TR Forums

With Hasewell-E on the horizon I'm finally feeling the upgrade itch after probably the longest stretch about not really caring about high end processors I've ever experienced (plenty of cool stuff happening at the low end thankfully)... I think a lot of you are in the same boat as Sandy Bridge and specifically 2500-2600Ks have aged very, very well. Realistically I still don't think I'll upgrade anything for at least another year because I doubt I'll run into anything 4.2Ghz of Sandy can't handle BUT when I look at a game like GalCivIII and it's 64bit only nature and heavily multithreaded AI I start to feel like I need to build a system that will just crush it, you know? If I decide that's a game I want to play and it turns out that extra cores and memory bandwidth can reduce my turn times that becomes something I care about more than an extra few FPS I might get by putting money into a new CPU instead of GPU.

So what makes you want to upgrade? CPU features? CPU intensive gaming? A new platform/chipset? Power efficiency? Overclock-ability? Handbrake or other non-accelerated video encoding? Professional work like FEA or some other number crunching process? If you've got a perfectly capable CPU right now, how big of a performance improvement (and in what arena) would it take to justify the expense?

When it gets hard to justify upgrading your own machine, consider the trickle-down benefits. When a family member lets the magic smoke out of theirs or begins to notice that their dinosaur isn't keeping up, that's additional justification for me to hand down my slightly-old system to them and build a new one for myself.

I feel you in the thirst for more power. Part of me wants to get Haswell-E, but I know that it is not something that will yield a huge performance increase.

The 2600k is still a beast when overclocked, you'd be hard pressed to run into a CPU bottleneck on a single-GPU rig. When you start getting into SLI though, a 6-8 core could be very useful.

The next upgrade to my i7 3820k x79 system will definitely be a new GPU, replacing the 2 GB GTX 670 with a GM204 or GM200 GPU.

drfish wrote:

Handbreak (handbrake) or other non-accelerated video encoding?

this is what does it for me

you cant really get much faster than my 5ghz 3930k atm but when haswell-e arrives and if i see that the 8 core overclocks well then i will be jumping on one fast

I'm waiting until Intel offers a consumer CPU with 8 true cores before I replace my FX-8350.

SuperSpy wrote:

I'm waiting until Intel offers a consumer CPU with 8 true cores before I replace my FX-8350.

Here's what happened when AMD stopped being competitive at the high end.

JustAnEngineer wrote:

Here's what happened when AMD stopped being competitive at the high end.

Yeah, considering the quad core Haswell version of that CPU can be had for less than 10% of the price, it makes me think I'm going to be sticking to my FX for a looong time.

I would wait unless someone else in the family needs a high end rig that your 2600k still is.

With Mantle and Direct X12 coming reducing CPU overhead I think it would be wise to wait and see what if much performance in games is gained.

I see my upgrade coming when PCI-e 2.0 at 8x becomes a bottleneck to a video card on SLI configurations and such. I think that will have a worthy reason to upgrade then.

If Sandy was PCI-e 3.0 I would honestly buy another Sandy system used, save a ton of money and have plenty of cpu power.

I only want a CPU upgrade when it starts to bottleneck my GPU heavily in a game, nothing else that I do on PC pushes it as hard as gaming except video conversion but that isn't time critical for me. Most of my GPUs have been in the sweet spot of price:performance so I don't need to upgrade CPUs very often.

Before and after my last CPU upgrade I did a few basic game tests to see how much performance was improved. At the time I was desperate to get better DayZ performance (heavily single-threaded), thankfully I got it but I quickly grew old of the unfinished game. The extra performance has been beneficial for Natural Selection 2 (FPS went from hovering around 90 to 120) although I didn't measure averages at the time.

SuperSpy wrote:

JustAnEngineer wrote:

Here's what happened when AMD stopped being competitive at the high end.

Yeah, considering the quad core Haswell version of that CPU can be had for less than 10% of the price, it makes me think I'm going to be sticking to my FX for a looong time.

Kind of apples and oranges, that is a Xeon E5, 2P capable. Need to find an Opteron equivalent for comparison.

vargis14 wrote:

With Mantle and Direct X12 coming reducing CPU overhead I think it would be wise to wait and see what if much performance in games is gained.

drfish was looking at a simulation style game where CPU requirements are not going to go away. Your regular shoot-em-up's? Perhaps.

Speaking of holding onto stuff: I'm still good with my first gen i7-875K.

