Intel's Core i7 870 & i5 750, Lynnfield: Harder, Better, Faster Stronger

343 Comments

View All Comments

• Inkie - Sunday, September 13, 2009 - link

  "Another really stupid remark was how AMD processors would only make sense if they drop the price. Have you forgotten AMD processors are coupled with excellent IGPs? That's been the big selling point for a while, and this hasn't really changed that, too much. IGP platforms are really big in sales too"
  
  You are really dumb and confused (to copy the kind of reply you make to others). Anand mentioned that the Phenom II X4 965 BE needs to fall in price. How many people are buying a $245, 140W TDP processor for use with an IGP?

• Anand Lal Shimpi - Tuesday, September 8, 2009 - link

  Let's go through this point by point shall we?
  
  "One thing that's obvious is, after reading this, that the esteemed reviewer modified his review in order to fulfill his preconceived ideas of what he wanted to say."
  
  I could actually say the same thing about your comments :)
  
  "Most people know this site isn't technical, but still, I always thought it was unbiased. It's biased, big time, just not for a company. It's about making a review say what they want."
  
  Any thoughts on what the motivation to do that might be?
  
  "For example, I doubt Anand is really as stupid as he sounds in this article. LGA1366 has so many advantages, certainly he knows this. But why the idiotic remarks about asking why it exists? "
  
  I believe I listed the advantages in the conclusion. But honestly, for the majority of users, the i7 920, 940 and 950 are made redundant by the Core i5 750, i7 860 and i7 870.
  
  "Why did this review ignore the fact that most people here overclock? Why did it ignore that most people reading these reviews won't use DDR3 1066 memory? Did this site just try to become PC Magazine without the writing talent, and go stock? I think not. It was because they only wanted to give information that was consistent with their preconceived idea. "
  
  I included two pages on overclocking investigation for Lynnfield, concluded that Bloomfield was a better overclocker and we've consistently looked at Bloomfield overclocking in all of our previous articles. I don't believe we've ignored overclocking at all.
  
  As far as memory speeds go, I try to use the fastest memory officially supported/validated by the platform. Of course we've run Bloomfield at higher speeds and Lynnfield at lower ones. I have DDR3-1066 data for Lynnfield:
  
  SYSMark 2007 (Overall/E-Learning/Video Creation/Productivity/3D)
  
  Core i7 870 (DDR3-1333): 233 199 245 235 257
  Core i7 870 (DDR3-1066): 230 195 246 230 256
  
  You're looking at less than 1.5% in the overall performance test.
  
  Given that several readers have requested it, I'll look at Bloomfield at DDR3-1333 in an upcoming article. It's not going to be a huge difference but it's worth addressing since there's interest. Remember what I said in the article: it takes three fully active cores running bandwidth intensive code to stress Lynnfield's dual-channel DDR3-1333 controller. Bloomfield doesn't change that, it already has enough memory bandwidth at 3 x 1066MHz, upping that to 1333MHz shouldn't account for much.
  
  "But, let's look closer. The memory controller on the Bloomfield is actually faster than the Lynnfield when using the same memory, even when running in dual channel mode. Why wasn't it mentioned here? "
  
  I don't believe that's the case, the Lynnfield memory controller is actually very similar to what's in Bloomfield.
  
  "Why wasn't a clock normalized comparison between the LGA1366 processors and LGA1156 processors made, or even attempted, to get an idea of what the architecture changes accounted for? Strange that this very important data is missing? I think so. "
  
  This was an oversight on my part, I didn't realize many would want it after the results from the 750 and 870. It's something that I'll include in a follow-up sometime next week.
  
  "Why weren't overclock processors compared? I mean, will anyone here buy an i7 920 and not overclock it? Probably not many. Since the only real advantage is the more aggressive turbo mode, this was what the article was based on. But, in reality, for readers here, it's not important, since people are going to overclock, and the i7 920 would wipe the floor up with the Lynnfields in the configuration that would be used. Why no mention of this? "
  
  The i7 920 and Lynnfield parts we reviewed here today can generally overclock to about the same levels (they just need voltage). I don't believe anyone is missing any part of the story here. Are you suggesting that users would only buy an i7 920 and overclock it but not overclock an i5 750?
  
