Upgrade to Broadwell or wait?

dogbyte_13

[H]ard|Gawd
Joined
Dec 21, 2004
Messages
1,486
Current machine in signature, i was wondering if i should just wait for skylake or get a broadwell set up now? i mean my system isn't bad i just think 2 generations later i should upgrade. Any thoughts or suggestions appreciated.
 
GPU upgrade before CPU. IMO. Then again it depends on the games you play.
 
I'm wondering what this all means and I'm confused.

I'm looking at this and it seems like Skylake needs a new chipset (H110, Q150, Z170 and so on) and has a new socket, namely Socket 1151, with support for DDR3 and DDR4.

Now I'm confused, why would Intel launch the Broadwell K-series "high-end" desktop CPU's with DDR3 on H97 and Z97 boards along with the Skylake locked "low-to-mid-end" desktop CPU's with DDR4 on H170 and Z170 boards ? Are they making us choose between 'new tech but slower variants' and 'old tech but faster variants' ?

I seem to miss something, as I don't see why they would do this.
 
Intel's primary threat is ARM. It makes sense to fasttrack the release of their more efficient designs while tweaking things for their high power designs (no, manufacturing tech is not the same for high power and low power).
 
But it still doesn't make sense. If they wanted to focus on delivering more power/performance efficient designs, why even release Broadwell-K at all when it's the same quarter as Skylake ? What is Broadwell even bringing to the table when they still have a clear lead on AMD in the relevant areas ? It seems to me they'd better focus on Skylake and the 100-series chipsets with DDR4 instead of having a filler-Broadwell K-series lineup.
 
Broadwell is basically Haswell on 14nm, and it was delayed by Intel's 14nm process delay. If Skylake architecture is ready, then it's prepared to be release once 14nm is ready for production. This may explain why Skylake may be released very close to Broadwell.

If Skylake is initially released for the low power segment, then Intel will need Broadwell to fill the high performance segment. I don't think Skylake will immediately fill everything from low power to high performance segment.
 
It does not make sense to wait for 10-15% improvement. The question is also how long wait. Intel priority is now mobile/low power, there is no competition in high performance desktop so any announced release date may easily get postponed.
 
My main beef with all this is that DDR4 is just for the X99 and 100-series chipsets.
It just seems contested when you have the X99 chipset with Haswell-E with DDR4 and the latest Z170 chipset with a low-end Skylake flying with DDR4, extensive PCIe storage support and USB 3.1 while your "high-end" CPU's of 2015 (Broadwell-K) is the only one left with DDR3 and USB 3.0 on the 90-series chipset from 2014.

I wouldn't mind much but 2015 will be the year DDR4 and PCIe storage starts to hit a large part of the enthusiast consumers and you are basically forced to choose between a mid-range Skylake CPU or an "end-of-life" socket 1150 platform.
 
But it still doesn't make sense. If they wanted to focus on delivering more power/performance efficient designs, why even release Broadwell-K at all when it's the same quarter as Skylake ? What is Broadwell even bringing to the table when they still have a clear lead on AMD in the relevant areas ? It seems to me they'd better focus on Skylake and the 100-series chipsets with DDR4 instead of having a filler-Broadwell K-series lineup.

Did you miss my statement where the manufacturing tech for low power and high power are different? A low power manufacturing tech does not scale well to high power, and vice-versa.

In general, the low power tech is ready before higher power tech at a given node. That's why you see mobile ARM chips using smaller tech before high end GPUs.
 
i mean my system isn't bad i just think 2 generations later i should upgrade. Any thoughts or suggestions appreciated.


If it isn't bad then there is no reason to upgrade. We can no longer play the generation game because 3 generations later the performance gains from Sandy > Haswell is a joke on the CPU side of things, especially if your SB part is OC'ed.

Chances are we aren't going to see anything more than the yearly 5% IPC boost on the CPU side of things for a long time. Those tiny improvements aren't going to relieve bottlenecks if they ever pop up without moving to a higher generation part of course.
 