Gone are the days when you needed to replace your system every couple years just to be able to run programs at all, or without pulling your hair out waiting. Since about 2008, I've only chose to upgrade when it offers a NOTICEABLE improvement (a couple seconds in a rendering/encode benchmark is NOT noticeable. 30% or more improvement in a rendering/encode benchmark, or 15+ fps using the same GPU IS noticeable). Here are some categories that drive CPU upgrades for me, in order of importance:

1) A dramatic change in processor performance: A significantly improved architecture (AMD K8 / Intel Core / Sandy Bridge), > 10nm process node reduction, adding 2+ cores (if software can efficiently utilize them), significantly increased clock speeds, significantly improved IGP (if no dGPU), etc.
2) A new tech is introduced/integrated on the platform: DDR3/DDR4, USB3.0, SATA 6Gbps, PCIe 3.0, M.2, etc.
3) Some of these categories are on the fence and you've found someone to buy your existing setup.
4) A new tech is introduced to the CPU: Turbo Boost, QuickSync, IGP added, etc.


Category #1 is the most influential because it can drive an upgrade even in the same CPU generation. (ie, Pentium dual core -> i5/i7) Maybe the system was built on a tight budget, and within a year or so, funds became available for an upgrade.

Flying Fox wrote:

vargis14 wrote:

With Mantle and Direct X12 coming reducing CPU overhead I think it would be wise to wait and see what if much performance in games is gained.

drfish was looking at a simulation style game where CPU requirements are not going to go away. Your regular shoot-em-up's? Perhaps.

Yeah, for GalCivIII (and maybe others) we're talking about some serious number crunching that needs to get done. Late game Civ5 bogs down enough as it is, something like GalCivIII could start to get on my nerves if I'm spending a lot of time waiting for turns to process.

Factors:

1. A new CPU will give me noticeable performance.
2. I can afford a new CPU.

CPU upgrades are pretty much impossible.
The only scenario I can think of when upgrading the CPU would not be completely retarded would be to buy a Pentium Anniversary Edition this year with a view towards replacing it with an unlocked i7 in a year or two.

In just about all cases, we must upgrade the entire platform. So I upgrade for major platform features.
I upgraded for USB 3 and for DirectX 11.
My next upgrade will be for USB 3.1 with the new connector, DDR 4 and DirectX 12.

I upgrade when my Pro Tools sessions start to lag. When starting playback causes it to hang for a few seconds, or I can't use many sampled virtual instruments at once, or it hiccups too often at a reasonably low latency, then something needs to change.

Milo Burke wrote:

I upgrade when my Pro Tools sessions start to lag. When starting playback causes it to hang for a few seconds, or I can't use many sampled virtual instruments at once, or it hiccups too often at a reasonably low latency, then something needs to change.

Pretty much this is what motivated me to go from an i3 2100 to an i5 3570K and then OC it to 4.5GHz.

Also there are lots of ways to conserve CPU power in PT11. The two I do most both come from older versions of PT but go much faster in 11.

* trim your audio, since plugins not in use will get turned off until they're needed
* bounce VIs to audio tracks, then inactivate the MIDI/instrument track

Supposedly that second one will be rendered obsolete in the next version of PT, based on a couple articles on Pro Tools Expert. It should be a single-click Freeze button like in Logic or Sonar.

I upgrade CPUs very infrequently, only when I'm really feeling the bottleneck in games. Going from a 3.2Ghz E6750 Conroe to a i7-3770 was insane. Went from sputtering around 30fps at medium settings to constant 60fps at max. Also SATA 6Gbps, USB3, and other new chipset features are most welcome and things like that will likely spur my next CPU/platform upgrade.

Plus my brother's girlfriend games perfectly well on the old LGA775 system (I replaced the E6750 with a QX6700 I got for $70 on ebay).

I'd still be using my i7 920 if it hadn't been such a pain in the ass over the years.

When I was active in Folding@home upgrades were motivated by getting core count on a budget, since the F@h SMP client scaled well and points awarded on the SMP work units were a non-linear function of how quickly you turned them around. The CPUs being replaced either trickled down to other family members or remained in use as headless Folding@home bots.

Since I no longer fold actively (participation now tends to be limited to using it as a burn-in test for new systems and the annual TRFrankenbot thing), and don't game, the pace of CPU upgrades has slowed to a trickle. When they *do* happen they tend to be a generation behind the curve (I don't think I have spent more than $200 for a CPU in around 10 years).

