Does AMD Need Six-Core Microprocessors for Desktops?

Does average Joe even need quads? Should Intel even make them? Should AMD make them? During a recession, having a viably priced product is pertinent to sales, not "top of the line". In todays world, when we hear more and more often, "its good enough" or, "its fast enough" for peoples needs, or, "why overhaul my entire rig for so little gain?", its amazing this question even crops up, should AMD do this or not?

"As reported previously, there will be no TWKR processors released commercially. "

http://www.xbitlabs.com/news/cpu/display/20090626180759...
This is final? I havnt heard that. I think this is a bloated piece, for hits n giggles

Also, since Intels own predictions on "Intel Corp.’s previous-generation Intel Core 2 Quad processors," will carry more than 90% of their market on DT til 2010, whats the point?

This whole article is full of hype n holes

To me, this is setting up as AMD being incapable, and thats simply not true. This isnt their core of theirmarket, and with the conomy, and since theres really nothing out there but the hype from this article itself, why should they?

If someone needed that power, they'd be using Opterons by now.

I'm guilty of that  

I'm having quite a lot of fun rolling in the land of low-powered HTPC & NAS builds actually. **All high-powered work has been put on hold until further notice  **

And in the server side of things, sometimes more cores, more RAM...etc. for virtualisation isn't the answer to all.
Look at those babies! There are Atom variants as well from SuperMicro. They're not as energy efficient nor denser than traditional blades, but they're a cheaper solution for the tasks they're designed to carry out.

And what about Larrabee?

The new Larrabee cards are x86-based. It will be interesting to see how programs and games will see the card without any special software.

The Larrabee cards will take the load off the CPU. Havok physics on the main CPU? No, do it on Larrabee.

I can see many programs like WinZIP, transcoding video apps, Adobe etc making it possible to use BOTH the main CPU and the Larrabee at the same time.

Of course, this is Intel's goal. It would be nice to see laptops and generic low-powered PC's that have a high-speed graphics interface so you can just add a Larrabee "box" externally and poof you have a high-end processing machine.

As for "do we need" 6-cores? That's kind of obvious. If people buy it they'll build it.

So you need some help overclocking then?

As long as Intel keeps with their crazy charging for their next higher up model, the i920 will continue to get squeezed more n more by P2, as stock clocks will continue to climb, and therell virtually be no difference between the 2. They may already be there if the 965 comes out at 3.4 as rumored. Ocing isnt what drives the main market. Cutting edge isnt either.
I see this article as fail. They want AMD to push ahead in a terrible economy towards the highend, where few sales will occur, and little SW benefits on DT. Dont they/we see the slow take up for i7 overall showing a higher priced scenario, with skt change etc , and having your cpus spread drastically in price (ala the i940,EOL) isnt really where its at?
For us, the enthusiast, itd be great, but here again, they knock AMD for their TWKR edition. Its just fail as an article, plain and simple, + some anti AMD potential as well

Me as well!  

With the power of the upper even low end of the i7 and the Phenom II X4 how much of that power can even be utilized in most task from every day task to CPU intensive applications? Is there a point where we have surpassed the need for more power? Not that it's a bad thing to have power.  

Its more a, the HW has arrived, wheres the SW thing. Remember, unlike the past where we saw huge gains in clock speeds and even IPC, now we simply double the core count, and we have theoretically double the perf.
Since some apps or portions of apps will remain too serial in nature to MT to, theres going to be diminishing returns no matter how many cores we have.
On the other hand, since MT is still relatively new to SW prog and devs, theres still alot of low hanging fruit out there as well. It all takes time, and the article wants a struggling AMD to make a huge charge in this direction? In a much smaller market? Costing?

According to the article even with 6 cores at a range of 2.4GHz to 3.0GHz AMD still won't be able to compete against the Core i7. It would and should only help in multi core applications and for that if your going AMD just stay with the X4 in your desktop (that's just my opinion) that's in less your going to use the extra cores most of the time. But then again i don't really know how many people would buy it. Is it worth the effort are should they just keep working on perfecting their Phenom II X4?

