• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Why is SNB impressive?

Arachnotronic

Arachnotronic

Lifer
So, I checked out that THG clock-per-clock, core per core comparison, and it looks that SNB is *barely* faster than Nehalem (10%, right?). What was Intel doing for the ~5 years of the development of SNB? Why didn't they just shrink Nehalem and clock it higher? (So don't tell me "oh, SNB overclocks well!")
 
Intel did shrink Nehalem, it's called Westmere. AVX is the big thing in Sandy Bridge, but it's going to be a while before we see software take advantage of it. Lets revisit this thread once that happens k?
 
I said this before SB isnt a big deal, its a glorified die shrunk nehalem.

Not knocking it but its not the leap core 2 was over P4.
 
yottabit said:
people seem to forget that there was a 'core' in between p4 and core 2. Core solos are miserable
Click to expand...

Nobody forgets that. The core architecture was a mobile platform, not a desktop platform. Even then, Core 2 had some pretty heafty gains over Core 1.
 
yottabit said:
people seem to forget that there was a 'core' in between p4 and core 2. Core solos are miserable
Click to expand...

Which, BTW, only lasted around 5 months before it was replaced by Core 2. That's why people don't really remember. And the Core Duos were decent. The problem they had was very low clock speeds and that they were 32-bit.

That, and Core Duo and Solo (microprocessor, not microarchitecture) was only on mobile devices (laptops).
 
Intel17 said:
So, I checked out that THG clock-per-clock, core per core comparison, and it looks that SNB is *barely* faster than Nehalem (10%, right?). What was Intel doing for the ~5 years of the development of SNB? Why didn't they just shrink Nehalem and clock it higher? (So don't tell me "oh, SNB overclocks well!")
Click to expand...

Lets not forget that SNB has an IGP. The core architecture itself only has incremental improvements (unlike the big changes brought by C2D/Nehalem), but there were certainly efforts put into graphics integration.

The other big changes like AVX are helpful for the programs which are optimized for it but won't show in many benchmarks.
 
GammaLaser said:
Lets not forget that SNB has an IGP. The core architecture itself only has incremental improvements (unlike the big changes brought by C2D/Nehalem), but there were certainly efforts put into graphics integration.

The other big changes like AVX are big for the programs which are optimized for it but won't show in many benchmarks.
Click to expand...

Core is a microarchitecture. Core 2 is a microprocessor line. Core is also a microprocessor line. Why is Intel's naming scheme so confusing?

The big improvement in Sandy Bridge is over-clocking headroom. We've gone from average 4GHz OCs to 4.5-4.6GHz.
 
Intel17 said:
So, I checked out that THG clock-per-clock, core per core comparison, and it looks that SNB is *barely* faster than Nehalem (10%, right?). What was Intel doing for the ~5 years of the development of SNB? Why didn't they just shrink Nehalem and clock it higher? (So don't tell me "oh, SNB overclocks well!")
Click to expand...

It's a modest improvement in performance, power consumption and clockability but these things all add up. It's the most powerful single x86 core ever made, but it's also the most power-efficient. It's has high overclocking limits while remaining at reasonable power consumption. It has both exceptional single-threaded and multi-threaded performance; CPU wise, it has no weaknesses and is strong in everything.

For all the talk of Bulldozer giving nearly two cores of throughput with a design that isn't much larger than a single core, Sandy Bridge is already there, at least compared with current AMD cores.
 
yottabit said:
people seem to forget that there was a 'core' in between p4 and core 2. Core solos are miserable
Click to expand...

What was so miserable about it? Core Duo/Solo were based on the Pentium M, which was a kick ass CPU. It had poor floating point performance but for some applications it had excellent performance, not to mention it put P4 and K8 to shame in terms of power usage.

There was a desktop motherboard with an adapter that let you use a Pentium M CPU. One site benchmarked the Pentium M overclocked to 2.4 GHz. It actually whupped the P4EE and came very close to the A64FX in a lot of gaming benchmarks.

Here I found the link: http://techreport.com/articles.x/7927/1
 
Last edited:
Intel17 said:
So, I checked out that THG clock-per-clock, core per core comparison, and it looks that SNB is *barely* faster than Nehalem (10%, right?).
Click to expand...

It is faster while drawing less power. See below.

Maximilian said:
its not the leap core 2 was over P4.
Click to expand...

That's a bad comparison because Netburst (P4) was intentionally long pipeline (lower IPC) to try for high clock speeds. Thus, of course if you compare it MHz to MHz it loses badly. However, at the launch of Conroe the fastest P4 was over a GHz higher clocked than the highest clocked Conroe.

Compare that to Sandy Bridge, which at launch is already higher clocked than Nehalem, plus it draws less power. I'm sure if you measure how much work per Watt consumed, it would compare very favorably.
 
Zap said:
It is faster while drawing less power. See below.



That's a bad comparison because Netburst (P4) was intentionally long pipeline (lower IPC) to try for high clock speeds. Thus, of course if you compare it MHz to MHz it loses badly. However, at the launch of Conroe the fastest P4 was over a GHz higher clocked than the highest clocked Conroe.

Compare that to Sandy Bridge, which at launch is already higher clocked than Nehalem, plus it draws less power. I'm sure if you measure how much work per Watt consumed, it would compare very favorably.
Click to expand...

