Curse

Burst Cancel · Feb 12, 2009, 04:25 AM // 04:25

First of all, pull up benchmarks for Nehalem on desktop applications - that's your i7 not producing the mindblowing performance increases that were expected.

Those Xeon benchmarks are worthless - you're using a server platform to benchmark server applications ... yeah, that's totally indicative of real-world desktop performance. Nevermind that usage patterns and app loads are totally different.

If you read the Dailytech/Anandtech articles you love so much, the top-end 32nm chips are expected to get ~25% clock-for-clock performance boost over i7. Not exactly double, and not exactly i5. Furthermore, the new 32nm i5s will be dual-core, quad-thread. The quad-core/octo-thread will be Lynnfield on the old 45nm process, basically a mainstream refresh of the completely underwhelming i7 with a dual-channel memory controller instead of the triple-channel on the i7.

You're also painting a rosier picture of Intel's release schedule than the reality. First of all, the 32nm process ramp was moved forward by half a year at best, given that it was slated for the first half of 2010 and moved to q4 2009. Second, 32nm was supposed to be 2009 (Westmere) to begin with according to Intel's original tick-tock development cycle - it was delayed to 2010 because they're slipping. Third, part of the reason they're moving up their 32nm process is because their 45nm SoC with integrated GPU (Havendale) got delayed too long to matter - Intel figures that if they can push the 32nm process up, they can roll the whole thing together and just do 32nm SoC. If Intel was actually ahead of schedule, we'd be seeing Havendale.

I've got no problems with Anandtech, but try reading their more relevant articles: http://www.anandtech.com/cpuchipsets...spx?i=3513&p=1

Lord Sojar · Feb 12, 2009, 05:48 AM // 05:48

Nehalem is a great performer. It has proved that time and time again. While the Core i7 920 may not be 100% better than the Q9450 in every single test, it is consistently better in EVERY SINGLE TEST. In addition, the Core i7 920 uses far less power while at the same time doing more than the Q9450. Of course, you are only considering Firefox, Microsoft Word, the OS, WMP, and other desktop apps as far as I can tell. Core i7 wasn't designed for pathetic tasks like those. Its performance is better than its Core2 cousins in those, just not to the mystical degree you somehow expect. Core i7 was designed for enthusiasts looking for a very high end PC that can do anything. Lynnfield is designed for gamers on a tighter budget, while still being an amazing performer. Part of that will be thanks to the integrated PCIe controller, which will absolutely help when it comes to gaming.

And actually, you misread. The 32nm fab gets a 22% percent clock for clock boost above the 45nm original fab planned for release.

And yes, Core i5 is dual core, with HT 2.0. I never said it would be quad core, did I? Nope. Lynnfield is not a refresh of the current Core i7, that is incorrect. Gulftown will be the next refresh of the current iteration of Core i7, and it will have 6 cores with HT 2.0; 12 threads. Lynnfield is simply the mainstream version of Core i7, while Core i5 is more budget oriented, targeted to the budget and midrange markets (it will still perform extremely well) And no, wrong again. The 32nm fab for these particular chips was moved up nearly a full year (9-10 months to be specific). The 32nm fab in and of itself is the story, not the processors themselves. Keep THAT in mind.

I am unbiased when it comes to CPUs and GPUs (albeit I do have stances depending on the situation). You assume I am an Intel fanboy? Really now? If I am anything, I am in the corner of my own company, and am an nVidia fanboy. This release is stunning, and shows the Intel that I am really beginning to appreciate. Dirk Meyer might be able to get AMD back on track after Ruiz botched the company, but it won't be easy. Intel is highly competitive; this isn't the P4 Intel we used to love to hate.

Oh, and the Xeon benchmarks are not worthless. They still show tremendous gains in performance, and despite the desktop variants not being quite as stunning in performance boosts, you can still safely assume a large part of that boost will be present. The same can be said when we test GPUs; while their performance thresholds are typically better in a non x86 environment, the vast majority of the time those thresholds are at least semi accurate in the the x86 environment.

The point is, with all the releases due at the end of this year, upgrading now would be rather silly. That point is irrefutable based on a technological and economic standpoint.