I agree. I Think perfecting multi core apps would perfect the multi core CPU.

I think most everything lands in the hands of the software developers. And by the apps supporting more cores the demand of 4,6 and even 8 cores would grow. Then i think the cost of a CPU with that amount of cores even at the low end of the core speeds would be able to justify it's self.

Heh, since the cheapest Istanbul CPU is what - $500? - AMD releasing a DT version would sorta blow a hole in your "Intel doesn't get it" ideology now, wouldn't it  

Presumably AMD has done some market research and concluded there is currently not large enough a market in DT for it. Or they are afraid that it won't cinch the honors as the highest-end CPU currently available. Intel, OTOH, obviously wants the bragging rights which is why they released the i975 and next year the 6-core Westmere DT CPU...

IMHO since software (Win7 in particular as mentioned in the article) is making progress to use more cores, real or virtual, then it will make more sense as time passes to progress to 6 cores. As Buzz Lightbeer used to say, "To 6 cores and Beyond!"  

I understand pushing ahead is paramount for us enthusiasts, and I totally agree with that. But, heres where we see some differences of what this article says, reality, and average Joe.
Where has AMD let down the enthusiast market? Show me a Intel quad going 7Ghz. Show me where AMDs current gen doesnt perform close to the top in gaming. Thats 2 huge segments of the enthusiast community where AMD is fitting the bill

Now, lets move to reality.
http://www.tomshardware.com/reviews/multi-core-cpu,2280...
It shows just how important multi cores are "becoming", and how far we have yet to go, and what currently is effective. 6 cores goes beyond cleeves findings, while his findings seem to show were barely scratching the surface of usage of quads.

Now, lets move to average Joe. Will he need 6 cores? Can he find any usage currently for it? Will he also pay more for something that he may never use?

As to pricing, 500$ doesnt change anything for DT does it? Since its NOT a DT solution, but enterprise, where we usually see a much higher cost/expenditure anyways, so, if either Intel or AMD price their future 6 cores at low end of 500$, all I can say is, good luck to them both. And it doesnt have anything to do with whether my Intel doesnt get it or not, unless the second higher 6 core is only 30-50$ more after release, and IF they do that, then Intel will be "getting it"

Q what is DT ?
sorry for hte dump Q

Desk Top

Well the difference between a suicide shot oc and the "enthusiast" market is that many more people could afford the $1K needed to buy the highest-performing CPU, than could afford to buy 500 liters of liquid Helium and the other equipment for achieving monster oc's. Besides, I believe the ancient Intel P4 holds the WR for overclocks anyway - where is AMD's single-core record holder?  

As for "reality", that is a time-dependent variable. The THG article shows what is the case with recent software, not what we will have even a few months from now. As the OP's article pointed out above, software such as W7 is trending towards more effective use of multiple cores. So here's a case where perhaps AMD "doesn't get" the impending future  .

This is nothing more than a DX11 scenario, or any DX thats preceded it.
While HW may exist, theres no immediate SW follow up. These things take time, and what theyre saying IS now, now later, when even and also, Intel plans to release theirs as well, when its more user friendly, and theres a market and demeand for it.
Makes sense to me, and not to berate AMD, like a rock n a hard place by saying, "they should release it now", or the obvious alternative of ignoring markets, or to imply AMD doesnt get it. Its a no win scenario as given from a article that either doesnt know markets, doesnt know trends, doesnt know AMD, and their particular market with the goods they have, and good common sense, but whatever.
Like I said, the article has fail written all over it, with some anti AMD potential as well, as being witnessed here and now