I did pick up a CPU/motherboard bundle at Microcenter recently -- FX-8350 + Asus M5A97 R2.0 (one of the nicer mid-range AM3+ boards) for $220. Not really much of an upgrade CPU-wise though (my existing one is an FX-8320). The rationale is that it will allow me to upgrade my main desktop to Ubuntu 14.04 and SSDs (from Ubuntu 12.04 and mechanical HDDs) without taking the old system down first. I can build and test the new system in parallel, and switch over to it after it has been fully tested and burned in. There's a touch of nostalgia too, as I believe the FX-8350 is AMD's swan song in the performance desktop segment (the uber-high wattage FX "stunt" CPUs don't count).

derFunkenstein wrote:

Pretty much this is what motivated me to go from an i3 2100 to an i5 3570K and then OC it to 4.5GHz.

Also there are lots of ways to conserve CPU power in PT11. The two I do most both come from older versions of PT but go much faster in 11.

* trim your audio, since plugins not in use will get turned off until they're needed
* bounce VIs to audio tracks, then inactivate the MIDI/instrument track

Supposedly that second one will be rendered obsolete in the next version of PT, based on a couple articles on Pro Tools Expert. It should be a single-click Freeze button like in Logic or Sonar.

Thanks for the tips. If a session is loaded with virtual instruments and is starting to lag, bouncing them out to tracks seems like a no-brainer to me. But, as my Pro Tools sessions seem less and less about recording per-conceived songs, and more and more about composing as I go, it's nice to have sufficient headroom for a number of virtual instruments. (And I'm currently more interested in modeled pianos than sampled pianos due to the phase issues on basically every sampled piano. Modeled pianos don't require much in terms of system resources.)

My old system, an AMD Athlon X2 5000+ with 4 GB of RAM, was starting to suffer. After a lot of manual timeline editing and micro-fades for a song's worth of drummer and bassist mistakes, even just pressing play took sometimes 10 seconds to spin up. That was on Pro Tools 8.

Now I've got a mildly overclocked 3570k and I'm irritated that Pro Tools and virtual instruments are utilizing so little of my 16 GB of RAM. Other than getting all of my sessions on one SSD and my virtual instruments on another, I don't have an upgrade itch currently. (Except to replace my AMD 5870, but that's a different topic.)

Well, I'm also irritated so little of my old plugin collection made the transition to Pro Tools 11. That's a different type of upgraditis with a different (and equally small) budget.

derFunkenstein wrote:

Milo Burke wrote:

I upgrade when my Pro Tools sessions start to lag. When starting playback causes it to hang for a few seconds, or I can't use many sampled virtual instruments at once, or it hiccups too often at a reasonably low latency, then something needs to change.

Pretty much this is what motivated me to go from an i3 2100 to an i5 3570K and then OC it to 4.5GHz.

Also there are lots of ways to conserve CPU power in PT11. The two I do most both come from older versions of PT but go much faster in 11.

* trim your audio, since plugins not in use will get turned off until they're needed
* bounce VIs to audio tracks, then inactivate the MIDI/instrument track

Supposedly that second one will be rendered obsolete in the next version of PT, based on a couple articles on Pro Tools Expert. It should be a single-click Freeze button like in Logic or Sonar.

This used to be me too, except with Cubase/Nuendo/Maschine/Ableton/Sonar.

But in the most recent 7 years, there has been virtually no problem with older CPUs so now I upgrade when graphic rendering takes too long, or when I want to bump up my F@H throughput.

I dunno having a few extra bucks and just figuring it's time for something new. Last upgrade was just cause I wanted to move into the Haswell line.

I upgrade every 5-10 years these days, I just grab the best deal in the highend when there's a good sale going on. My excuse for upgrading from a 4 core Athlon64 was switching from HDD to an SSD drive for the OS, which in turn made me switch from winxp to win8. Then I thought might as well upgrade the memory to 16gb, and from there I figured I might as well get a newer CPU and mobo while I'm at it, so went with an i7 4770s that was on sale, kinda a snowball effect.

So yeah it wasn't exactly a better CPU that got me into an upgrading mood, that was not really holding me back personally, had more than enough performance for my needs. It was mostly the fact that SSDs dipped under $0.50 in price, and it triggered an interest in building a newer rig that I'll keep till the next 1 or 2 console cycles show up.