Burst Cancel · Feb 12, 2009, 05:56 PM // 17:56

Quote:

Nehalem is a great performer. It has proved that time and time again. While the Core i7 920 may not be 100% better than the Q9450 in every single test, it is consistently better in EVERY SINGLE TEST. In addition, the Core i7 920 uses far less power while at the same time doing more than the Q9450. Of course, you are only considering Firefox, Microsoft Word, the OS, WMP, and other desktop apps as far as I can tell. Core i7 wasn't designed for pathetic tasks like those. Its performance is better than its Core2 cousins in those, just not to the mystical degree you somehow expect. Core i7 was designed for enthusiasts looking for a very high end PC that can do anything. Lynnfield is designed for gamers on a tighter budget, while still being an amazing performer. Part of that will be thanks to the integrated PCIe controller, which will absolutely help when it comes to gaming.

Nehalem is an incredible performer, actually - if you do the right stuff with it ... problem is, most people won't be doing the right stuff with it, which is what I've been saying the entire time. The relevant question is never, "is it better", but rather, "is it better enough for the price". Unless you're doing HPC/server, 3D rendering, etc., the answer is almost always "no" for i7 - which is why the mainstream market didn't give a shit when Nehalem launched.

Anandtech lists i7 performance improvements of 0~15% over Penryn on apps that aren't well-threaded - i.e., the majority of desktop applications. In many situations, including gaming, getting a higher-clocked Core 2 Duo gives you more performance than an i7 system at a significantly lower price-point. It's even outperformed clock-for-clock in some situations (e.g., GRID) due to its poor cache.

Quote:

And actually, you misread. The 32nm fab gets a 22% percent clock for clock boost above the 45nm original fab planned for release.

Even better, given that I said ~25%.

Quote:

And yes, Core i5 is dual core, with HT 2.0. I never said it would be quad core, did I? Nope. Lynnfield is not a refresh of the current Core i7, that is incorrect. Gulftown will be the next refresh of the current iteration of Core i7, and it will have 6 cores with HT 2.0; 12 threads. Lynnfield is simply the mainstream version of Core i7, while Core i5 is more budget oriented, targeted to the budget and midrange markets (it will still perform extremely well) And no, wrong again. The 32nm fab for these particular chips was moved up nearly a full year (9-10 months to be specific). The 32nm fab in and of itself is the story, not the processors themselves. Keep THAT in mind.

You're basically rephrasing what I said about Lynnfield and Gulftown, except you don't like my use of the word "refresh". Lynnfield is still the mainstream i7, which is exactly what I called it.

The 32nm fab would have ramped up in 2009 originally if Intel had stuck to their schedule. Move along, nothing to see here.

Quote:

Oh, and the Xeon benchmarks are not worthless. They still show tremendous gains in performance, and despite the desktop variants not being quite as stunning in performance boosts, you can still safely assume a large part of that boost will be present. The same can be said when we test GPUs; while their performance thresholds are typically better in a non x86 environment, the vast majority of the time those thresholds are at least semi accurate in the the x86 environment. This is nothing new...

They are worthless. If you'd actually look at desktop benchmarks rather than trying to assume that servers and desktops are similar, you'd notice that they are in no way comparable. See:

Quote:

Originally Posted by Anandtech

We don't see a huge performance increase thanks to Nehalem, we're looking at an average boost of 7 - 12% at the same clock speed as Penryn. The Core i7-965 managed a 7% performance advantage over the QX9770, while the i7-920 pulled 12% on the identically clocked Q9450.

The biggest performance boost is naturally in the 3D suite, the rest of the applications are showing 5 - 10% performance boosts at the same clock.

Hmm ... 7-12%? And 5-10% outside of 3D? Yeah, that really matches up well with those amazing SAP benchmarks ... To say that the new mainstream chips will double performance is nothing sort of intentional misinformation - i.e. unwarranted hype.

Quote:

The point is, with all the releases due at the end of this year, upgrading now would be rather silly. That point is irrefutable based on a technological and economic standpoint.

Sure, upgrading now is silly if you don't need to do it now. If you do, there's nothing to wait for, especially if you're a mainstream user. For them, CPU advances are largely solutions looking for problems. You're going to have a hard time convincing people who are perfectly fine on their Core 2 or even Athlon X2 systems that zomg they need to jump on the new tech bandwagon - unless you lie and hype.