Hmm, I don't think that's accurate. The fastest Conroe at launch, the X6800, was clocked at 2.93GHz. The fastest Pentium 4, the EE 3.73, was clocked at... 3.73GHz. Even then, the performance difference between both was astronomical.

Anyway, by the time Ivy Bridge comes out in Q1 2012, we'll probably be at similar clock speeds to the "fastest" Pentium 4.
 
Well -- I missed the leap from Wolfdale/Yorkfield to Nehalem-I7.

So I can't make any comparison at all with the generation I missed.

But perhaps I chose to miss it because I was waiting for this improvement -- which some other poster insinuated to be "incremental." It may be. But I have to say -- Sandy Bridge "K" processors are "rich."

And -- see -- I wouldn't say it's just about the cores. Even for the fact that the core includes iGPU. Putting the memory controller on the processor -- that was one of the biggest factors with Nehalem. I'm guessing that the people at Intel probably were dropping in to these forums -- here and at other places -- just to think about "how to do it better."

So we have an entire menu of improvements. There's Sandy Bridge -- even for being meant for "mid-range" "mainstream" consumers (as opposed to "extreme" enthusiasts.) there's the Z68 chipset -- not to dliminish the P67/H67 that preceded it. But all this stuff adds up:

-- Several hundred Mhz in OC'ing headroom, at low voltages
-- This (new to me, anyway) "Turbo-Mode" OC-ing
-- bottlenecks opened wide with things like ISRT
-- the prospect of utilizing iGPU and dGPU together . . . and this LucidLogix-Virtu thing . . .

I just resolved my stupid mystery -- if I wasn't as old as I am, I'd try to hide it. I got my 2600K and Z68 system up and running . . . . got the ISRT to work with the OS installed. Kept going into BIOS and wondering if I'd been Rip Van Winkle asleep for 30 years, or why the old BIOS' of LGA 775 seemed like Model T's when compared to the Star-Trek Generations of Z68. And thinking "I can figure this out . . . I can figure this out. . . . "

Finally got all the peripheral voltages set -- for safety. I'd set the "Turbo Ratio" or multiplier -- even just to 40 -- didn't care about bumping up the bCLCK to 103+. Every time I'd save and re-boot, Windows CPU-Z and AI-Suite would tell me "3.4 Ghz." Or "3.5 Ghz with the bCLCK at 103 . . ."

Three days . . . . reading that stupid (as always, with ASUS) manual, looking at forum posts . . .

Then today, I finally opened the ASUS monitor and PRIME-95 simultaneously. Apparently, even the recent-last CPU-Z doesn't tell you what you need to know other than current VCORE. And . . . I get it . . . "Turbo" mode . . . . 4.4 Ghz.

And this is just . . . preliminary . . . for my conservative inclinations about OC'ing.

Almost takes the fun out of it.
 
There's not much out there to push Intel to try for huge performance gains, AMD is so far behind and keeps pushing their architecture change back.

These CPUs are faster than most people news, which doesn't help in getting them to push the performance envelope.
 
LOL_Wut_Axel said:
Hmm, I don't think that's accurate. The fastest Conroe at launch, the X6800, was clocked at 2.93GHz. The fastest Pentium 4, the EE 3.73, was clocked at... 3.73GHz.
Click to expand...

Who buys EE chips. :\

I just checked and Intel lists a non-EE 4GHz P4 with 2MB cache (double the normal amount) on a 1066MHz FSB (also higher than normal) while the fastest non-EE Core 2 was the E6700 at 2.66GHz.
 
Zap said:
Who buys EE chips. :\

I just checked and Intel lists a non-EE 4GHz P4 with 2MB cache (double the normal amount) on a 1066MHz FSB (also higher than normal) while the fastest non-EE Core 2 was the E6700 at 2.66GHz.
Click to expand...

Hmm, I can't find a P4 being launched with 4 GHz or higher stock frequency, but there is the P4 570 and P4 670 at 3.8 GHz.
 
Last edited:
nickb64 said:
There's not much out there to push Intel to try for huge performance gains, AMD is so far behind and keeps pushing their architecture change back.

These CPUs are faster than most people news, which doesn't help in getting them to push the performance envelope.
Click to expand...

Hmm . . . seemed like a few years ago, AMD was "ahead." It's a duopoly, no formal barriers to entry but an entire regime of usage, software and customer-base that insures to make it harder for a third entrant.

Given that, they have enough returns on their product to keep up their R&D effort. And if people don't need or even perceive extreme performance, or they don't perceive a need for it, both companies can just saunter along . . . Even the duopolist is driven by the customer-base . . .

Spoke to an engineer for Intel up in Tacoma about six years ago -- who was leaving that particular career behind. He was saying, all incredulous and wearisome about it -- "It's getting down to the molecular level!!"

Figure they have other envelopes to push. The latest generation of "everything" has sort of pushed the "reliability" envelope.
 
BonzaiDuck said:
It's a duopoly, no formal barriers to entry but an entire regime of usage, software and customer-base that insures to make it harder for a third entrant.
Click to expand...

You mean like an x86 license?

I think there are significant barriers to entry for the CPU market, maybe the most significant of any industry...
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top