SPCR

Thanks for the review. It was informative.

As of now, it's looking like I'll be holding out for cheaper ddr3 ram and lower prices on the i7 920s. That..and waitin for those damn ssd's to come down to earth....

Long live 775!

It's all nice and cool and all, but if you have a dual core and a decent/good 775 mobo supporting 45nm and Quad CPUs and capable of OC, you are still good for a year or so, maybe more. Unless, of course, you wish to play your games with 6% better framerate or zip your files 12 sec faster.

Another thing is that 775 quads and the higher clocked 2-core CPUs will now get cheaper. Kinda win-win. You want an entirely new setup - go for i5, you want to upgrade to 4 cores - stay with 775 and find a cheap(er) new/2nd hand Q9550. Or even 9400 for the time being and switch to sth with more cache a little later. At least that's my plan Heck, if you don't mind a little extra heat & noise find a Q6600 - still a capable CPU that OCs to 3,6 with no effort (but yes, it does need a good cooler).

I was trying to figure out why your benchmarks had these new chips performing so poorly compared to the other reviews published today, then I noticed you had turbo mode disabled. I don't understand the rationale behind disabling it while benchmarking the chips performance. No sane person is going to buy a chip then run it with one hand tied behind its back - wouldn't it make more sense to test the configuration people are going to actually use?

I have to agree with RBBOT - why test with Turboboost disabled? Calling it "overclocking" is like calling Cool N Quiet or Speedstep "underclocking and undervolting". It's an integral part of the chip, and is enabled on a per-load basis based on thermal and power factors. Why test idle power consumption with power-saving features enabled when you don't test load consumption with the performance-enhancing features enabled?

please fix this glaring bug... Lynnfield core is exactly the same as Bloomfield. Even the idea of it not having any L2 cache at all makes me ROTFL.

mentawl wrote:I have to agree with RBBOT - why test with Turboboost disabled?

I guess the power figures would fluctuate quite a bit then. However, a short comparison table for one CPU with Turbo mode would be nice.
That reminds me of my good old 486DX with the "Turbo" Button on the case - 33Mhz instead of 16... ah, good times...

Back to topic: These processors look like serious trouble for AMD. There are now even less reasons to get a Ph II quad-core, which leaves only the X3 parts and the HPC market where AMD is objectively better.
Good for the consumer in the short term, bad for competition in the long term... *crosses fingers for the upcoming ATI cards*

K.Murx wrote:Back to topic: These processors look like serious trouble for AMD. There are now even less reasons to get a Ph II quad-core, which leaves only the X3 parts and the HPC market where AMD is objectively better.
Good for the consumer in the short term, bad for competition in the long term... *crosses fingers for the upcoming ATI cards*

I have to agree with this one big time. The Phenom II X4-955 is listed at $189 on newegg right now versus $210 for i5-750 which is a very narrow price gap. However - the i5-750 is way faster, like 20% faster in pretty much every respect. Worse still, the 955 generally only overclocks about 18% and I've seen reviews of the i5-750 saying 50% O/C should be possible on a good mobo and top air cooling.

So let's see - the Intel chip costs a little more than the 955, but is still $30 cheaper than AMD's 140W 965, the Intel chip is a lot faster, it uses less power, should be easier to cool quietly, has a socket with more upgradeability, and on yeah - overclocks way more than AMD too, making the already faster chip ahead by an even larger margin, to people for which that is important. AMD is in trouble here. They keep saying their next chip architecture, Bulldozer, won't be ready until 2011 now. The question is - will AMD even be in the CPU business by then, or will they have given up and gone GPU-only?

RBBOT wrote:I was trying to figure out why your benchmarks had these new chips performing so poorly compared to the other reviews published today, then I noticed you had turbo mode disabled. I don't understand the rationale behind disabling it while benchmarking the chips performance. No sane person is going to buy a chip then run it with one hand tied behind its back - wouldn't it make more sense to test the configuration people are going to actually use?

I agree with RBBOT on this. Lynnfield is defined by its Turboboost. I don't understand why anyone would purchase one with the intention of disabling it except for perhaps some unique circumstances.

