How much better is a Broadwell i3 over a haswell for dsktops that is?

[john5220]

Hi currently I find the performance of an i3 to be lacking in haswell department. Its not worth an upgrade if it barely matches a FX 4300 AMD in multi threading.

So i was hoping for the broadwell i3 being better?

If a broadwell i3 could match a 2300 core i5 sandy bridge I would be all good with that.

Broadwell will fit in 1150 socket right? note I am not looking for any GPU performance only CPU

My aim is to run DayZ stand alone

 

[Roland00Address]

Broadwell I3 will only be up to 5% faster than Haswell I3 at the same clock frequency more or less. It is a shrink not a newer architecture.

 

[john5220]

Jesus what a waste. No competition since AMD is in the dumpster and intel decided to keep selling trash for many years to come.

 

[Yuriman]

Jesus what a waste. No competition since AMD is in the dumpster and intel decided to keep selling trash for many years to come.

Why not buy an i5 or i7 if you're unhappy with your i3?

Broadwell is bringing major improvements to the mobile space. Performance on desktop has been going up consistently too, despite power usage dropping with every generation. Raw performance is not the focus of either company now.

 

[poohbear]

Jesus what a waste. No competition since AMD is in the dumpster and intel decided to keep selling trash for many years to come.

Not trash, theyre great cpus for what they do. Like the ps4 or xbox 1 have better processors? If u want a beefy cpu, get an i7. Coming on here & saying an i3 is trash?!? Lol

 

[ShintaiDK]

No such thing as a Broadwell i3 for desktop. There will only be Broadwell-K for desktop. Else its Skylake i3 (LGA1151.).

 

[PowerYoga]

uses i3, complains about performance.

wut

 

[sm625]

Broadwell i3 will be trash because its the same as haswell, only shrunk even more. A haswell i3 is already ridiculously tiny, at least the cpu part anyway. Broadwell will be a total joke. Most competitors will be selling their chips for 10 times less (per xtor). This is just too greedy to be considered good. A broadwell i3 should have 6MB of cache. Even then it would still be a tiny chip to be asking over $100 for.

 

[BSim500]

A haswell i3 is already ridiculously tiny, at least the cpu part anyway.

You should see the Core-M (a dual-core chip on a 5-core die...) I'm sure they could (if they wanted) easily release an iGPU-less "PK" unlocked quad-core for gamers with dGPU's (a bit like the i5-3350P but "no GFX" rather than "fused off GFX") in the same die space as an i3. And couldn't they fit (almost) 6x cores in the same space as a quad iGPU-less Haswell?

I guess the question needs to be asked is How 'free' is today's 'free' iGPU in terms of how many additional cores & cache could be added in the same space / what savings could be passed on with a smaller die? I'm all for improvements, but I really don't see the point in forcing GT3 across the board. One massive waste for most non-gamers (even a GT1 would suffice for office / netbox use) plus gamers with dGPU's alike.

 

[ShintaiDK]

You should see the Core-M (a dual-core chip on a 5-core die...) I'm sure they could (if they wanted) easily release an iGPU-less "PK" unlocked quad-core for gamers with dGPU's (a bit like the i5-3350P but "no GFX" rather than "fused off GFX") in the same die space as an i3. And couldn't they fit (almost) 6x cores in the same space as a quad iGPU-less Haswell?

I guess the question needs to be asked is How 'free' is today's 'free' iGPU in terms of how many additional cores & cache could be added in the same space / what savings could be passed on with a smaller die? I'm all for improvements, but I really don't see the point in forcing GT3 across the board. One massive waste for most non-gamers (even a GT1 would suffice for office / netbox use) plus gamers with dGPU's alike.

You can start by asking yourself how many more cores you can add under a 84W TDP without reducing speed. To cut it short, the answer is 0. But even if we go far we can say 1 core. On the other hand it will also increase thermal density giving you yet another problem.

And again, if you dont want the IGP. Then there is a purchase option for that too on LGA2011-3.

 

[Yuriman]

You can start by asking yourself how many more cores you can add under a 84W TDP without reducing speed. To cut it short, the answer is 0. But even if we go far we can say 1 core. On the other hand it will also increase thermal density giving you yet another problem.

Excellent point. TDPs would go up *a lot* if you tried to fill a 200mm^2 chip with high clocked cores.