Buy what you need, when you need it, at the best price you can find at the time.

Kattar · Feb 12, 2009, 06:14 PM // 18:14

If you'd like to hurl insults, please do it via PM.

snaek · Feb 12, 2009, 06:29 PM // 18:29

uh oh o__o
*eats popcorn*

rahja may be hypn...
but u'd be on the other end of the spectrum burst cancel
acting as if new tech like this releases every single dam month

btw, the nature of this forum section would appeal to enthusiasts, not jus mainstream consumers
well thats wut i'd like to think anyways

Lourens · Feb 12, 2009, 06:55 PM // 18:55

No need to get scared I'm sure if Intel beats AMD in PWW1 (Processor World War 1 lolwhut?) Mister Obama will rescue amd like they did with Chevrolet

Lord Sojar · Feb 12, 2009, 07:03 PM // 19:03

Burst, the majority of our forum users are gamers, not standard desktop users.

I will simply give you the old adage that a picture is worth a thousand words:

Here we have results from Anandtech, your favorite. Notice... this is FarCry2, a modern game, and many games to follow will use similar programming in respect to threading (because software engineers don't dwell on old technology, they are focused on quad/octa core now)

So, that said... for GAMERS, which this forum is representative of, Core i5/i7 technology (whether that be Core i7 Nehalem, or the more mainstream Lynnfield due out later this year) benefits them greatly in the applications they use. I agree with you, who cares about Firefox performance? I mean, it is already fast enough...

snaek · Feb 12, 2009, 07:50 PM // 19:50

its important to note that, that is vista dx10
the game under xp dx9 i think would have different numbers
(but im jus assuming)

*looks forward to windows 7 and dx11 with wutever processer is the best for it*

Lord Sojar · Feb 12, 2009, 07:53 PM // 19:53

Quote:

Originally Posted by snaek

its important to note that, that is vista dx10
the game under xp dx9 i think would have different numbers
(but im jus assuming)

*looks forward to windows 7 and dx11 with wutever processer is the best for it*

Windows 7 and its subsequent applications will thrive with a processor build around multithreading. DX11 will require cutting edge, very efficient GPUs to get the fullest out of it. DX11 also isn't as bloated as DX10 is. Suffice to say, we will deliver a card to you (the gamers) that will synergize with DX11 extraordinarily well; you have my word on that.

snaek · Feb 12, 2009, 08:25 PM // 20:25

good to hear! ^___^

Burst Cancel · Feb 13, 2009, 01:56 AM // 01:56

You're deluding yourself if you think the majority of the gaming market is running $2k sli/crossfire setups. Did you miss all the guys still playing GW on P4s and X2s? You want to know where gamers spend their money? On their video cards. Because unless you're running at retardedly low resolutions, you're not going to be CPU-limited.

I also love how you continue to pull a single benchmark and try to pass it off as representative. Hey, I can do that too:

Or how about this one:

Or maybe you'll just take Anand's word for it:

Quote:

Originally Posted by Anandtech

Overall in gaming tests the situations where Nehalem was faster than Penryn outnumbered those where it didn't, but upgrading to Nehalem for faster gaming performance doesn't make sense. We were entirely too GPU bound in all of these titles, if you want Nehalem it should be because of its performance elsewhere.

You know what the best part of all of this is? The platform costs for Core 2 and i7 aren't even comparable. You see that E8400 outperforming that i7 920? The E8400 costs $165, while the 920 costs $290 - that's a 40% difference in cost on the processor alone. Combine that with the fact that x58 motherboards ($300+) and DDR3 are still significantly more expensive than P45 and DDR2, and you start running into the question of, "why bother?"

You know what? Forget it. Once the product actually launches, it will be pretty damn clear whether it's worth buying or not. And until it launches, this is all speculation - and hype.

I'm sure it will be every bit as awesome as you claimed. Like, double performance and totally low cost and everything. I can hardly contain my excitement.

Kattar · Feb 13, 2009, 02:09 AM // 02:09

As I said before, if you want to be hostile, take it to a pm.

With that said, I'm locking this. Tech Corner is not the place for this sort of behavior. Leave it alone, or go somewhere else.

