Does it make sense to buy Haswell now?

[ShintaiDK]

Is there a noticeable difference between a 4 core with HT & full on 8 core? Doesnt the system just see 8 threads either way?

Performance wise there is. But for anything outside performance no. Its just 8 cores seen from the system.

 

[Chesebert]

I have finally gotten my system up and going. I ended up with 4790k. It's currently sitting at 4.8Ghz at 1.249v in Prime95; temp is around 80c on the hottest core but only 71 on the coolest. I think this chip can use some lapping in 8 yrs or so when additional OC is necessary. Did I get a good chip?

Encoding is off the chart...and LR5 is very snappy.

 

[PliotronX]

I have finally gotten my system up and going. I ended up with 4790k. It's currently sitting at 4.8Ghz at 1.249v in Prime95; temp is around 80c on the hottest core but only 71 on the coolest. I think this chip can use some lapping in 8 yrs or so when additional OC is necessary. Did I get a good chip?

Encoding is off the chart...and LR5 is very snappy.

Certainly did, son! Nice going. I'm jelly. I would forgo the lapping and delid it if you really want the optimal thermal performance, lapping wouldn't drop temps much and it's more time intensive.

 

[NTMBK]

Wait, can you still delid the 4790k? I thought it had solder?

 

[Headfoot]

Wait, can you still delid the 4790k? I thought it had solder?

No, just less terrible TIM with a smaller gap between IHS and die and some more voltage regulation circuitry. Solder kicks in once you hit Haswell-E

 

[thilanliyan]

4.8GHz?!! My 4.1GHz @ 1.25v 4670K is pathetic lol...it's even watercooled but still gets up to 75-80C...
Might need to consider a delid.

 

[RussianSensation]

I have a very ancient system (Q6600@3600) and I am itching for the longest to upgrade.

So I am looking into a 4690k on an Asrock Z97 Extreme6 which I can hopefully overclock to 4.5 or so.

Now I read Skylake is already around the corner, although possibly delayed. Obviously I feel odd now upgrading if an entire new architecture might come out late next year? (Here I simply want to ignore Broadwell, not that I see any sense in Broadwell anyway...)

On the other hand, something new is ALWAYS around the corner...so what should I do? Get a recent Haswell system and then (maybe) upgrade again once Skylake has come out?

I am going to have to go against the recommendation to upgrade to a Haswell 4690K for 2 reasons:

1) You waited this long to upgrade but you missed the perfect timing to get i5 4670K/4770K 1.5 years ago. In other words, you waited nearly 2 years to get 97% same tech as 4690K/4790K, but Skylake is probably just 7-8 months away. If you are going to get new DDR3 and pay $240+ for an i5, might as well wait for the latest tech that's around the corner and get DDR4 that will be viable for 5+ years, while DDR3 is basically worthless after 2H 2015. Based on Intel's CPU trends, usually more often than not the best time to buy a desktop Intel CPU is at the beginning of the next Tock cycle (Jan 2011 SB, June 2013 HW, and soon Skylake). Ticks rarely bring anything worthwhile with minimal IPC increase, minimal extra overclocking headroom, but nearly 1 year later. In the case of 4690K/4790K, it's even worse since they aren't even Broadwell-K refreshes, so you don't even extra features, a lower node or reduced power usage.

2) Given your upgrade pattern (or lack thereof) of holding on to a 2007 Q6600 (a nearly 8-year-old CPU!), if you are going to upgrade, might as well spend the extra $ for the 5820K or 4790K. The extra $ over the next 5-8 years you'll likely hold this system will not mean much to you in terms of total ownership cost over such a long period of time but the extra performance from 50% more cores/HT will surely come in handy over the next 5-8 years. Considering 4790K works at 4.4Ghz out of the box, buying an i5 4690K and overclocking it to 4.5-4.6Ghz with an after-market cooler is worth than getting a stock 4790K and running stock cooler. If you have a MC near you, then 5820K with Asrock Extreme 4 board is a better option than 4790K over such a long period of holding the system. If X58 Xeon drop-ins are any indication, X99 chipset could benefit from some incredible Xeon bargains at the end of its useful life too.

