Sandy bridge vs haswell, 30% better clock per clock?

[Makaveli]

Another thought - I did come from P8P67 (not Z77) boards to Z87s on some of these. Obviously the fast boot is faster, but could the Z87 (2 generations newer) new RST features (accelerated storage) make a difference?

I think that is possible.

Thinks like this are hard to judge there are no hard metrics to base it on.

 

[el etro]

Not really, but is like ~15% more clock per clock. Is very good for a 2-year update.

 

[mindbomb]

the big difference is that haswell has improved hyperthreading and supports avx2. If your program is optimized well for avx2 and hyperthreading, you'll see the biggest difference.

 

[Ertaz]

So is a Intel Pentium G3220 3GHz Haswell Dual-Core Processor faster than an i7-920?

 

[dragantoe]

So is a Intel Pentium G3220 3GHz Haswell Dual-Core Processor faster than an i7-920?

stock speeds easily

 

[Concillian]

stock speeds easily

I don't think this is the case:
http://anandtech.com/bench/product/677?vs=47

Ivy i3 still gets beat by a small margin in most multithreaded benchmarks, a Pentium would fare even worse.

Probably a ~3.5 GHz dual haswell with HT at least until it starts winning out in MOST multithreaded benchmarks.

Also not sure why 'el etro' bumped a several month old thread here.

 

[nwo]

So is a Intel Pentium G3220 3GHz Haswell Dual-Core Processor faster than an i7-920?

Faster in what?

Haswell Pentium will be faster in single threaded performance but certainly not faster in multi threaded performance where the i7 920 will have a massive advantage since it is 4 cores and 8 threads versus Pentium's 2 cores and 2 threads.

Sandy bridge is roughly 30% slower clock per clock compared to Haswell in multithreaded applications. However, the majority of gaming benchmarks do not reflect anywhere near those numbers. The higher the resolution, the less the CPU makes an impact on your fps. If you are talking about 1440p, I would expect less than 5% performance difference between Haswell and Sandy at the same clock speed.

 

[ShintaiDK]

Single and multithreaded is terrible terms when used with CPUs with multiple cores. The G3220 will certainly be faster in 1 and 2 threads. And possible 3. But is also depends on the scaling factor of the application/game. It doesnt matter much if 4 cores is used, and it only offers 10% over 2 cores.

 

[nwo]

Single threaded performance usually refers to apps that run off of one thread, or 1 core vs. 1 core performance. Multi-threaded refers to applications which can fully utilize all the threads you throw at it.

I'll leave it at that and let this comparison do the rest:
http://www.cpu-world.com/Compare/254/Intel_Core_i7_i7-920_vs_Intel_Pentium_Dual-Core_G3220.html

 

[BonzaiDuck]

I can't tell how much the OP has exploited his OC'ing potential. Maybe I missed something in the posts here.

I'd been very cautious about voltage and diligent about temperatures as I tried to pick up a Centigrade here and grain of rice there. Now, I've given my i7-2600K a shakedown test sequence, hopefully ironed out some shortcomings and tweaked the BIOS some more after finding another sweet-spot at 4.7 Ghz. I think I'm ready to shoot for 4.8, but I don't think it's going to be a 24/7 setting: My best estimate has the load vCORE bumping up too close to 1.4V.

This being said, one could become more casual about pushing V's and Mhz a little higher, knowing that you're not likely to do too much damage and that you "eventually" plan on upgrading to some newer configuration.

The one thing that impressed me, noted in my posts to some other similar threads, is the Cinebench 15 results I got in reaction to the general call to post scores, as made in another thread several weeks ago. On that one, single benchmark, I'm neck-and-neck with an i7-4770K @ 4.4Ghz. Another survey conducted here recently shows that 4.4 is about the most probable preferred clock for a Haswell K.

You can't discount instruction-set enhancements and other features of Haswell performance at seemingly lower clocks. But any difference just doesn't seem worth it now from a sane, benefits-v-costs, practical perspective. It would be different if you had an LGA-775 C2Q, or maybe even a low-end, OC'd Nehalem. A "mainstreamer" with a five year old system might have a hardware failure in need of total replacement, and there I would say "Go for it! Get a Haswell system!"

Might as well "consume" your SB-K as you eyeball technical development. Personally, I'm planning now to build a Haswell-E. And for some time after the fall release, I'm going to take a lot of time thinking about it before I make final parts selection and place my orders. I might even wait until some months into 2015.

I can only say that my SB 2600K with Z68 is the best darn system I've ever had.

 

[Headfoot]

Single threaded performance usually refers to apps that run off of one thread, or 1 core vs. 1 core performance. Multi-threaded refers to applications which can fully utilize all the threads you throw at it.

I'll leave it at that and let this comparison do the rest:
http://www.cpu-world.com/Compare/254/Intel_Core_i7_i7-920_vs_Intel_Pentium_Dual-Core_G3220.html

That's conflating terms. Single threaded is a single thread. Multi-threaded is any number of threads greater than one. It does not mean it scales infinitely, that is an entirely different thing, commonly called "embarrassingly parallel." A program with 80% of the work falling into thread 1 and 20% of the work falling into thread 2 is multi-threaded. A program with 25% falling evenly onto 4 threads is also multi-threaded. There is no commonly used single term for this as far as I know. People usually just say "this program is more parallel than that one."