Did you miss my statement where the manufacturing tech for low power and high power are different? A low power manufacturing tech does not scale well to high power, and vice-versa.
I didn't miss it, but I chose to ignore it because it doesn't answer the question why Intel apparently choose to split the desktop lineup of 2015 into:
Low-end: 100-series chipset + Skylake (probably along 80-series + Haswell/Broadwell)
Mid-end: 100-series chipset + Skylake (probably along 80-series + Haswell/Broadwell)
High-end: 90-series chipset + Broadwell-K and Haswell-K
Top-end: X99-chipset + Haswell-E (because Broadwell-E is 2016 apparently)

While I understand that you might mean that Skylake-K just won't be ready in time for June 2015, this doesn't explain why they are wasting everyone's time (whom would care) with the Broadwell-K in the first place. It's not like the i7-4790K is in any trouble of being insignificant in a year's time, they could just as easily bin it to 5-10% higher boost-clocks and call it the i7-4890K. Because that's what I'm expecting from Broadwell-K: just another slight bump that only show in some specific benchmarks and a slightly reduced TDP.

If it isn't bad then there is no reason to upgrade. We can no longer play the generation game because 3 generations later the performance gains from Sandy > Haswell is a joke on the CPU side of things, especially if your SB part is OC'ed.

Chances are we aren't going to see anything more than the yearly 5% IPC boost on the CPU side of things for a long time. Those tiny improvements aren't going to relieve bottlenecks if they ever pop up without moving to a higher generation part of course.
Yeah I agree on this, the main reason to upgrade though is the chipset. If we put a Z77 chipset across a Z97 chipset or a Z170 chipset, the jumps in supported technology are much bigger than the CPU performance from those respective processors. That's why I'm eagerly awaiting the Z170 chipset with DDR4, USB 3.1 and extensive PCIe storage. But I'm not sure what Skylake will bring to the CPU table if the successors to the i5-4690K and i7-4790K are going to be Broadwell processors.
 
I didn't miss it, but I chose to ignore it because it doesn't answer the question why Intel apparently choose to split the desktop lineup of 2015 into:
Low-end: 100-series chipset + Skylake (probably along 80-series + Haswell/Broadwell)
Mid-end: 100-series chipset + Skylake (probably along 80-series + Haswell/Broadwell)
High-end: 90-series chipset + Broadwell-K and Haswell-K
Top-end: X99-chipset + Haswell-E (because Broadwell-E is 2016 apparently)

While I understand that you might mean that Skylake-K just won't be ready in time for June 2015, this doesn't explain why they are wasting everyone's time (whom would care) with the Broadwell-K in the first place. It's not like the i7-4790K is in any trouble of being insignificant in a year's time, they could just as easily bin it to 5-10% higher boost-clocks and call it the i7-4890K. Because that's what I'm expecting from Broadwell-K: just another slight bump that only show in some specific benchmarks and a slightly reduced TDP.


Yeah I agree on this, the main reason to upgrade though is the chipset. If we put a Z77 chipset across a Z97 chipset or a Z170 chipset, the jumps in supported technology are much bigger than the CPU performance from those respective processors. That's why I'm eagerly awaiting the Z170 chipset with DDR4, USB 3.1 and extensive PCIe storage. But I'm not sure what Skylake will bring to the CPU table if the successors to the i5-4690K and i7-4790K are going to be Broadwell processors.

Because they're going to be making Broadwell chips anyways, and they need to sell those dies somehow. In addition to that, there will be people that always upgrade to the latest and greatest, no matter how soon the next one is coming.

Besides, if Broadwell follows Intel's typical tick-tock cycle, there won't be any IPC increases, only GHz increases.
 
Due to the recent stagnation in desktop CPU, it is the first time in 20 years that I keep a CPU (3770K) for more than 2 years.
Frankly, I should be good for another 1-1.5 year.
 
Besides, if Broadwell follows Intel's typical tick-tock cycle, there won't be any IPC increases, only GHz increases.

Even that is highly doubtful. The biggest recent change in generations was introduction of 8 core Haswell-E from the previous 6 core processors. This in fact was nothing new since it was just repositioning of one level higher Xeon processor for consumer market. But stock clock rates came down.
 
2500k @ 4.5ghz is still overkill for 90% of stuff out there, even games.
 
Besides, if Broadwell follows Intel's typical tick-tock cycle, there won't be any IPC increases, only GHz increases.

Ivy had 5% IPC over Sandy so I'd expect Broadwell to have some improvements too.