Edit: At the request of another user, I am opening this back up. The discussion may continue. If there are any remarks posted that are even remotely antagonistic or hostile, this thread will be closed and all users involved will face disciplinary action.

This is your final warning.

nebuchanezzar · Feb 13, 2009, 06:27 PM // 18:27

This is a topic that should be discussed civilly. Both Rhaja and Burst have valid points in their respective views. The hostility of the views does need toned down just a hair though I agree.

The big news is NOT just that a new CPU is coming out in the latter half of this year. The big news is the 32nm die fabrication and Rhaja did try to point that out. The i5 is designed to be exactly what i7 is not, and that is a mainstream CPU. Ci7 was targeted squarely at extreme enthusiasts or businesses that need raw CPU power. Burst has a very valid point the the average Joe User is not that audience. The i5 however is targeted at the Joe User.

Rhaja is also very correct in pointing out how truly aggressive this is from Intel. While it is an incontrovertible fact that the smaller die process is cheaper to manufacture, Intel has notoriously charged a premium on their products. The only exception to this is when AMD has a competitive product. A great case in point is how the Q9770 was cut 40% when it was shown that the PhenomII's were competitive with the Core2's. So while we can assume that the i5's will be substantially cheaper than the i7's are, no one knows how much Intel will attempt to bleed the consumers on it.

This is going to greatly affect computing in many many ways. The northbridge will go extinct. This will make the R&D and manufacturing costs of motherboards go down a bit(substantially in some cases) and that should also pass down to consumers. DDR3 memory will have matured by this Fall also which may make it an attractive upgrade price/performance-wise. These changes are just as important as they have industry wide affects.

To ignore these facts does not change them. The i7 is only 1% of Intels sales and that is not an issue with them at all. That seems to be the design all along as i5 will run a seperate socket(I forget what, 1156 or something)from the i7s socket 1366. This alone shows a strong separation intended from the very design of the chips.

This is very worthy of discussion. The "take it right away" or "hold back until you absolutely have to" is a personal decision. I have put together computers for both reasons. Some, no purpose besides OC'ing until I break it. Some for a functional purpose and I intend them to do that until they fry out in a decade or more. What you choose is of course your choice but, if you are considering a new PC in the near future than waiting is a wise decision.

We can make the argument for price/performance all day but in the end that is a subjective matter for each of us. This news is most relevant to those gamers that are STILL on a P4, or socket 754, or even 939. These are the people that will need an entirely new system, not just an upgrade. There is a very real possibility of getting a better performing PC for the same or perhaps even a lower price by waiting until this fall. If anything Rhaja is speaking out against i7 as i5 WILL be a better fit for most user and should be less expensive all around.

For those on Socket 775 or AM2/AM2+ this information is less important. Those users are still in the position where simply purchasing a newer and stronger graphics card will make the most substantial difference on their PC's and the cost conscious will do just that. However, the vast majority of those boards are DDR2. DDR3 will replace DDR2 and that makes any upgrade paths more convoluted. They can replace their motherboard and ram and graphics card to move up to DDR3 should they choose to. That will probably not happen until this Fall however and it may be a better cost/benefit decision to also upgrade to an i5 CPU at the same time.

I apologize for my wall of text and would really like to see a more open discussion on this. Many of the arguments are moot points really. Yes, in gaming a dual core performs equal to and sometimes better than a quad core..today. That will change but it is up to each of us to decide if it is worth the expense now, or the upgrade expense later. Information like this helps us to make a more informed decision and should be used and discussed.

Burst Cancel · Feb 13, 2009, 07:16 PM // 19:16

I have difficulty being "civil" when I see statements like this:

Quote:

Well, it means buying should absolutely be held off until this release. The technology leap will be enormous, and pricing should reflect the near 50% reduction in production costs Intel will see from moving to 32nm.

and

Quote:

Core i5 will DOUBLE the performance of current Core2 product lines.

Blatant misinformation, followed by blatant misrepresentation of facts (i.e., using i7-based server platform SAP benchmarks as representative of desktop i5 performance).