Each program is going to have its own unique thread usage patterns and thus the importance of single thread performance vs number of available threads is going to be a sliding scale, and not binary

 

[tweakboy]

Hey all, getting bored with my 2500k @ 4.5ghz, is haswell 30% better performance wise at the same clock speeds? So 4.5ghz haswell would be 30% faster? Is that a correct assesment? Im asking for gaming @ 2560x1600 resolution. Talking about the 4 core i5 haswells here. Thanks in advance!

Your wrong sir, its 50 percent gain. NOT!!!!!!!!!!!! There is no real world difference, 5 to 10 percent if that.

 

[cytg111]

You will not get more than 5% on average at 1440p. Better get more money and get a 6-core Haswell next year

argh .. my eyes decieve me, surely you mean 8 core haswell

 

[BonzaiDuck]

argh .. my eyes decieve me, surely you mean 8 core haswell

It's my understanding that they won't offer a 4-core E chip like the IB-E i7-4820K Skt-2011. It is likely the Haswell-E will come in two flavors: 8-core and 6-core. . . . Unless I misread something about that. . .

 

[BrightCandle]

Doesn't help for gaming but there is evidence at Phoronix and a few other places that Haswell is pretty significant for developers. Compilation has a lot of branches and the increased ports and prediction cache really help compilers out with more speculation resources and better prediction. I have been seeing 20-30% been used which is quite a bit, but about the best we should expect from Haswell compared to IB.

 

[Concillian]

I can't tell how much the OP has exploited his OC'ing potential. Maybe I missed something in the posts here.

Don't worry too much about the OP. He started this thread like 4-5 months ago. Someone recently bumped it.

I think the OP got what he needed from the thread a few months ago and has done whatever he's going to do.

 

[BonzaiDuck]

Don't worry too much about the OP. He started this thread like 4-5 months ago. Someone recently bumped it.

I think the OP got what he needed from the thread a few months ago and has done whatever he's going to do.

This seems to be a volatile topic spread across a few threads the address it. I'm more concerned that the desktop PC may go the way of the steamboat, and I'll just become some mainstreamer-come-lately to mobile devices and the burden of extra wireless account subscriptions.

 

[Makaveli]

This seems to be a volatile topic spread across a few threads the address it. I'm more concerned that the desktop PC may go the way of the steamboat, and I'll just become some mainstreamer-come-lately to mobile devices and the burden of extra wireless account subscriptions.

Blasphemy !!

 

[Torn Mind]

A 30% increase in two generations would require require Ivy Bridge to be 1.15 times faster than Sandy Bridge and Haswell to be just a hair under 1.13 times faster than Ivy Bridge or some other combination of performance gains that result in a 30% increase over two generations. .

 

[The|Hunter]

Well in Cinebench15 single threaded benchmark

my 4770k @ 4.7ghz, ram 2133mhz acted as 4930k @ 5Ghz 2133mhz quad channel ram, we both got 188cb.

 

[sxr7171]

IMHO Ivy is the sweet spot. Better IPC than sandy and I get 4.8GHz on a corsair H80. I think it would be hard to get that performance from haswell without delidding. I may be wrong.

 

[BonzaiDuck]

IMHO Ivy is the sweet spot. Better IPC than sandy and I get 4.8GHz on a corsair H80. I think it would be hard to get that performance from haswell without delidding. I may be wrong.

Did you de-lid your Ivy Bridge? What you say about the processor may well be true except for their choice of TIM-goop instead of solder with the non-E processors. The only thing I can say again (with my SB 2600K): You may get high clocks; you may get Cinebench scores close to the Haswell 4770K; but there are enhancements to the instruction set and other features that you don't get with the earlier processor.

That being said: I still love my 2600K. As for your IB, that's pretty darn good -- 4.8Ghz with a corsair H80. My recollection of comparison reviews seemed to show my D14 falling short by about 0.9C from the H80, and I know that replacing the two Noctua fans with a single Akasa 140R gives me a 5C advantage over the review benchmarks.

 

[piasabird]

I think part of the problem is that when you do a fresh install or an OS for a while everything seems a little faster just because there is no junk on your computer yet.

 

[pw257008]

This seems to be a volatile topic spread across a few threads the address it. I'm more concerned that the desktop PC may go the way of the steamboat, and I'll just become some mainstreamer-come-lately to mobile devices and the burden of extra wireless account subscriptions.

so, a topic fit more for an i7-920 than a Pentium G3220?

 

[BonzaiDuck]

so, a topic fit more for an i7-920 than a Pentium G3220?

Not sure exactly what you mean by that. Sure -- if you had an i7-920 and wondered about upgrading, it's been maybe at least four years -- whenever socket-1366 was released.

Are we concerned about what to do with upgrade possibilities for an $80 Haswell? Can't say. If it turns out the $80 Haswell falls short in some way, it's the user's fault for not picking what he needed from the processor line.

I can see upgrading a bunch of C2D Wolfdales with some of the more modest Haswell offerings. But even that is all about "want" versus "need" versus budget and common-sense.