 

[xthetenth]

With a long upgrade cycle like this, it makes it hard. The probably less than but likely close to a year space between now and Skylake is a reasonably long period with a significant delta in performance, but if you're going to try to stretch it out to nearly a decade, that length of time is a really big factor to multiply the difference between Skylake and Haswell. There's a compelling argument for each, but I'd say to definitely give an i7 a good hard look if you want to make the computer last a long time. The price difference isn't as huge percentage-wise if you're buying a totally new platform, and with things slowly starting to take advantage of more cores this will be a performance increase likely greater than the generational advantage could be.

 

[flexy]

Ok, I now got a rather reasonably priced, USED i4770k bundle with an Asus Z87 board. My reasoning was that I will likely want Skylake but this is still at least a year to go. So this is not seen as an investment that should me last "7-8 years" (not with DDR4 basically around the corner)...rather than a "cheap" upgrade so that I can get my system current while awaiting Skylake. Even if I were to keep the i4770k system for longer I am sure it would be sufficient for what I am doing.

 

[DilipVasan]

Its still got some life left

 

[Bradtech519]

I think it would be a big jump for you. I mainly play games & do a lot of distributed computing on my CPU/GPU. The 4770/4790ks are beasts at doing the same projects I do now in DC. I notice at times they complete 5 minutes per task faster in projects I do. It's been temping for me even with an AMD FX 8350. Pull the trigger & enjoy!

 

[Chesebert]

Just to close the loop on my part of this thread. I am still sitting at 4.8ghz @ 1.265v (x264 16 threads 48hour loop stable). The OC Haswell is on average about 3x faster than Q6600 @ 3.6ghz. The difference is not obvious under average windows/browsing loads, but once I crank up x264 or doing any Lightroom work, the performance gain is obvious.

I think this chip can easily get over 5ghz with some work (not inclined to do anytime soon). I think I am good until Lightroom can be run on more than 8 threads (currently the max is 6 threads).

 

[StrangerGuy]

I am going to have to go against the recommendation to upgrade to a Haswell 4690K for 2 reasons:

Even if you want a i5 you can just buy a 4690K (or non-K? Not sure whether that works) and run it max turbo 4 core 3.9GHz on a ~$60 MCE enabled budget mobo. Like anyone is going to notice a difference in a real world gaming situation between that versus like, let's say a 4.4GHz 4690K. I would rather use the money saved from not overclocking on far better things like GPUs and SSDs.

I wish that stupid reddit/linustechtip advice of recommending "i7 is useless, spend $200+ mobo/cooling to overclock a 4690K" would just stop, especially towards first timers. OCing as a bang-for-buck value activity is now dead.

 

[DrMrLordX]

It isn't dead, it's just been segmented in such a way that you can't outperform the top-of-the-line Intel CPU with an overclocked budget CPU anymore. You pay to play or you struggle with something that will never be a 5960X (or even a 4790k).

 

[StrangerGuy]

It isn't dead, it's just been segmented in such a way that you can't outperform the top-of-the-line Intel CPU with an overclocked budget CPU anymore. You pay to play or you struggle with something that will never be a 5960X (or even a 4790k).

The part about OCing to reach the speed of flagship chip is a nice bonus but ultimately irrelevant now.

It's dead not because you can't OC anymore, it's dead because putting money into OCing Haswell is a terrible investment. Heck its the worst it has ever been since the good old Celeron 300A, we don't have unlimited budgets and when there are far better things to buy for the dollar elsewhere in a build. I find it especially disingenuous (and frankly dumb) whenever somebody says "You should overclock because is free!" like this concept called 'opportunity cost' doesn't exist. The truly surprising thing is hardly anyone takes a step back and realize it themselves.

 

[jkauff]

I have a feeling the 4790K will be the last enthusiast chip we'll see for awhile. Intel's focus at this point is on power savings, not performance.