As for that 32nm process all of you keep talking about - check the transition from 65nm to 45nm during the Core 2 generation for the actual consumer effect of process shrinks, especially in the absence of a competitive AMD. How much performance did that shrink get us, and how much was the cost reduction "reflected" in the end-user purchase price? Here's some relevant reading to get you started:

http://www.anandtech.com/cpuchipsets...oc.aspx?i=3069
http://www.anandtech.com/cpuchipsets...oc.aspx?i=3137

And again, Intel moving 32nm up to 2009 isn't significant for anyone except Intel. If they'd stuck to their original tick-tock schedule, 32nm in 2009 was the plan all along; don't act like Intel is doing you favors.

The really interesting question here is whether AMD will have anything to compete with Intel anytime soon (8-ball says, "Outlook not so good"). Competition is what actually drives prices down - otherwise cost reductions just translate into higher margins for the manufacturers.

It's not just a simple disagreement. Please consider what this looked like from my perspective. We know Rahja is something of an enthusiast; he's up-to-date and knowledgeable about hardware. Then he claims something sensational like "DOUBLE PERFORMANCE" and when asked to back it up, uses SAP benchmarks that he knows aren't representative - instead of readily-available desktop benchmarks. Given Rahja's tech background, that didn't strike me as being wrong - it struck me as lying.

My positions in this thread might ultimately be wrong, but all statements made herein of my own knowledge are true, and all statements made on information and belief are believed to be true.

In any case, I'm done. I think I've presented my viewpoint on this release and sufficient background material that anyone who wants to weigh the merits of my argument can do so for themselves. Ultimately, people will think what they want to think.

nebuchanezzar · Feb 13, 2009, 09:08 PM // 21:08

I was not trying to diminish any of your points Burst.

The possibility for a double performance increase of Core2 is possible simply from HT2.0 whereas the Core2 has no HyperThreading at all. This is of course a theoretical maximum increase and real world results never produce that. That's like saying your SATA 2.0 HDD has read/write of 300Mb/sec. Any who look into it know that is not the case.

Quote:

As for that 32nm process all of you keep talking about - check the transition from 65nm to 45nm during the Core 2 generation for the actual consumer effect of process shrinks, especially in the absence of a competitive AMD. How much performance did that shrink get us, and how much was the cost reduction "reflected" in the end-user purchase price? Here's some relevant reading to get you started:

hmm, I'm pretty sure I said the same thing

Quote:

While it is an incontrovertible fact that the smaller die process is cheaper to manufacture, Intel has notoriously charged a premium on their products. The only exception to this is when AMD has a competitive product. A great case in point is how the Q9770 was cut 40% when it was shown that the PhenomII's were competitive with the Core2's. So while we can assume that the i5's will be substantially cheaper than the i7's are, no one knows how much Intel will attempt to bleed the consumers on it.

Your last statement echoes Rhajas last statement in his OP albeit in a different light.

Quote:

The really interesting question here is whether AMD will have anything to compete with Intel anytime soon (8-ball says, "Outlook not so good"). Competition is what actually drives prices down - otherwise cost reductions just translate into higher margins for the manufacturers.

That's very valid and I agree. AMD is about 1 year behind Intel in development terms imo. This would have stayed constant if Intel only introduced a limited amount of 32nm CPUs in mid/late Q4 of this year. They have instead tossed the 2 cores that would have been designed at 45nm for mainstream computing. Instead of releasing a high-end 32nm first like has been their history with die shrinks, they are putting out the mainstream CPU's first and holding off on the Socket 1366 replacement until 2010. They also put in the 2010 plans of the on chip GPU. That puts AMD between 1-2 behind in development, again imo.

The PhenomII is what Phenom should have been, flat out. Now Intel is putting out the on chip(not quite on die but close enough for all intents and purposes) GPU that AMD had been working on and should have had out this year also BEFORE Intel. That's not going to happen though.

fwiw, if anything I am more of an AMD fanboi than Intel although I have computers from both camps and probably too many at that. I am quite pleased to see AMD's PhenomII finally competitive. They have however totally had to hand Intel the "extreme high-end" market. Intel now wants the whole thing and appears willing to aggressively price those 32nm mainstream i5 CPU's. Not just for desktops but for laptops even moreso. Many sources are already presenting the same idea Rhaja put out for laptop users. Laptops are actually outselling desktops last I checked.