  "The technical savvy will still opt for it in a lot of instances, since they will overclock it. If you have to recommend something for a friend that won't, these things are fine. "
  
  Again, the overclocking community will be split into those who overclock but don't care to increase voltages and those who do. The latter group will be served just fine by Lynnfield and save money. Do you not agree?
  
  Our readership spans both overclockers and not, so it's important that we characterize performance at stock as well as overclocked. Lynnfield, regardless of clock speed, will be similar to Bloomfield performance.
  
  "Have you forgotten AMD processors are coupled with excellent IGPs? That's been the big selling point for a while, and this hasn't really changed that, too much. IGP platforms are really big in sales too, so, I'd agree with the premise that AMD probably should lower their prices a little, but not that they don't have advantages even where they are."
  
  AMD does have better IGPs, but not $50 better. If you take the difference in price between a Core i5 750 and a Phenom II X4 965 BE you can buy a better discrete graphics card than any IGP.
  
  Take care,
  Anand

• iocedmyself - Wednesday, September 9, 2009 - link

  I have to agree that the technical end of this review, even to the extent of hardware/software configs in benchmarking suite is pretty non-existent
  
  As for any semblance of an unbiased approach, the only way this review could have possibly been more pro-intel is if it were replaced with a picture of a guy on his knees with a hand gingerly wrapped around an intel logo while licking his lips and gazing dreamily at a belt buckle looming slightly above the logo.
  
  As for the logic behind AMD IGP's not being "$50 better when taking into account the price of the 965BE" it's so flawed it makes my panda very very sad. BE's are unlocked...they are for overclockers, and a 200mhz difference is not incentive enough for anyone with even a whisper of a clue to shell out $55 over the 3.2ghz 955BE. I have an undervolted 945 at 1.2875v that is rock solid at 3.6ghz and the half dozen PII chips i've overclocked have all been able to undervolt while still hitting 3.4.
  
  Also something that should be obvious, even buying a 965 BE is still still an upgrade part for AM2+ MB users, where as it's intel's bi-yearly "here's our latest architecture cut down for mainstream..with the briefly anticipated opertunity to buy another new intel cpu socket!"
  
  Then there is also the fact that anyone looking to upgrade machiness in the corprate world would not want a discreet graphics card, in fact they pretty much need an igp which also makes the argument of the 965BE cost irrelivent really as i doubt quads would be in the running for consideration, but oh well.
  
  95w 3.0ghz 945 BE costs $169,
  65w 2.5gjz 905e costs $179(likely choice for corporate quad solution)
  125w 3.2Ghz 955BE $187
  140w 3.4Ghz 965Be $239
  
  Being that 780G is stong enough to run Vista ultimate it's more than strong enough for win7 which will be the OS of bussiness shortly and would also handle even a 965BE so MBs are...
  
  780G Starting at $59-$94
  785G starting at $69-$99
  790GXstarting at $79-$114
  790FXstarting at $89-$188
  
  4 gigs of DDR2 1066 $44-$59 if using a 780G/785G
  4 gigs of DDR3 1600 $79-$99 if using a 790GX/790FX
  
  Well gosh....it looks like those unlocked BE chips that short of damage or inadequete cooling can exceed 3.4ghz without a voltage bump. Then their is also the fact that intel doesn't have a comparable IGP to go up against AMD's....the i5 boards infact don't have any IGP's at all nor will they for a good long while. Meaning that if you want to use this cpu you are obligated to also buy a motherboard, and a discreet video card and most likely ram.
  
  So what exactly is this "not $50 better"BS? As it stands the motherboard is included in the price of the I5 putting the price around $320 or so with combo deals.
  Which is about the same price as a 790GX+965BE,
  or $65 more expensive than a 955BE+790GX
  or $100 more expensive than the 955BE+790GX+ cost of discreet card for intel system.
  But no, this was very impartial and professional....
  
  

• goinginstyle - Thursday, September 10, 2009 - link

  So your suggestion is to buy a $240 965BE, install it on a 780G motherboard (good luck with that), and that should offer all the performance I need in a system. So this setup will handle everything including gaming, not playing poker on the internet gaming, but running Crysis or Flight Simulator X?