I've been eagerly awaiting this release for months now so I'm glad it's living up to expectations. For a silent-pc/gamer enthusiast it hits all the right notes. My current Athlon 64 (socket 939!) system can retire in style.

I guess the power figures would fluctuate quite a bit then.

I can understand having it off for measuring idle power, but not for the rest of the review.

Ugh, it looks like every review site got sent an i5-750 and an i7-870, but I can't find any benchmarks comparing the i7-920 to the i7-860, which share a price point. Obviously, the 860 will be the winner in power consumption, but I still want to see the performance numbers.

swivelguy2 wrote:Ugh, it looks like every review site got sent an i5-750 and an i7-870, but I can't find any benchmarks comparing the i7-920 to the i7-860, which share a price point. Obviously, the 860 will be the winner in power consumption, but I still want to see the performance numbers.

HardOCP included an i7-920 in their review. Bottom line - at the end of the review they point out Intel no longer shows the i7-920 in the product lineup going forward because it's redundant against the i7-860.

HardOCP CPU review: i5, i7-860, i7-920, Phenom II 955, C2D 3.2ghz

mczak wrote:please fix this glaring bug... Lynnfield core is exactly the same as Bloomfield. Even the idea of it not having any L2 cache at all makes me ROTFL.

Sorry about that, I made a typo in the chart and then went by that. Fixed.


With regards to Turbo Boost we were trying to give a better indication of what the performance difference is clock-per-clock. Sure, TB is a built-in feature, but it does give it an unfair advantage compared to the other processors compared.

We can't make that our default testing state as its an unfair advantage over the other processors compared, even though its a trivial process to overclock them manually. We prefer not to overclock at all since there are plenty of other sites that can cover that more extensively.

Umfff. These CPUs are superficially interesting, but too damned expensive (and don't get me started on the motherboards). £400+ just to upgrade the CPU + mobo + RAM of my current system (yes Virginia, that is $640), and that's not really even pushing for the higher spec parts.

That said I'm glad they're on the market and so performance competitive at the price points they're at - hopefully they'll prompt AMD into some savage price cutting!

Lawrence Lee wrote:With regards to Turbo Boost we were trying to give a better indication of what the performance difference is clock-per-clock. Sure, TB is a built-in feature, but it does give it an unfair advantage compared to the other processors compared.

We can't make that our default testing state as its an unfair advantage over the other processors compared, even though its a trivial process to overclock them manually. We prefer not to overclock at all since there are plenty of other sites that can cover that more extensively.

I'm sorry to keep harping on about this, but how do you define it as overclocking? It's a standard feature on the CPU, enabled by default - in many ways it's exactly like Speedstep or CnQ. I just don't understand how you can refuse to test load performance/power consumption with Turboboost forcibly disabled, while you'll test idle power consumption with Speedstep or CnQ enabled - doesn't that seem kinda odd to you?

Personally, I don't think you can call it overclocking when it's built into the CPU and it is part of the default behaviour of the chip. Calling it "unfair" is kinda backwards too, as it's an architectural improvement that means the CPU is a fair bit better than almost any other CPU out there (both in terms of efficiency and performance) - surely it's a decent chunk of the reason to buy an i5/new i7?

Just seems like a glaring omission that I'm struggling to understand.

Thanks for the review. I'm impressed by the idle performance as shown in this and other reviews. There's a great review (that I can't seem to find a day later) that goes into Intel's process/design approach to power management on the 4 cores...using somewhere in the neighborhood of 1M very low leakage transistors to act as breaker bar style kill-the-current switches for unused cores. Thus, the incredible idle power.

Now, if only the GPU guys would get their act together and enable similar idle power performance.

swivelguy2 wrote:Ugh, it looks like every review site got sent an i5-750 and an i7-870, but I can't find any benchmarks comparing the i7-920 to the i7-860, which share a price point. Obviously, the 860 will be the winner in power consumption, but I still want to see the performance numbers.

Guru3D has the only review I've found so far that includes the i7-860.

*Edit* There's also the bit-tech review.

I ordered an 860 yesterday, but with a couple cashback discounts, so the price comes out pretty reasonable.