I'm going to stop writing as I end up dropping this mountainous walls of text as if I am speaking in person and it's terrible for forum communication. I'm sorry that I'm not able to put thru my statements more clearly.

I read the link you offered here is the last page. I think the conclusion is pretty much spot on and will echo what I already posted. Which happens to be a middle ground between your apparent "FUD" and Rhaja's apparent "hype."

*edit*
I do agree that Rhaja's posts did seem a bit hyper-inflated from my point of view but he offered the links where he got his information from and did not simply make it up. You both used some non-representative charts also that would favor your side. Grid was on medium settings where-as Rhaja's were Ci7 tests that fully utilized the inherent advantages of the CPU in ways that desktop users simply don't. You are absolutely valid to point that out, but I would be remiss to not point out you did something similar.

Fril Estelin · Feb 13, 2009, 09:57 PM // 21:57

Very impressive, but: 1) the market has proven that pure power is not enough to make a whole platform a "success"; 2) is Intel dominance a good thing? (for all those that hate "trusted computing", I can tell you it's not)

Burst Cancel · Feb 13, 2009, 10:49 PM // 22:49

Your defense of the double performance claim is without merit. First of all, if that had been the reason, why the need to dig up SAP benchmarks and ignore desktop benchmarks? Second, I'm sure Rahja, being a tech enthusiast, already knew that HT doesn't give doubled performance; it's not exactly a state secret. Doubling the number of cores doesn't even double performance (except in perfectly threaded applications, and then you have to take in communication overhead, cache, etc.), and on poorly-threaded applications (like ... most desktop applications), the performance improvement approaches zero. Do you see people claiming that quad-cores have double the performance of dual-cores? A blanket claim of "double performance" based on hyperthreading is something even the most spin-happy and irresponsible marketing departments would be proud of.

And as for my use of non-representative charts, you must have missed the sarcasm in my post. I specifically said:

Quote:

Originally Posted by Burst Cancel

I also love how you continue to pull a single benchmark and try to pass it off as representative. Hey, I can do that too:

(emphasis added). In other words, I specifically stated that I was going to do what he did. You know, to like, prove a point.

---

Phenom II might be what Phenom "should have been", but it's not good enough to allow AMD to compete on anything except price. While that's good for the consumer in the short term, it's not exactly good for their bottom line or long-term competitiveness. AMD has also been conspicuously absent in the netbook market, where much of the sales growth in the industry has been. As I've said, we've moved beyond the point of home users needing new machines - right now, "good enough" is pretty damn good. That's a problem when you're trying to convince people to shell out high-end chips like Nehalem, or even Lynnfield.

The integrated GPU is most interesting to the mainstream non-gaming market, particularly for mobile and low-power applications. Unless there's a benefit I'm not aware of, the IGP isn't at all interesting to gamers ... especially not an Intel GPU that has a history of significantly underperforming IGPs from AMD (780) and nVidia (8x00 series).

Lord Sojar · Feb 14, 2009, 01:01 AM // 01:01

Those with budgets will want to look to Core i5 to get the most bang for their buck, hands down. That was my point, really. Users who are looking for a PC that can play modern games at medium settings with little worry to upgrade 3 years from now will not want to go with a DDR2 bound solution. That is the primary issue. This release from Intel gives users the opportunity to buy a 32nm solution ahead of schedule, and not have to upgrade for quite sometime assuming that they are buying Core i5 for the right reasons.

While I agree that the boast of 100% performance gains on the desktop environment might be over the top, it is valid to assume a very high degree of improvement based on what Intel learned from Nehalem vs Core2, and what the newest Xeon's performance ratings are shaping up to. If a user can get a system that has 25%+ performance boosts over current gen technology, for a low price, and that will have expandability 2-4 years in the future, they should absolutely go Core i5. From an economic stand point, it just makes sense. But, if that user already has an E8600 with 3-4GBs DDR2 800/1066, then it would be their discretion to upgrade or not.