• TA152H - Tuesday, September 8, 2009 - link

  Anand,
  
  The first part of your response is pretty lame. I really have no vested interest in the P55 being better or worse. I just dislike misinformation. I don't think it's bad, I just don't think it's all good or all bad. You went way out of your way to make it all good, with just little mention of the bad so it seemed unimportant.
  
  Yes, the motivation would be that you built it up, and you probably were shmoozing with Intel, and really wanted this thing to be all that and a box of cheese crackers. Reality is different though, but you didn't want to see it. After all the build up, if it sucked, you'd have eaten a bit of crow. Really, it's not bad, but you really exaggerated your remarks so much, I don't know how you can defend them. There's no point in anything below the 975? What? The 920 is a great processor for so many people. It's got much better i/o for a i/o based server, overclocks like a devil, has better video performance in dual card mode, rapes the Lynnfield where memory bandwidth is important, etc... No uses? You really should have qualified it better, and when you don't, you come off as biased.
  
  I will agree with your remark that the majority of users aren't going to find the i7 920 attractive. The reason being, the majority of users don't overclock, or have seriously i/o heavy applications, or use dual video cards. If you had worded it that way, I would have agreed. You didn't.
  
  You missed the point on the overclocking, and I suspect that's on purpose, but let's assume not. You compared normally clocked processors on all benchmarks, with inferior memory for the Bloomfield. This is what people look at. A few pages at the end saying the Bloomfield overclocks better doesn't have the same impact, especially when, if you overclocked and took benchmarks, the i7 920 would wipe the floor up with the Lynnfields. The reason is the turbo mode advantage would evaporate. I agree you should have shown them normally clocked (although with equal memory), but also overclocked since most of your readers will overclock. You didn't. You showed Lynnfield in the best light possible.
  
  I agree the memory controller is very similar, but the Bloomfield seems to beat it slightly, very slightly. It's only meaningful in the context that it is faster, instead of slower as indicated in some of your benchmarks.
  
  With regards to them overclocking to the same speed, again you miss the point. If you overclock them to the same speeds, the turbo mode advantage of the Lynnfield disappears, completely changing the picture, and giving people a different story. So, yes they would both overclock, but it would help the i7 920 a lot more.
  
  I agree these processors are fine for a lot of people. That's not my contention. You basically exaggerated their abilities, by aforementioned reasons, and then boldly said the i7 920 and i7 950 basically have no purpose (although, with regards to the 950, maybe I don't disagree :-P). So, it's not me saying these things are pointless, it's more my reaction to you saying the stronger LGA1366 platform isn't useful outside of the context of a $1000 processor.
  
  You do make a valid point with the AMD versus Intel IGP situation, especially considering electrical use. I'm really frustrated with AMD not selling a better designed chip, although, just as many people are frustrated with Intel having such poor IGPs. I'll concede this though, you make an excellent point.

• Anand Lal Shimpi - Wednesday, September 9, 2009 - link

  I've responded below:
  
  "I really have no vested interest in the P55 being better or worse. I just dislike misinformation. I don't think it's bad, I just don't think it's all good or all bad. You went way out of your way to make it all good, with just little mention of the bad so it seemed unimportant. "
  
  I tried to highlight the good points: turbo, power consumption, price as well as the bad points: multi-GPU scaling, overclocking. The contention here appears to be that you view Lynnfield's memory performance as being much worse than I do.
  
  "Yes, the motivation would be that you built it up, and you probably were shmoozing with Intel, and really wanted this thing to be all that and a box of cheese crackers. Reality is different though, but you didn't want to see it. After all the build up, if it sucked, you'd have eaten a bit of crow. Really, it's not bad, but you really exaggerated your remarks so much, I don't know how you can defend them. There's no point in anything below the 975? What?"
  
  One thing I believe is that you always have to check your ego at the door. Intel never briefed me on Lynnfield, the turbo excitement and my interest in the processor all came on my own. I got my first chip on my own, outside of Intel's jurisdiction and saw a lot of potential in it. I believe it met my expectations. Had it not, you would've heard here. I wanted the first value SSDs to be amazing, but they weren't and that's what folks got to read about here.
  