amd has come to the realisation that they dont have the megabucks for r&d that intel has so it thinks to itself, should i spend my time and money developing a six core twelve threaded cpu with 12mb of cache that can only support a line of under 50 boards where the cheapest board is $169 .00 and the cheapest cpu if found for $250.00 is concidered a bargain and only one chipset that i make supports it, oh and it only supports the latest memory technology, or should i create a line of cpus that the other 90% of humanity can afford, where the cheapest board at $50.00 supports from the lowest end sempron to the highest end phenom II, i can go from a board with 1 pcie slot that supports ddr2 for $50 to a board with 4 pcie and ddr3 and keep my cpu instead of having to buy a new cpu immediately if i want any type of performance gain, or maby i could buy a $90 cpu and a $180 board and upgrade the cpu later andhave a backup cpu if i ever fry my cpu, interchangability is the name of the game in maistream where if somethig goes bad, i can dig through my box of computer parts and be up and running in minutes, instead of needing to come up with $200, i wouldnt fault you, im planning to move to i9 when theyre released, and financially im already preparing, but you can bet that if amd says wait till q2 and theyll allow me to simply replace my x2 4400+ and and put in an x6 with the same board and everything with a minor perf hit when compared to intels offering...? what would u do?

Does AMD Need Six-Core Microprocessors for Desktops? Nope.

Neither does Intel. Marketing exists to sell you (or convince you of the benefits of) something you don't really need.

The biggest upcoming ripoff is Win WVSHB - "What Vista Should Have Been"

As fun as The Core Wars have become they will 'fail' to the superior parallel processing power of the GPU with improved instruction sets.

"Multi-Core" will soon mean a couple of CPUs, a couple of GPUs and a couple of S(pecialized)PUs for individual needs ...

It's simple enough for AMD right now. They have the clock speed myth well in their favour, and with that they should exploit it even though they are behind in actual speed. Intel did it for long enough.

Not only that, AMD have clock speed *and* cores advantage. When the ignorant see 6 cores at 3ghz, they'll be queuing up to buy it. 6x3ghz = 18ghz 'wow'. That is how the non-pc savvy work it out.

Intel would love to have something like that to market right now instead of languishing Nehalem sales which are mostly down to it being a mediocre upgrade over the core2, 2.4ghz (99% of buyers do not overclock so they have no clue about how well i7 overclocks) and seriously overpriced with that in mind.

The real fight is at the core2/phenom2 market. Nehalem hardly even matters.

But how long will Intel want to push old technology that they have already made there money on? Like the Core2/Quads. Its all about getting the on aware and clueless buyers to upgrade to the newer technology even if they don't need it. The same goes for AMD.

Sadly, cpu's are already maxed out in real useful terms. The software will never catch up, apart from scientific stuff which will generally use all the power it can for number crunching.

This is something that intel just have to deal with in future. The average person had no need for a mainframe-like cpu sitting on their desktop 10 years ago, now they have one almost by default and it's going to be a really hard sell to convince anyone that they actually need to spend money 'upgrading' in future.

But, thats where gpgpu comes in. Cheaper overall, better perf in alot of apps, Open Cl and W7 and DX11 etc all pointing in this direction, and Intel making LRB, spending untold millions if not more for it, and yet, they say its all about the cpu, where anyone who can see farther than their nose knows were heading towards a blending of parallel and serial computing, and that the future is mainly gfx applications, sound apps etc

Im talking for saleability as far as gfx etc goes, not other apps, which arent gfx related, which as jennyh pointed out, are mainly scientific in usage

OK , in short......... Yes, Intel and AMD need to continue on multi-cores. We could bash this back and fourth for days and days, but it will remain that it is a must.........!!!!!!!! As for the average joe using six to eight cores daily.... not very likely. Soon enough that will change and personaly ill be ready long before that day comes and most intense gamers will defintely upgrade or be left behind. For that matter more people will follow the trend than we can probably fathom. HATS off to our hardware GURU's.

The problem is software vendors need to catch up to even dual core. Once all the software developers get used to writing multi-core optimized products, we will see more of a need for multi-cores. Right now the only thing that can take my C2Q over 15% usage is F@H and games.