 

[AtenRa]

Deactivate Hyper-Threading and you can double the cores from 22nm going to 14nm keeping the same TDP and clocks

ahhh, add a little bit more L3 cache and you are set

 

[BSim500]

You can start by asking yourself how many more cores you can add under a 84W TDP without reducing speed.

I thought 14nm 4-core Broadwell-K's TDP got reduced to 65w? That's 16w per core x 6 cores = 96w or about the same as an i5-2500K... The only reason Haswell runs so hot is due to a combination of that "gap" between the TIM & heat-spreader combined with the FIVR's auto-100mv-overvolt-on-AVX. Since the latter doesn't apply to most games / software (and may be removed anyway from what I heard), and I'm pretty sure Intel could improve the "gluing" process on the former (assuming there's some genuine technical reason why they can't solder it which would outright cure the problem), I really don't see a problem with Sandy Bridge TDP levels. Haswell gets hot at ANY given TDP - ie, even "55w" i3's have hit 100c due to bad gluing whilst others don't even breach 55c).

Another way of looking at it : 84w (22nm) -> 65w (14nm) = a 25% reduction in TDP from the new process. So a 6-core 20M cache Xeon E5-2643 (135w on 22nm) = barely 100w on the new process - or again within 5% of a 2500K. And that includes HT plus a ton of cache. If 14nm turns out like that in practise, I think they could get a full speed hex-core down into the 90's (TDP w).

 

[ShintaiDK]

I thought 14nm 4-core Broadwell-K's TDP got reduced to 65w? That's 16w per core x 6 cores = 96w or about the same as an i5-2500K... The only reason Haswell runs so hot is due to a combination of that "gap" between the TIM & heat-spreader combined with the FIVR's auto-100mv-overvolt-on-AVX. Since the latter doesn't apply to most games / software (and may be removed anyway from what I heard), and I'm pretty sure Intel could improve the "gluing" process on the former (assuming there's some genuine technical reason why they can't solder it which would outright cure the problem), I really don't see a problem with Sandy Bridge TDP levels. Haswell gets hot at ANY given TDP - ie, even "55w" i3's have hit 100c due to bad gluing whilst others don't even breach 55c).

Another way of looking at it : 84w (22nm) -> 65w (14nm) = a 25% reduction in TDP from the new process. So a 6-core 20M cache Xeon E5-2643 (135w on 22nm) = barely 100w on the new process - or again within 5% of a 2500K. And that includes HT plus a ton of cache. If 14nm turns out like that in practise, I think they could get a full speed hex-core down into the 90's (TDP w).

Feel free to provide the links that broadwell-K is 65W.

And we are not talking about a shrink either. But about IGP estate.

The 100mv for 256bit AVX got nothing to do with the FVIR. It would be exactly the same without. And remember throughput is almost doubled.

Haswell only get hot when its throughput is almost twice of SB/IB.

 

[BSim500]

Feel free to provide the links that broadwell-K is 65W.

http://www.fudzilla.com/home/item/35391-desktop-broadwell-lga-is-socket-1150

 

[ShintaiDK]

http://www.fudzilla.com/home/item/35391-desktop-broadwell-lga-is-socket-1150

So no credible link? Its quite obviously from the i3 and Pentium part that he doesnt have a clue.

I didnt know people even read FUD anymore.

 

[Roland00Address]

I guess the question needs to be asked is How 'free' is today's 'free' iGPU in terms of how many additional cores & cache could be added in the same space / what savings could be passed on with a smaller die? I'm all for improvements, but I really don't see the point in forcing GT3 across the board. One massive waste for most non-gamers (even a GT1 would suffice for office / netbox use) plus gamers with dGPU's alike.

You assumed Intel wants to sell you dollars per transistor. Intel does not want to do this. They want to sell you dollars per performance and by limiting the core count they can cause people who need much higher amounts of cores since there work is highly parallel to pay more and move up to things like xeons, xeon-phis and such.

On normal workloads that most "humans" do on their personal computers do you really need more than 8 threads?

Intel purposefully wants to keep the consumer market and the industrial market seperate for the industrial market will pay thousands of dollars per cpu while customers do not want to pay more for a cpu core than a bottle of wine or a couple cups of starbucks.

 

[BSim500]

So no credible link? Its quite obviously from the i3 and Pentium part that he doesnt have a clue.

I didnt know people even read FUD anymore.