  It's interesting that you view my comments as exaggerated, as that's what most of your Lynnfield criticisms appear to be. Perhaps this is all an issue of misinterpretation, hooray for internet communication :)
  
  "The 920 is a great processor for so many people. It's got much better i/o for a i/o based server, overclocks like a devil, has better video performance in dual card mode, rapes the Lynnfield where memory bandwidth is important, etc... No uses? You really should have qualified it better, and when you don't, you come off as biased."
  
  Let's go through this point by point.
  
  1) The Core i7 920 is an excellent processor. I'm not sure what you're referring to by it has better i/o for an i/o based server. It does have QPI, which only really helps it in talking to other processors (which it can't do right now, only Xeons get that ability). It also has more PCIe lanes thanks to X58, but I'm guessing that most people who are pushing several GB/s over PCIe for something other than graphics would opt for a server board and Xeon platform instead.
  
  2) It does have better performance in multi-GPU mode, especially with 4 GPUs, but I already mentioned that.
  
  3) It does have more memory bandwidth, but not significantly more. Xbit did a comparison with Lynnfield and Bloomfield at DDR3-1333 at the same timings. Their Everest results showed the following advantages for Bloomfield: 6%, 14%, 4% (read, copy, write). Now those aren't real world performance advantages but rather bandwidth advantages. Very few desktop applications are bandwidth constrained, so if you're lucky enough to find one then you might see half of that in a performance boost. More likely than not you'll see something more negligible.
  
  4) The gains are expensive. The 920 costs $284, you can buy an i5 750 for around $80 less plus opt for a cheaper motherboard. Overall savings? At least $100, most likely more. Never do I say that the 920 is a bad processor, just that it's difficult to recommend it given the price savings. You may view that as biased, but I'm trying to recommend the more affordable platform that delivers a large percentage of the same performance.
  
  "I will agree with your remark that the majority of users aren't going to find the i7 920 attractive. The reason being, the majority of users don't overclock, or have seriously i/o heavy applications, or use dual video cards. If you had worded it that way, I would have agreed. You didn't. "
  
  That was my intention in the conclusion where I said:
  
  "As I see it, LGA-1366 has a few advantages:
  
  1) High-end multi-GPU Performance
  
  2) Stock Voltage Overclocking
  
  3) Future support for 6-core Gulftown CPUs
  
  If that list doesn't make you flinch, then Lynnfield is perfect. You'll save a bunch on a motherboard and the CPUs start at $196 instead of $284. "
  
  I didn't include the bit about an i/o server but the rest seems spot on with what you're asking for. If the reader doesn't find themselves concerned about items 1 - 3 then Lynnfield is a good option. I'm sorry it didn't come off that way, but that was my intention with that paragraph.
  
  "You missed the point on the overclocking, and I suspect that's on purpose, but let's assume not. You compared normally clocked processors on all benchmarks, with inferior memory for the Bloomfield."
  
  Again, I tested both at their highest officially supported frequencies. I'll point you to the Xbit data for now. We're talking about an average performance advantage of 8% in a memory bandwidth test for Bloomfield at DDR3-1333 vs. Lynnfield. I'll have a follow up with some real world data, but you're not going to see that translate into anything significant in real world apps. Most apps are not 100% memory bandwidth limited, if they were we would see no performance scaling at all with CPU speed.
  
  "This is what people look at. A few pages at the end saying the Bloomfield overclocks better doesn't have the same impact, especially when, if you overclocked and took benchmarks, the i7 920 would wipe the floor up with the Lynnfields. The reason is the turbo mode advantage would evaporate. I agree you should have shown them normally clocked (although with equal memory), but also overclocked since most of your readers will overclock. You didn't. You showed Lynnfield in the best light possible. "
  
  Lynnfield and Bloomfield appear to have similar max overclocks when overvolted. With turbo disabled, and both chips around 4GHz, it boils down to a memory controller comparison once more. I don't believe you'll find the huge performance differences you're looking for.
  