Plus Broadwell will be best cpu that works with ddr3 - everyone who invested into good ram might find it atractive because of that.
 
Ivy had 5% IPC over Sandy so I'd expect Broadwell to have some improvements too.

Plus Broadwell will be best cpu that works with ddr3 - everyone who invested into good ram might find it atractive because of that.

Ivy was the exception. Most previous generation die shrinks had no increase in IPC or change in architecture.
 
Current machine in signature, i was wondering if i should just wait for skylake or get a broadwell set up now? i mean my system isn't bad i just think 2 generations later i should upgrade. Any thoughts or suggestions appreciated.

I'm just curious as to what Boradwell CPU you would buy right now? I can't find any Broadwell desktop CPU's for sale. I have been trying to hold off on building my Plex server until I can get a CPU that sips power because it will be on all the time but still has enough power to transcode two or three simultaneous streams. The low power Haswell's don't have enough processing power and the standard Haswell's use too much electricity.

If you could point me to a Broadwell CPU I can buy right now it will give me something to keep an eye on for Black Friday/Cyber Monday deals. I live near a Microcenter and traditionally Black Friday at MC will yield me the best price for the next year.
 
My bad, i noticed after i posted this that there wont be a broadwell for a while and confused the lga 2001-v3 for it. I meant should i upgrade to either a 4xxx socket 1150or a 5xxx socket 2011 series, also i just bought my gpu in june and i am not upgrading that yet.
 
i mean my system isn't bad i just think 2 generations later i should upgrade..

Kill that thought. Kill it thoroughly forever. Never upgrade without a reason. Ever. There is no such thing as future-proofing. If a game eventually comes out that your computer can't handle, upgrade at that time. The hardware available at that time is going to be better than the hardware out right now that you don't even need.

As for Broadwell, Intel can keep it.
 
In my opinion the i5-2500k is a good CPU. I'm still running an i7-920. Intel hasn't really been offering a lot of performance increase since the i series have some out. It's been such a disappointment I haven't upgraded. To me the only reason to upgrade are the new features in the motherboard chipsets. For example my X58 chipset doesn't support AES encryption on my SSD.

Here is my advice to you:
Are you doing a lot of video encoding that would benefit by upgrading to 6 or 8 cores?
Are there features in the new chipsets that you would use that you are missing out on?
If you answered no to both questions then don't upgrade now. Wait for Skylake.

At this point they may as well throw the desktop Broadwell in the trash. Skylake is due Q2 2015. They already failed in delivering Broadwell on time. If they do hold Skylake back so they can sell an obsolete cpu for 6 or 9 months that will be a second failure. Consumers will forgive the first one because it was out of their control but the second is just taking advantage of your customer.

I understand the feeling that you need to upgrade because you have had that computer for a while. Believe me, my computer is 5 years old now. It blows my mind that Intel hasn't made a worthy upgrade by now. They pulled so far ahead of AMD with the i series that they have been phoning it in for years. It's all been about using less power. My i7-920 uses a lot of electricity, especially overclocked. If I switched to Haswell I would save a lot on my power bill. If I bought an 8 core Haswell-E I would see significant improvements in video encoding times. That's about it though. In gaming and everyday use I wouldn't see any difference.
 
At this point they may as well throw the desktop Broadwell in the trash. Skylake is due Q2 2015. They already failed in delivering Broadwell on time. If they do hold Skylake back so they can sell an obsolete cpu for 6 or 9 months that will be a second failure. Consumers will forgive the first one because it was out of their control but the second is just taking advantage of your customer.

Isn't the Skylake that comes out at the same time as the desktop Broadwell the low end stuff?
 
thanks i appreciate the input guys and gals, maybe ill wait till late 2015 or 2016 and see if anything is better at that time, who knows maybe AMD will get off their ass and make a good chip again.
 
Due to the recent stagnation in desktop CPU, it is the first time in 20 years that I keep a CPU (3770K) for more than 2 years.
Frankly, I should be good for another 1-1.5 year.

I'm on a OCed 2600K. Recently upgraded video card (AMD 7950). I have no reason to ugprade. I haven't gone this long without an upgrade in a long time has to be 20+ years as well... 486). Don't know if we've hit the limits or if it's lack of competition or if there just isn't any software taking advantage of what we have so there isn't any reason to go faster, anyway...