I have to disagree about Turbo being overclocking. It the future of CPUs. You have to abandon the thought of 'quadcore' and 'dualcore' concept. They are merely resources. You could turn the tables around with these things. They are fast dual-core chips that can act as a slower quad-core when the app is well threaded. The point is to go back to the 'idle' state as fast as possible. It's not like turbo increases power usage too much since it doesn't increase voltage.

The only reason to actually disable turbo is if you wish to overclock these things into FAST quadcores. Which is kind of in conflict with considering turbo overclocking.

You're not running these with HT disabled either?

Haven't read the entire thread, but at glance I can see a lot of argument over why turbo wasn't enabled. To give a different perspective, I don't understand why enabling it is so important. Lawrence wrote that SPCR was trying to give an accurate clock-for-clock comparison against other cpus. To me this is a legitimate reason. To include turbo just because it's a "standard" feature doesn't make it a fair comparison. This is similar to SPCR's policy of using a standard reference fan in their heatsink reviews so we can compare the innate heat-transferring ability.

And for that point about SPCR testing using Speedstep/CnQ, it's because that is what SPCR is interested in. SPCR is not interested in testing the number crunching performance limit, but it IS interested in testing how low the power consumption can go.

Remember, a "fair" comparison isn't the same as testing each product as-is.

Scoop wrote:It's not like turbo increases power usage too much since it doesn't increase voltage.

The multiplier is locked on the current Lynnfield CPUs so turbo mode can only work by increasing the cpu base frequency which you could call clock speed. You may not agree but by any definition that is overclocking. While voltage may remain constant there is a significant increase in the system power draw, which can only arise from an increase in amps.

For the 870 for example the increase from turbo mode is 666Mhz, i.e. it rises from 2.93Ghz to 3.59Ghz. You can also of course manually overclock. One reviewer, Guru3D, overclocked an 870 to 4.1Ghz. This increased power draw for their system at idle by 13w, at load it was 133w. So on that basis turbo mode could result in about an 80w increase in system power draw. All these numbers are AC at the wall. Even allowing for that, these are significant figures which cannot be simply ignored.

Guru3D comment "....What a lot of you do not realize that overclocking a processor can consume heaps of wattage...". Sorry but we do, because we have been this way before with the Core i7 CPUs. And that I think is the point, which to a degree you also make, the Lynnfields are a Nehalem variant and in terms of power requirements and cooling should be treated like the LGA 1366 Core i7s and not as a sort of improved Core 2 Duo or Quad Core.

As to a Lynnfield being a fast Dual Core and a slower Quad Core. Not really. What you really have (with the 8xx Lynnfields) is up to 4 actual cores and up to 4 virtual cores. So faced with a demanding single threaded app a Lynnfield will put 3 of its actual cores into effectively a sleep state, ignore the virtual cores and concentrate system power on just one core. This can include implementing turbo mode, yes on just one core. Depending on the app, it will then bring in further cores up to the actual maximum of 4 when it will start to deploy the virtual cores. If turbo mode is turned on it will implement turbo mode before bringing another core into use.

On the 7xx series Lynnfields there is no HyperThreading so it just works on the 4 actual cores, but otherwise it works in just the same way. You could of course run a 7xx with turbo mode off, but given the way the chip works because it will simply use more cores instead although the lack of turbo mode will ultimately limit system performance.

PartEleven wrote:Haven't read the entire thread, but at glance I can see a lot of argument over why turbo wasn't enabled. To give a different perspective, I don't understand why enabling it is so important. Lawrence wrote that SPCR was trying to give an accurate clock-for-clock comparison against other cpus. To me this is a legitimate reason.

Clock-for-clock comparisons usually aren't very interesting, since architectural differences mean you'll get different real-world performance per-clock from different CPUs. Ofcourse, Bloomfield and Lynnfield have pretty much the same architecture, but the biggest difference between them is arguably the much improved turbo mode. Disabling that leads this reviewer to a very odd conclusion:

It would seem that these LGA1156 processors bridge the performance gap between LGA1366 and LGA775, offering a nice performance boost over Core 2 Quads, but falling short of the Bloomfield i7's.