  "So, it's not me saying these things are pointless, it's more my reaction to you saying the stronger LGA1366 platform isn't useful outside of the context of a $1000 processor. "
  
  Of course the Core i7 920 isn't a useless processor, even at stock speeds it's seriously one of the fastest things you can buy and in many cases it's even faster than any Phenom II/Core 2 Quad. The LGA-1366 platform loses a lot of its appeal (or usefulness) because there's now something that delivers very, very similar performance, at a noticeably lower platform cost. That was the entire point of the article and something that is echoed in many other reviews of the processors (including the THG review).
  
  Take care,
  Anand

• TA152H - Wednesday, September 9, 2009 - link

  Oh, I forgot to ask a couple of questions.
  
  I read on another site that the Lynnfield lowered the L1 cache latency to three clocks instead of four, but X-bit shows the same speed as the Bloomfield at four clocks. Do you have any idea which is correct? I still don't know why this processor has to be slow with the L1 cache, so I'm hoping they bring it down to three clocks. Yes, it's only a few percentage, but it's still bizarre. It's not a high clock speed processor, (as opposed to the Prescott), and the cache isn't so big (it's smaller than the AMD processors), but it needs four cycles? Strange. I think they used slower memory so they could lower power use, but I could be wrong. I'm not really sure of that.

• goinginstyle - Thursday, September 10, 2009 - link

  TA152H (or should we say Rich from Toms Hardware),
  
  You have not answered one important question. Considering all of the major sites came to the same conclusion and tested in the same manner for the most part (including your website), why do you not copy and paste your ramblings here at other sites. You are very good at copy and paste routines based on your articles.
  
  Simply put, it is so obvious you are posting here to try and make this site look bad due to your affiliation with another site.
  You cannot claim to be fair and impartial when your actions state otherwise. I ask once again, when are we going to see your ramblings at Toms, Tech Report, Xbit, and others?
  
  In the meantime, prove to us that a Bloomfield platform is better to the point than a Lynnfield platform that I should spend extra money for similar performance or worse actually. Where is your article describing all of these points you continue to make? Where is your proof? If you are so good, lets see an article from you proving all of your ramblings.

• TA152H - Wednesday, September 9, 2009 - link

  Hi Anand,
  
  On point one, I'd agree some would get a Xeon, but sometimes, we buy parts (I do this all the time), and when they get older, we find additional uses for them. The superior i/o of the i7 920 can really come in handy, and it's going to be cheaper than a Xeon in most cases. There are still a lot of i/o based applications that can surpass the crippled i/o of the P55, especially with SSDs.
  
  With respect to point two, Xbit really screwed up their article by being imprecise when they measured the memory timings. This is what I find so irritating, computers are precise, articles about them need to be equally so. My contention is based on this - the Bloomfield being used isn't really clear. Unless it's the EE version, the uncore is running slower than on the i7 870. Based on the L3 cache scores, it wouldn't seem so, but it's not at all clear. Also, were the processors allowed to go into turbo mode? Even for memory speeds, this can change the results slightly. It's just not clear how they benched, and they really needed to be for this data to be valid.
  
  On the other hand, when I looked at a lot of benchmarks on other sites, you see some really clear performance leads with the i7 920 on real benchmarks. I'm not as quick to dismiss the additional performance as you are. Maybe what you consider insignficant, I don't.
  
  As far as the i5 750, to me this isn't a competitive processor with the i7 920. It's in a different market, and I don't believe it's useless. Within the context of people that overclock and such, I think anything higher starts to become less supportable, because you start running into a better platform that has features you simply can't get with the P55. You can just overclock the i7 920, but you can't add additional PCI-E lanes, or magically get an additional 64-bits of memory bandwidth with the P55. To me, once you step over that line, unless you're going purely stock (which most people do), these processors are just too expensive to compete with the Bloomfield. The i5 750, however, I'll agree, competes against a different animal, the Core 2, and therefore is attractive even to people who know how to overclock, due to cost.
  
  With respect to overclocking, again, you just aren't seeing what I'm saying. Most people here WILL overclock a 920. If you compared at 920 and the P55s overclocked, the story would be better. You're speculating about the memory bandwidth, whereas you had the ability to produce concrete results. You're way too experienced to know that everything you think should happens with computers, does. We're always surprised, right? But, more to the point, the results you give show an advantage for the Lynnfields because they overclock, when in the configuration most people would use, they would be slower.
  