Core i5 isn't built for those with E8x00 or better CPUs. It is designed with Athlon X2 or Core 2 E4400 processor users in mind. Those who have been holding off for as low as possible with a Pentium4/D, or an Athlon64 can also buy this Fall knowing they are getting a long lasting product that Intel has commited itself to (socket 1156). Intel's roadmap for LGA-1156 is robust, no denying that. It is a safe upgrade, and one that won't break the bank. Performance will be great (not double in many tests, but still within enough of a margin to be worthwhile), and cost will be very low (32nm fabs are extremely cheap to produce, and mainstream products have very little of the "early adopter fees" associated with enthusiast level hardware)

And I agree with Kat. The Tech Corner is not for arguing. A healthy debate without hostility is fine in my book, but let's leave the semantics out (that applies to myself as well).

I think upgrading should be postponed unless a user absolutely needs a new system now, or if a user wants something that is enthusiast level. If you disagree, explain why. I am open to reason and logic.

Burst Cancel · Feb 14, 2009, 02:37 AM // 02:37

Our first and largest problem is that we're trying to talk about too many different markets at once - Core i5 and even i7 might make sense for some markets, but absolutely no sense for others. Even narrowing to "mainstream" doesn't say anything, because a lot of performance improvements are application-specific; whether or not you use Photoshop, for example, might change everything, even if the rest of your usage patterns remain the same. That's why I almost never agree with blanket recommendations (maybe in the Athlon v. Core 2 days ...). One guy might get a 40% boost in his most-used applications, while someone else might wait for half a year for a 5% increase at 30% more cost.

The move to DDR3 depends on pricing and time horizon. Memory isn't a big factor in performance, which is why people on a budget tend to shell out for high-end video cards instead of high-end overclocker memory. The pricing delta between DDR2 and DDR3 is quite steep, and while it isn't quite as ridiculous as it was when Nehalem launched (what was it, like 5x? 10x?), it's far too large for a comparatively trivial component. This is particularly true at the high speeds where DDR3 really starts to benefit - at lower clocks, DDR2 has performance parity with DDR3 due to DDR2 having tighter timings. Right now, you can pick up stacks of DDR2 on the cheap - enough to tide you over for several years while the next technology refresh settles down. If you have relatively short upgrade cycles (1~3 years), or if you tend to replace rather than upgrade, grabbing a DDR2 solution now at its dirt-cheap prices makes more sense.

Given the current economic wildcards and an uncompetitive AMD, it's hard to predict where i5 pricing will be when it launches. Furthermore, keep in mind the overall platform cost is the number that matters; the i7 processors themselves aren't overpriced compared to Core 2 Quads - it's the motherboard and DDR3 price premiums that push it over budget. A low-cost processor isn't going to make much of a difference if other prices don't fall in concert.

"Upgrade path" was the same argument used for the Nehalem launch - basically, some people were convinced that you absolutely had to wait for i7, because Core 2 was being dead-ended. That argument isn't any more compelling today that it was a year ago, and for the same reason: Core 2 is solid enough that being dead-ended won't matter. As I said earlier: "good enough" is pretty damn good, nowadays. Many of those who decided not to wait for Nehalem bought Core 2's last year, which they expect to last them all the way to Sandy Bridge - they're skipping the Nehalem/Westmere generation entirely.

To wait for Core i5, you're basically gambling that the performance improvements over what is available today, in the applications you use, will be large enough to justify waiting almost a year for a product with a currently unknown price. Whether that's a good gamble or not is ultimately going to depend on the individual.

Blackhearted · Feb 14, 2009, 03:38 AM // 03:38

Quote:

Originally Posted by Rahja the Thief

And Blackhearted... regarding the IGPU on chip in the Core i5. These chips are, again, aimed at midrange consumers looking for excellent performance/dollar. 95% of the world's PC users do not need anything more than an Intel IGP. The Core i5 will simply improve the basic functions these users utilize, without adding to the cost (if anything, it will reduce the cost)

I highly doubt that adding on the igp is going to come free. As one can't add more and more to something without it becoming more expensive to make. Which in turn, is always put on to the consumers price. Even with the 32nm die shrinking I can still see it being very likely that igp-cpus will be more expensive, by atleast a little bit, compared to their cpu only counterparts.

And while you may be right that alot of users don't need a real gpu, having intels trashy gpus inside cpus is bad. We need someone competent at graphics to do this first. Not someone who's chips struggle at even basic video features. As i can see even futher integration of intels crap gpus into pc's can only hurt it's capabilty to be called a gaming platform.