Most other reviews are saying that Lynnfield is in fact a pretty good "base clock-for-clock" improvement on Bloomfield, since the turbo modes go a lot higher and have a much improved implementation. While it's true that the fastest i7-9xx processors still outperform the currently available i7-8xx models, most of the current i7-9xx line-up is made redundant by the i7-8xx models. So the conclusion that they fall short of Bloomfield seems very much unjustified.

TLDR: While the reviewer set out to do a fair clock-for-clock comparison, he ended up comparing apples to oranges, as the improved turbo modes make for very different real-world (as in the speed these CPU's will end up doing most of their work at) clock-for-clock situation.

Luckily this is a minor quibble as most of the article focuses on power draw, for which turning off turbo mode was a good thing.

Don't forget as well that Turbo mode has the potential to considerably effect power efficiency of chips, in the strict "work done per unit of time" definition. Given that the benchmarks used will vary in how multithreaded how they are, the ability of a Lynnfield chip to ramp up one or 2 cores and then quickly clock down to idle will produce different (and in my mind, more realistic) results. Just check out the power efficiency results on the Tech Report review.

Thanks for the review! Yours and Anandtech's were the two I most looked forward to.

I hope to see you guys test the Gigabyte and MSi uATX s1156 motherboards. Would love to see how they stack up voltage-, stability-, and performance-wise.

Also, word is Lynnfield i7 chips with Win7 are quite a pair, given the way the turbo mode and hyperthreading are managed in Win7, providing a very efficient use of time and cores, and thus power usage.

As far as motherboard pricing, you can look here to get a better idea of specific prices:
http://www.newegg.com/Product/ProductLi ... 55&x=0&y=0
http://microcenter.com/search/search_re ... rtby=match

Thanks for the review, but I have to agree with RRBOT and gang that disabling turbo boost does not give a fair representation of how these CPUs will be used in the real-world. If anything, I guess we should expect the actual power consumption under load to be a bit higher than reported here.

I also don't really agree with calling turbo "overclocking" since the definition (as I always understood it) is running a chip over its specifications. I don't think that applies here at all, even if the net effect of keeping it on does give you a clock boost if certain conditions are met. I think mentawl hit the nail on the head with the analogy to Speedstep.

CASteve wrote:Now, if only the GPU guys would get their act together and enable similar idle power performance.

Yes, if only...

AZBrandon wrote:... - the Intel chip costs a little more than the 955, ... the Intel chip is a lot faster, it uses less power, ... AMD is in trouble here.

I'm more keen at looking in the other direction:
Phenom II X4 945 is same TDP (95W), slightly faster nominal clock (although less efficiently using that clock and slower than the Turbo boost).
It's also quite a bit cheaper than i5-750 with an even larger price difference given the cheaper motherboards available for socket AM3 compared to socket 1156.

That's where I think an interesting price/power/performance comparison can be made!

Cheers
Olle

Would someone please send me a PM when SPCR review Lynnfield?
I was surprised to see that their Servers have been hacked by an AMD fanboy who published a review of a crippled Lynnfield!

Maybe you need to retire your system administrator as the alternative suggests that your features editor should fall on his sword.

I had to impede the progress of this troll war, but...

Lawrence, can you comment a bit on the 775 vs 1156 socket?

I understand the dimensions are about the same...since Thermalright sells a 1156 bolt thru kit, do you think we can continue to use our socket 775 favorites and just add the bolt thru kit for conversion? Or, doesn't the topside bracket align...

I agree with Vinz and the others in this thread, not using turbo mode in the performance and load measuring tests is completely mis representing these cpu's.. Turbo mode is one of the defining features of the Lynnfield platform. Not considering it leads to false summaries as mentioned by Vinz.

SPCR:
If you already have a fairly high speed quad core setup, there isn't any reason to jump ship to LGA1156.

This is false, the big reason to jump to Lynnfield IS Turbo mode, it makes it so much more efficient by tuning the clockspeed and power distribution based on the code being run and resources being used. No other cpu on sale can do this so effectively. Anandtech stated that this is even enough to improve the user experience, so I'd imagine that, along with the bandwidth and power consumption improvements would be enough to make Core 2 Quad / Original Phenom owners consider an upgrade.