  Anand, since when did you become stock????? I mean, that's really not what your site is about, yet, this is what this article is. It's just not your typical fare. You can understand why I'd be suspicious? I mean, really, anyone reading this article is going to find your memory used worthless, because no one will use DDR3 1066 for the Bloomfield. We have to extrapolate, with any tests, from what you used to what we would consider using. It's always best if it's the same, and in this case you chose a memory that your readers will not use. You have to know this. DDR3 1333 is cheap enough no one will get 1066. Again, when did you become Mr. Stock? It seems to me to show the Lynnfield in the best possible light. Where I am, wouldn't you think the same?
  
  Take a look at your own Excel benchmark. The Bloomfield destroys the Lynnfield, even the much higher clocked 870(especially if it's turboing). Would your memory numbers have indicated this?
  
  Let's look at some of your remarks in the article -
  
  A) Lynnfield's memory controller: Also faster than Bloomfield.
  - In fact, it isn't, unless you use faster memory. It's really slightly slower.
  
  B) Discovery: Two Channels Aren't Worse Than Three
  - Yes, they are. A lot of benchmarks show this. It's only a matter of degree.
  
  By the way, the benchmark also disproved your theory about memory bandwidth only mattering for three or more. Since, by your own indications, only two cores were running, and the Bloomfield still had higher numbers, running slower memory, this would necessarily be false.
  
  C)The next thing that the Core i5 750 does is it finally ends the life of LGA-775.
  - This is plain wrong. It COULD, if Intel wanted it to, but, really, it's not even close to that. I know you're trying to be dramatic here, and respect that, but, really, it's just wrong. For one, most people buy IGPs, and there isn't one for the P55, yet. On top of that, the IMC complicates IGPs a bit, so, LGA-775 will do fine right now. This really changes very little. If Intel wanted to keep it around, it's a lot smaller, and could be cheaper. It's really up to them what they do with it.
  
  I'd agree it makes a whole lot of Core 2 processors unattractive, but you didn't say that. In terms of market impact, you made a gross exaggeration. When the IGPs come out, we'll see more impact. But, you didn't use a future timeline in your statement.
  
  D)If you didn't have a well threaded application, Bloomfield wasn't any better than a similarly clocked Penryn.
  
  OK, now that's flat wrong. I mean, really, you ran enough benchmarks to know this. It has a totally different cache architecture, and much faster main memory bandwidth. Your review of the brain-damaged i5 750 even proves this. Even without turbo mode, it would beat the Penryn in almost every benchmark by a decent bit. Strange, strange, strange remark.
  
  E)The Core i7 870 gets close enough to the Core i7 975 that I'm having a hard time justifying the LGA-1366 platform at all.
  - This is the one that really got me. For a good proportion of people reading this, there are only two processors that make the most sense. The i5 750 if you going kind of low end, or the i7 920. The market for all the others is really small. Why? If you start getting into Bloomfield territory, you should get the real deal and overclock it. You get better performance, flexibility, and expandability. I hate to use absolutes, so there are going to be exceptions, but really, for most of us reading this, the i7 920 still makes a lot of sense is so many situations.
  
  You can't justify the LGA-1366 platform anymore, except in very limited situations? You don't think that's worded a bit too strongly?
  
  A couple of other things. If you use the same memory, you'd see bigger differences in the benchmarks where you do video benchmarking probably. Since the main use of this bandwidth is transferring memory, the slower memory obfuscates the important advantages of the Bloomfield, especially since using main memory creates contention with processor.
  
  Anyway, this is my book writing for the day. Whew, and you thought you were wordy.

• Lifted - Wednesday, September 9, 2009 - link

  Please don't waste your time replying to this child. He's making a mountain out of a molehill. He compared the difference between Bloomfield and Lynnfield with such beauties as "wipe the floor" and "raping". He constantly mentions overclocking, so I assume he's referring to games with his colorful comparisons, which makes them even more ironic since most games are GPU limited and often show ZERO benefit from different high end CPU's. Even he has his computer encoding h264 24/7, I don't see how somebody can get so excited over the few minutes they'll save each day.
  
  Feel free to ignore him, your time would be better spent writing articles, being with your family, or even picking your nose.