We won't know for sure until it gets released. Obviously. Considering the last process change was a 20% improvement (32nm 95w Sandy -> 22nm 77w Ivy), and from comments on leakage in Anandtech's article, I personally don't find 65-70w TDP quad-cores that "outrageous". Remember - i5's were already down to 77w on 22nm Ivy Bridge. And it's entirely possible Intel has tweaked how much over-voltage AVX gets fed in future.

 

[Lepton87]

We won't know for sure until it gets released. Obviously. Considering the last process change was a 20% improvement (32nm 95w Sandy -> 22nm 77w Ivy), and from comments on leakage in Anandtech's article, I personally don't find 65-70w TDP quad-cores that "outrageous". Remember - i5's were already down to 77w on 22nm Ivy Bridge. And it's entirely possible Intel has tweaked how much over-voltage AVX gets fed in future.

65-70W TDP quads are indeed nothing outrageous. My 22nm IVY BRidgr 3630QM is only 45W and it still effectively runs at 3.2GHz under all cores load. I know it's mobile but it shows what can be done.

 

[john5220]

Any chance of broadwell i3 being 10% faster than current i3?

What about its price?

 

[crashtech]

You should just sell your stuff and switch to AMD.

 

[Yuriman]

Historically, Intel has priced their new chips about the same as the ones they're replacing, and just phased the old ones out. I expect the new i3's to draw significantly less power and be clocked a bit higher, so you may be looking at 10% total from IPC and clocks.

My aim is to run DayZ stand alone

As I understand it, DayZ is built on the source engine and should scale pretty similarly to ARMA3.

^ I don't recommend "upgrading" to an FX for DayZ. A Haswell i3 is often as much as 30% faster than the i3 2100 in that list. Terrible value, huh.

 

[john5220]

^ yeah the more benchmarks I see its becoming very apparent the i3 is a way better buy than the FX 6300
And besides calling a i3 a dual core is kinda unfair, because its Hyper Threading actually does performance and actions of an actual core.
For example 64 player maps in battlefield 3 and 4 would jump from 12 fps min Phenom II X2 dual core to about 60 FPS on a intel i3

And would jump from 30 FPS min on pentium G to 60+ fps on i3 ( I don't use average or max I prefer use min). So it seems i3 is not a dual core at all its a quad core because BF3 runs fine but it needs 4 cores to do processing.

Infact from the benchmark you posted on my other thread, even though it dont have a pentium, its clear from the i3 results for advanced warfare that it blows out the FX 6300 and I from a estimation my Pentium G would actually match or beat the FX 6300

I had no idea intel was this good. AMD performance is really bad in lots of games I mean I am sure advanced warfare supports 4 cores yet it takes a FX 6300

I guess its intel's architecture? its just overall better than AMD

 

[Yuriman]

Some basic, theoretical maths:

1 Haswell core = 1.6 Piledriver cores
1 Hyperthread "core" = ~0.25-0.30 Haswell cores

Load up two Piledriver cores, and you lose 5-20% performance, so:
2 Piledriver cores in the same module = ~1.6-1.9 single Piledriver cores

1 Haswell hyperthreaded core ~ 60% of a loaded Piledriver module's core

~~
In terms of raw total theoretical processing power:
i3 vs FX-6300
1 thread, i3 has a 60% advantage
2 threads, i3 has a 60% advantege
3 threads, i3 has a 23% advantage
4 threads, i3 has a 16% advantage
5 threads, FX-6300 has a 1% advantage
6 threads, FX-6300 has a 15% advantage

Most games don't distribute their load evenly over all threads, though, so in those cases where you might have 6 or even 8 threads that can be taken advantage of, an i3 may still beat an FX-6300 or even an 8350 because it'll be waiting on one thread. In games that distribute the load very evenly over lots of threads, the FX chips excel and provide an excellent value.

Non-gaming workloads (scientific computing) often scale very well on FX-chips.

 

[john5220]

^ hey thanks can I ask you a question?

https://forum-en.msi.com/index.php?topic=182658.0

On that thread it says you can OC the pentium G 3258 to 4.4 GHZ with the stock cooler that comes with the intel CPU and the one that also comes with the G3258 has a copper center in the stock heat sink they say its the same heatsink that comes with the i7 CPU also

This is my mainboard btw

http://www.amazon.com/ECS-Elitegroup...mainboard+1150

So if I buy that pentium G can I then reach 4.3 or 4.4 ghz with stock cooler and my mainboard?

if you say yes I would jump to the ceiling for joy I could end up just buying that chip and sell my current G3220