There needs to be a push to create an IDE for developers that makes it much easier to program for multiple threads. Nobody is going to program for unlimited threads in notepad.

'The need to create an easier method of programming for multi-threading'

Haven't we been hearing this phrase for the last 6yrs or more?

And what that mythical reverse Hyper-threading from AMD?  

I don't know, I probably couldn't install Windows 6 years ago.

Six years ago I ran a dual Athlon MP machine, even at that time I heard people screaming better ways to program for multi-threading. Since then I've heard it being mentioned on a yearly basis.

There is still no easy way to do it. Programs we have today that does multi-threading were coded with the intent to take advantage of it i.e. coded very differently and complicated

The biggest bottleneck now is still I/O and the Operating system ... both bog down the cpu and GPU.

Stick two SSD's in RAID0 ... you might have to sell a kidney.

Plus ditch Vista ...

meh ...

Hmm, most interesting especially since the highest-clocked Istanbul is 2.6GHz, not "3". I'd be careful about tossing around flame words like "ignorant" if I were U  .

OK, one area where increased computing power for DT or mobile would be a big seller, would be realtime speech recognition that is accurate enough so that we could lose the keyboard & mouse entirely. Current SR algorithms are based on digitizing input speech using DCT or something similar, then categorizing each phoneme based on a probability comparison to a stored vocabulary. Fairly computation-intensive esp. when you have to account for background noise, differences in speakers which is why the better ones require a training period to adjust the model, etc. This should be ideal for some GPGPU or equivalent like Larrabee, and Joe Blow would probably pay a few extra bucks for this feature esp. for mobile or netbook, and if they can get the power low enough, a smart phone.

Somebody open a can of tuna fish or something??? There seems to be a lot of cat avatars here  

That was when Intel killed the Pentium HT...      

I forgot about Reverse HyperThreading. The new 6-core CPU's will only show up as 3-cores!

Yea, but 3 x 2.6 is still 7.8 Ghz.

]

Just what we need ...

A World full of Joe Blows walking around telling their netbooks and smart phones what to do  




This.

As an old SMP dude I've been hearing "it's right around the corner" since the Pentium Pro multi-sockets ...

My C2Q is better than that: 4 x 2.5 = 10 GHZ!!!!!

But it doesnt have reverse HT. And wouldnt that 8.7Ghz and 0.1?

Maybe, just like we see sometimes on SMT, they dont use reverse HT? Cuts the cores in half

LOL - well I'd rather have them speaking to their devices instead of keyboarding or texting while driving. At least their eyeballs would presumably be on the road ahead..

Trouble with present-day SR technology is that it is only about 80 - 90% accurate on average. Yeah that sounds impressive but whut yf eveby otter keesttoke om yoor c!mpoodur w z wrong??!?

A wonderful plan! Unfortunately doomed to failure because of the whole "spelling" and "grammar" thing.... I mean, whut yf eveby otter werd oot owt yoor mawf w z wrong??!?

LOL - I can agree with that sediment  

However, miraculous present day linguistics technology can correct quite a few grammar errors, same as spelling errors, since it relies upon rules & dictionaries. Of course that takes processing power & storage, even more so if you include a natural language translator.

IIRC, lack of digital processing power led to such things as 'analog' processing equivalents like with neural networks 20 years ago or so, plus the speaker had to train the system extensively, speak words separately with pauses inbetween, and use a quiet environment and good microphone to max the S/N ratio. That was OK if you were in a private office with no secretary and had no typing skills, but not so OK in most other situations.

It was even worse 50 years ago - on Star Trek (original series), Kirk gave orders to the Enterprise's computer via speech recognition. It was really tough to cram Majel Barret into that little speakerbox on the wall!

Sew their!  

I can see it now: 'Ebonics' language packs for Win CE, or a REAL version of the old Windows 2000 for Hillbillies!


<considers breaking out the "shizzolator" - but comes to his senses that some things are just too dumb for words>

Back in her day, I wouldve loved to...neer mind