Anandtech:
The turbo mode transitions happen fast enough to accelerate even simple actions like opening a new window. OS and application responsiveness is significantly improved as a result and it's something that you can actually feel when using a Lynnfield machine. It all works so seamlessly, you just always get the best performance you need. It's like Intel crammed the best single, dual and quad-core processors all into one package.
....
Lynnfield's turbo modes change the game. Say goodbye to tradeoffs, the Core i5 and Core i7 are now fast regardless of thread count. It speed that is useful, it speed that you can feel, it's what truly makes Lynnfield the best desktop microprocessor of 2009. It's not just faster, it's smarter, it's better. It's why today's title borrows from Daft Punk and not Star Wars; it's not more of the same, it's something futuristic and new.

Lynnfield shows us the beginning of how all microprocessors are going to be made in the future. Even AMD is embracing turbo, we'll see it with Fusion in 2011. Extend turbo to its logical conclusion and you end up with something very exciting. Imagine a processor made up of many different cores, large and small, CPU and GPU. Each one turning on/off depending on the type of workload, and each running as fast as possible without dissipating more heat than your system can handle.

I agree with the SPCR reviewer that Turbo Mode on the P55/Core i5/Core i7 platform is overclocking, but equally it is a default feature that operates as part of the way the whole hardware platform works. And incidentally the operating system is also involved too. So I thought the review should have included the figures for Turbo Mode both off and on, but simply turning it off - even though this was clearly stated to be the case - produced some potentially misleading data and conclusions. To the extent where I think that the review should be updated with the Turbo Mode on figures.

To its credit SPCR did not indulge in the P55 'previews' that seemed to be everywhere on the web, but clearly there is scope in the future for some of the P55 hardware to be looked at in depth. It is interesting for example that a performance oriented board like the MSI P55-GD80 includes a Green menu in its BIOS. This allows among other things control over both 3 pin and the CPU PWM fans to a degree that I have not seen before. A feature of this board is its 'super cooling' which includes extra large heatpipes and heatsinks. But like every other P55 board its cooling is entirely passive, and it is thanks to SPCR among others that that's the case.

lodestar wrote:

Scoop wrote:It's not like turbo increases power usage too much since it doesn't increase voltage.

The multiplier is locked on the current Lynnfield CPUs so turbo mode can only work by increasing the cpu base frequency which you could call clock speed. You may not agree but by any definition that is overclocking. While voltage may remain constant there is a significant increase in the system power draw, which can only arise from an increase in amps.

So turbo works differently on Lynnfield than Bloomfield? AFAIK, on Bloomfield turbo increases multiplier, you get a one step (133Mhz) increase in frequency. I would suppose it works the same on Lynnfield, like in your following example of 666Mhz is 133Mhz x5.

lodestar wrote:For the 870 for example the increase from turbo mode is 666Mhz, i.e. it rises from 2.93Ghz to 3.59Ghz. You can also of course manually overclock. One reviewer, Guru3D, overclocked an 870 to 4.1Ghz. This increased power draw for their system at idle by 13w, at load it was 133w. So on that basis turbo mode could result in about an 80w increase in system power draw. All these numbers are AC at the wall. Even allowing for that, these are significant figures which cannot be simply ignored.

Obviously overclocking to 4.1Ghz does not happen with default voltage so the numbers here are invalid. Additionally, turbo doesn't increase frequency on the 870 to 4.1Ghz even if you're using only one core. AND, the consumption of one core is not the same as four cores running the same frequency.

lodestar wrote:As to a Lynnfield being a fast Dual Core and a slower Quad Core. Not really. What you really have (with the 8xx Lynnfields) is up to 4 actual cores and up to 4 virtual cores. So faced with a demanding single threaded app a Lynnfield will put 3 of its actual cores into effectively a sleep state, ignore the virtual cores and concentrate system power on just one core. This can include implementing turbo mode, yes on just one core. Depending on the app, it will then bring in further cores up to the actual maximum of 4 when it will start to deploy the virtual cores. If turbo mode is turned on it will implement turbo mode before bringing another core into use.

I don't understand, you just stated why I called it that. Should I have said fast single-core? Running one thread on 870 results in a 3.60Ghz on that single core, bring on another thread and we have a dual-core running at 3.46. 3/4 cores and we have 3.20Ghz. And all of this IF the cooling is sufficient.