Intel 2013-2014 roadmap

[csbin]

https://twitter.com/asder00


Core i7/Core i5

Core i3、Pentium、Celeron

chipset

 

[RadiclDreamer]

Any idea what the 4771 will bring over the 4770?

 

[mavere]

Maybe Iris Pro?

 

[zir_blazer]

Any idea what the 4771 will bring over the 4770?

Two choices:

Faster GPU, MAYBE at the cost of features (Check Nehalem Core i5 660 vs 661, you have 166 MHz more on GPU at the cost of 14W higher TDP, and losing vPro and TXT.

A Intel Upgrade Services part.

 

[ShintaiDK]

Any idea what the 4771 will bring over the 4770?

+100Mhz baseclock is my bet.

 

[TazFTW]

This image is with them.

 

[zir_blazer]

A boring 100 MHz increase? Why the hell they didn't called it simply Core i7 4780 to be consistent with nearly all other models?

 

[ShintaiDK]

A boring 100 MHz increase? Why the hell they didn't called it simply Core i7 4780 to be consistent with nearly all other models?

Due to the 4770K.

 

[inf64]

It's not even 100MHz over 4770K, check the table TazFTW posted above. "New" 4771 brings what 4770 non-K supports today and that's VT-d, TXT and maybe TSX (in 4771)- adding 100Mhz over non-k's base clock is not going to be felt in any workload as it's ~3% difference while max Turbo is the same. The results will be identical in most applications. What users need is 4770K with not fused off ISA extensions.

 

[LogOver]

A boring 100 MHz increase? Why the hell they didn't called it simply Core i7 4780 to be consistent with nearly all other models?

At least "+1" to the model number for "+100MHz" to the base clock is much more sensible than "+1060" to the model number in exchange for "+700MHz" base clock.

 

[Abwx]

At least "+1" to the model number for "+100MHz" to the base clock is much more sensible than "+1060" to the model number in exchange for "+700MHz" base clock.

Obssessed by AMD....
Whenever one wants to downplay an intel s lackness
and need to trash AMD for this is just showing how
hatefull and biaised he is...

 

[inf64]

@ LogOver
This is about intel's roadmap and not about AMD model numbers. Please do not flame bait others.

 

[Fjodor2001]

So going from Sandy Bridge (January 2011) to Haswell Refresh (Q2 2014) we will only have seen something like 8% (SB->IB) + 8% (IB->Haswell) + 3% (Haswell Refresh) = 19% performance increase in more than 3 years? 🙁

Is there even anything on the roadmap further ahead that is likely to bring substantial performance improvements? Broadwell seems to be mostly about lower TDP and better IGP again (like Haswell)? The details on Skylake are not that well known though, maybe it will bring some CPU performance improvements? Or is the Desktop CPU development basically "completed" now as far as CPU performance goes. This is more or less as good as it gets performance-wise. We just have to get used to it?

 

[ShintaiDK]

So going from Sandy Bridge (January 2011) to Haswell Refresh (Q2 2014) we will only have seen something like 8% (SB->IB) + 8% (IB->Haswell) + 3% (Haswell Refresh) = 19% performance increase in more than 3 years? 🙁

Is there even anything on the roadmap further ahead that is likely to bring substantial performance improvements? Broadwell seems to be mostly about lower TDP and better IGP again (like Haswell)? The details on Skylake are not that well known though, maybe it will bring some CPU performance improvements? Or is the Desktop CPU development basically "completed" now as far as CPU performance goes. This is more or less as good as it gets performance-wise. We just have to get used to it?

How did you get Haswell refresh to 3%? Also you cant just say 8+8+3 and get 19% (assuming the numbers you came with.). It would be just over 20%.

New instructions is the way to higher performance. Over time the IPC for IB vs Haswell will only increase in the favour of Haswell.

 

[Fjodor2001]

How did you get Haswell refresh to 3%?

100 Mhz / 3400 Mhz => 3%. Assuming the 4771 is just a 100 Mhz CPU frequency increase over the 4770.

Also you cant just say 8+8+3 and get 19% (assuming the numbers you came with.). It would be just over 20%.

Fair enough. It should actually be 1.08*1.08*1.03 => 20.1%. But since the difference was so small I didn't bother being exact in my previous comment. It was just an approximation anyway.

New instructions is the way to higher performance. Over time the IPC for IB vs Haswell will only increase in the favour of Haswell.

Let's hope so. But we have unfortunately been disappointed before with the real life benefits of new instructions that have been introduced.

 

[ShintaiDK]

100 Mhz / 3400 Mhz => 3%. Assuming the 4771 is just a 100 Mhz CPU frequency increase over the 4770.

We already got the K model at that speed. So lets wait and see what the Haswell refresh is, before basing anything on it.

Let's hope so. But we have unfortunately been disappointed before with the real life benefits of new instructions that have been introduced.

Using legacy code perhaps. Try Linpack on a Nehalem vs SB vs Haswell and tell me what you get.

 

[Fjodor2001]

Try Linpack on a Nehalem vs SB vs Haswell and tell me what you get.

Most people don't use their computer to run Linpack. They are interested in real life performance increases... 😉

 

[ShintaiDK]

Most people don't use their computer to run Linpack. They are interested in real life performance increases... 😉

It was just an example. As long as you keep using legacy code for baseline, you will always be dissapointed.

UT4 engine for example will use new instructions.

 

[OBLAMA2009]

haswell is gonna be around a while, itll be a good upgrade once the corrected chips are out...

 

[Shmee]

What do we know so far about IB-E? Any ES OC results yet? Do we know how many cores? Any other upgrades?

 

[Edrick]

I want a 4770 with the L4 cache!!!!

 

[ArizonaSteve]

More mobile dross no doubt. I'll come back in 2020...

 

[inf64]

haswell is gonna be around a while, itll be a good upgrade once the corrected chips are out...

What corrected chips? There is nothing wrong with Haswell so nothing to correct (technically speaking there is always some errata but this is it when it comes to performance). Maybe if you are referring to IHS<->die contact, there is a room for improvement.

 

[uribag]

What corrected chips? There is nothing wrong with Haswell so nothing to correct (technically speaking there is always some errata but this is it when it comes to performance). Maybe if you are referring to IHS<->die contact, there is a room for improvement.

I think he is talking about the USB3 issue.

 

[inf64]

USB3 has nothing to do with Haswell cores but with a supporting chipsets(motherboards). This issue is actually non-issue and intel already started shipping new stepping chipsets that have no such problems (however rare they might be).