with a 920 and a 5770 over a 555->965 and 5870.

And yet again, I'm not saying the 555 is better for everyone and we should dump the 920. I'm only saying THERE IS A REASONABLE ARGUMENT for the tweaked 555.

Your platform: $360; my platform: $180

That's a $180 difference, not a sub-100 difference. Check your addition skills.

My platform will perform just about as well as yours for half the price. The saved money buys a much better graphics card.

As Learux said, the platform argument mostly died with the 690 and we buried its remains with the 780. Buy it, run it for a year or two until the next gen chips hit and drop in price, then craigslist it and get a new one.

FYI, 780G and H55/P55/X58 all seem to generally sit at four-star reviews on Newegg, indicating roughly similar quality.

Now stop being ridiculous and just admit there's a reasonable argument for people who'd rather change a BIOS setting than pay an extra $180 for a few more fps in a few games.

The 555-> 965 will keep anything short of a 5970 happy in most every game on the market and be just as functional. Your 4GHz Nehalem is certainly faster... but most games don't care at all. If you can keep the graphics card filled, you're good to go. 555->965 fills the card, 965 fills the card, 920 fills the card, and 920 @ 4Ghz fills the card. 920 will have benefits here and there, but the benefits won't be worth it to most people, especially gamers. If you have deep pockets and a real need for speed, then by all means grab a 920 and OC the crap out of it. For the rest of us, price/performance is important and there's very little payoff for super-fast processors.

Come on man, this is simple stuff. You usually make good posts - why is this so hard to comprehend?