I can easily OC mine to 4700, which I do when I know I'm going to be doing x264 encoding (otherwise I run it with stock Turbo mode). I could probably go higher if I wanted to apply more voltage, because my temps are quite reasonable at 4700, but just that much of an OC (and the extra threads) has greatly reduced my encoding times.

I don't know if extra cores vs. extra threads would make a difference in x264, maybe someone else has done the research.

 

[StrangerGuy]

I have a feeling the 4790K will be the last enthusiast chip we'll see for awhile. Intel's focus at this point is on power savings, not performance.

I can easily OC mine to 4700, which I do when I know I'm going to be doing x264 encoding (otherwise I run it with stock Turbo mode). I could probably go higher if I wanted to apply more voltage, because my temps are quite reasonable at 4700, but just that much of an OC (and the extra threads) has greatly reduced my encoding times.

I don't know if extra cores vs. extra threads would make a difference in x264, maybe someone else has done the research.

X264 is multithreaded but it doesn't scale particularly well.

1st pass: http://anandtech.com/bench/CPU/53 - Single threaded dependent = 4790K is #2

2nd pass: http://anandtech.com/bench/CPU/54 - 4790K has 52.2 FPS but the 6C/12T 5820K only has 60.1 FPS. Overall price/performance (CPU + mobo + RAM cost) 4790K wins by a long shot. Unless of course your time is worth a lot of money.

 

[coolpurplefan]

Get the Haswell and enjoy. I doubt we'll see Skylake-K until 2016.

I hope Intel reads this and tries to get things done on time. I want a Skylake to record the Olympics.

 

[Chesebert]

The part about OCing to reach the speed of flagship chip is a nice bonus but ultimately irrelevant now.

It's dead not because you can't OC anymore, it's dead because putting money into OCing Haswell is a terrible investment. Heck its the worst it has ever been since the good old Celeron 300A, we don't have unlimited budgets and when there are far better things to buy for the dollar elsewhere in a build. I find it especially disingenuous (and frankly dumb) whenever somebody says "You should overclock because is free!" like this concept called 'opportunity cost' doesn't exist. The truly surprising thing is hardly anyone takes a step back and realize it themselves.

I don't understand your animosity toward 4790K. It's basically a factory-overclocked 4770K. If you treat it as a 3.5Ghz processor that you can easily hit 4.5Ghz with, then it's not so bad. It's no Sandy in terms of headroom, but it sure does no worse than Q6600, which can be routinely OCed by 1Ghz with decent cooling.

Would I have liked a 5820k? Sure, but where I am (not in the States), a comparable x99 setup would cost $600 extra compared to a z97 build. The OCed 4790K is about as good as 5820k at stock on fully multi-threaded apps and decimates 5820k on less than optimal multi-threaded apps (which include most of the apps and games out there).

I believe OC still makes a lot of sense for most of the applications out there. For fully-threaded apps (i.e., those that can take advantage of as many cores as you can throw at them), it would make more sense to get more cores, as no amount of OCing can make a quad core perform at the level of an 8 core no matter how much you overclock the quad.

 

[2is]

Hope to speed up Lightroom 5 (probably should upgrade to SSD first and see if that helps) and reduce the encoding time for my occasional x264 needs.

On the gaming side, I need to get that min frame rate up on some of the newer games (e.g., bioshock infinity) and would like to upgrade to a better GPU (currently on OC 5850) that will not be bottlenecked to the death by the Q6600.

For those uses it is absolutely worth it. Especially with quick sync cutting encode times by a factor of 10+

 

[StrangerGuy]

I don't understand your animosity toward 4790K. It's basically a factory-overclocked 4770K. If you treat it as a 3.5Ghz processor that you can easily hit 4.5Ghz with, then it's not so bad. It's no Sandy in terms of headroom, but it sure does no worse than Q6600, which can be routinely OCed by 1Ghz with decent cooling.

Would I have liked a 5820k? Sure, but where I am (not in the States), a comparable x99 setup would cost $600 extra compared to a z97 build. The OCed 4790K is about as good as 5820k at stock on fully multi-threaded apps and decimates 5820k on less than optimal multi-threaded apps (which include most of the apps and games out there).