Kill that thought. Kill it thoroughly forever. Never upgrade without a reason. Ever. There is no such thing as future-proofing. If a game eventually comes out that your computer can't handle, upgrade at that time. The hardware available at that time is going to be better than the hardware out right now that you don't even need.

I have to remind myself of this all the time. I get that upgrade itch every so often. It's always fun to upgrade. But, the gains I'd get won't matter much at all. Yea, a little faster, but there is absolutely nothing I cannot do or struggle at. My PC is plenty fast for everything I do...

I'd wait. You're not held back on the CPU side.
 
I've been on x58 forever. I7 920 to i7 970 (then x5660, same speed, lower volts).

The only reason I'd upgrade is if my Rampage III Extreme died horribly and took the CPU with it. Even then....I might look into a used x58.

I was thinking Haswell-E...then Broadwell-E. But if my system last as long as it can, Skylake-E will probably be it.

Way to go Intel....you just competed against yourself.
 
I'd say you're fine with your current CPU for gaming duty since performance is going to be largely dictated by your GPU...and you've got the power there.

The only reason to upgrade to a currently available platform is to get new chipset features, but only if you can truly benefit from them. Otherwise, stay put, save up a lot of money over the next year or so, and upgrade to a 14nm 8C/16T. Heck, when your OC'd 2500K actually starts getting long in the tooth, maybe 10nm will be out by then. :/
 
Due to the recent stagnation in desktop CPU, it is the first time in 20 years that I keep a CPU (3770K) for more than 2 years.
Frankly, I should be good for another 1-1.5 year.

I'm still on a SB i7-2600K from 2011. That is unheard of for me, I'm usually upgrading every 1 - 1.5 years or so. From what I've researched, the difference between my Sandy Bridge and the Haswell equivalent is negligible.

The only thing I'm upgrading this year is my GTX 770 to GTX 970 SLI. Even that would be unnecessary if I had the 4GB version of the 770.

I keep wondering if I'm falling behind the times but it seems the reality is that:

1. Intel is getting/is comfortable with their position in the market.

and/or

2. We're starting to reach the limits of what current materials used in processors can provide in terms of performance.

I'm no expert but experience tells me we should have seen another gigantic leap in CPU technology by now.

I went from:

Pentium 4 Prescott > Pentium D Smithfield > Core 2 Duo Conroe > Core i7 2600K Sandy Bridge

Each one of those had a massive performance advantage over the other (P4 to PD not as much but still a good bit). At this point I see no reason to upgrade for at least another 2 generations.
 
Last edited:
1. Intel is getting/is comfortable with their position in the market.

and/or

2. We're starting to reach the limits of what current materials used in processors can provide in terms of performance.

I believe it is both. As for 2. looking from the outside 22nm tri-gate helped the low end power consumption versus 32nm but did not offer much improvement at 4GHz+ I hope that 14nm will solve this otherwise I worry that we will have more of the same until a breakthrough happens in the materials.
 
I believe it is both. As for 2. looking from the outside 22nm tri-gate helped the low end power consumption versus 32nm but did not offer much improvement at 4GHz+ I hope that 14nm will solve this otherwise I worry that we will have more of the same until a breakthrough happens in the materials.

Also, try to see it through the glasses of manufacturer (Intel): Market growth worth noticing is exclusively in portable where power consumption is priority and where the company position is rather weak, moreover deadly endangered by ARM. Solution: deemphasize desktop, keep the position in server/data center. 14 nm development push in portable, desktop just a sideline. Thus anybody waiting for Broadwells, Skylakes and followers may buy Haswell now and forget about 10% performance gains per core.
 
1. Intel is getting/is comfortable with their position in the market.

and/or

2. We're starting to reach the limits of what current materials used in processors can provide in terms of performance. .

Both. #1 is partially because AMD isn't competing. Also, R&D costs rise with time yet sales are not doing so well. People don't upgrade computers as often these days as they get faster. As someone said earlier in this thread, Intel has to compete with themselves i.e. they are having trouble getting people to upgrade from older Intel products (particularly X58 and newer, but in many cases people are satisfied with even Core 2 Duo/Quad machines still).

The only thing I really care to buy right now is more storage. Maybe a new GPU in another generation or 2, but even that would just be a midrange one.
 
Back
Top