I believe OC still makes a lot of sense for most of the applications out there. For fully-threaded apps (i.e., those that can take advantage of as many cores as you can throw at them), it would make more sense to get more cores, as no amount of OCing can make a quad core perform at the level of an 8 core no matter how much you overclock the quad.

I'm talking about Haswell, not Haswell-E. A 5820K has essentially a free OC even if you went for the cheapest route, however the same cannot be said for S1150.

 

[Chesebert]

I'm talking about Haswell, not Haswell-E. A 5820K has essentially a free OC even if you went for the cheapest route, however the same cannot be said for S1150.

5820k has a $600 premium over 4790k in terms of comparable build where I am; so even with a free OC, it wasn't really worth it. If it were $200 or less, I would have totally gone for it. However, for 720p x264 encoding, I am not sure 5820k has any advantage over 4790k with quick sync.

 

[BonzaiDuck]

None of the previous opinions are any worse than mine here.

I'd say if you're still running a Q6600, it's time to upgrade. I don't care . . . if it means the now-obsolete Skt-1155 and either a Sandy or Ivy Bridge, Skt-1150 and the first-gen Haswell's, or the Devils Canyon refresh.

Also, I and others speculated about a thermal "challenge" with the Devils Canyon and the 5820K. Now I see that it's not so bad, at least for the 4790K.

For some things, there's really no reason to OC the 4790K. You can squeeze as much as another 300 Mhz out of it compared to its stock turbo speed of 4.4 Ghz. But with the generational improvements in the CPU, you're better getting a good graphics card to supplement the 4790K.

 

[SteveGrabowski]

Even if you want a i5 you can just buy a 4690K (or non-K? Not sure whether that works) and run it max turbo 4 core 3.9GHz on a ~$60 MCE enabled budget mobo. Like anyone is going to notice a difference in a real world gaming situation between that versus like, let's say a 4.4GHz 4690K. I would rather use the money saved from not overclocking on far better things like GPUs and SSDs.

I wish that stupid reddit/linustechtip advice of recommending "i7 is useless, spend $200+ mobo/cooling to overclock a 4690K" would just stop, especially towards first timers. OCing as a bang-for-buck value activity is now dead.

Oh my God, I hate PC Master Race's infatuation with the i5-4690k. Everything below the 4690k is crap and everything above it is a waste of money because games don't use hyperthreading! I loved overclocking 15 years ago when you'd take a midrange CPU and overclock it to flagship performance, but now you have to buy the i5 or i7 with the flagship stock clock for the socket, spend $20 extra for the unlocked multiplier, $30 extra for the Z97 board instead of H97, $30 extra for the cheapest aftermarket cooler to get, what? 15% higher clock than the turbo if you get a nice chip? What's really funny is seeing people talk about how their 4690k is an 84W chip when the power consumption goes way up with an overclock to 4.5GHz. I can't believe how many people will go buy a $110 liquid cooler and pair it with an i5-4690k when you can just get an i7-4790k actually binned to run at 4.4GHz and a Hyper 212 EVO for a grand total of about $10 more. But I guess the closed loop water cooler looks nicer through the window?

 

[Lepton87]

The part about OCing to reach the speed of flagship chip is a nice bonus but ultimately irrelevant now.

It's dead not because you can't OC anymore, it's dead because putting money into OCing Haswell is a terrible investment. Heck its the worst it has ever been since the good old Celeron 300A, we don't have unlimited budgets and when there are far better things to buy for the dollar elsewhere in a build. I find it especially disingenuous (and frankly dumb) whenever somebody says "You should overclock because is free!" like this concept called 'opportunity cost' doesn't exist. The truly surprising thing is hardly anyone takes a step back and realize it themselves.

I think it's not irrelevant just that it's been relegated to high-end CPUs only. An Overclocked 5820K matches or beats the 5960X in MT and is a lot faster in ST. Of course it can't keep up with an overclocked 5960X in MT.