I upgraded from i5-750 to Haswell only when my H55 motherboard gave up its ghost and I couldn't find a replacement H55/P55 board. All I had to buy was a motherboard, a processor and two sticks of RAM. Everything else was reused from the original rig, except the Silverstone casing from the dying Core 2 Duo HTPC.

Evil names like Devil's Canyon

My upgrade itch is in hibernation. I used to upgrade every 3-5 years when the performance jump hit a sweet spot (Core 2 Duo launch, Core i7 launch, etc) but that kind of performance jump hasn't really happened in awhile now. The only reason I upgraded my 920 to a 4771 was because the old motherboard had some limitations and issues I didn't want to deal with. It would probably require those new 8-cores to drop to $500ish before I'll consider any upgrade from the 4771. Nice to see hex-core prices are dropping though, that's a good start.

Pretty much when my CPU becomes a significant bottleneck in games and I can get a significantly faster CPU without breaking the bank. I do some video encoding too but I tend to run that when I'm away from my PC or overnight, so absolute top-of-the-line performance isn't really necessary. I'm already encoding stuff far faster than I could ever watch it all.

For both CPUs and GPUs I like to get at least a 50% performance improvement in games before upgrading though, preferably at a similar price point to the previous CPU/GPU I bought. Any improvement smaller than that feels like a waste of money, especially if it also requires a new motherboard. Going from my E8400 to this 2500K it was actually closer to a 100% improvement in many games (combination of higher per clock performance, higher clockspeed, and double the number of cores)

The New CPU must be able to give a meaningful performance gain over my existing cpu.. Taking me from 50 fps to 58 fps is .. for example..... Irrelevant.
Gaming is my only performance concern as everything else is handled very adequately by moderate to low end tech nowadays anyway.

I am very intrigued by going above 1080p.... If I take that plunge, I will need a system to build around that gaming resolution.


If the New cpu consumes considerably less energy and has considerably more performance AND I can get a reasonable price for my old rig, I will buy a whole new system based on the new architecture.. GPU TOO.... Otherwise, don't waste my time with incremental gains... (LOOKIN AT YOU INTEL).

The new architecture better be worth the time and expense (re-load the o.s. buying setting up etc.) to go through or forget it!

Oh one final thing.. . I never wait till my old rig is about to self destruct. I will sell it while it is still worth something and stable for the next buyer.


This is exactly the kind of issue I would like to see tech sites do an analysis on... They should do Cross Generational Comparisons when doing CPU and GPU reviews instead of filling up the Benchmark Graphs with hardware that is at totally different prices... I usually stay in the mid-to upper range to my frugal nature. I will NEVER EVER EVER buy a high end SLI or Crossfire rig.. That is just not necessary... .However I do want to see what each generation brings to the table performance and TDP wise before I decide to sell my old hardware and buy new stuff!
I would much rather see cpu and gpu comparison reviews that go back 3 generations instead of showing me hardware I will never buy anyway..

matnath1 wrote:

This is exactly the kind of issue I would like to see tech sites do an analysis on... They should do Cross Generational Comparisons when doing CPU and GPU reviews instead of filling up the Benchmark Graphs with hardware that is at totally different prices... I usually stay in the mid-to upper range to my frugal nature. I will NEVER EVER EVER buy a high end SLI or Crossfire rig.. That is just not necessary... .However I do want to see what each generation brings to the table performance and TDP wise before I decide to sell my old hardware and buy new stuff!

I totally agree with you about that. As you said most people tend to have a price range they stay within, and very few people are going to go from a $200 graphics card to a $1000 one (or vice versa). I also understand why most sites don't do such testing though. The more cards you use in your testing the longer it will take, and many sites already struggle to get reviews out in time for NDA liftings. Unfortunately being first tends to get more page views than being good and informative, but late.

Anandtech has some nice "CPU Bench" and "GPU Bench" charts that go back to at least a couple previous generations, if you haven't already seen them:

http://www.anandtech.com/bench/CPU/1033
http://www.anandtech.com/bench/GPU14/815

The last time I upgraded (just) a CPU was from a Core 2 to a Core 2 Duo. That was worth it, and not enough of the rest of the system (SATA, USB, PCI, Memory speed, etc.) were worth a new mobo.

The next time I might upgrade a CPU will be when TSX works, and only then if Intel replaces my 3 Haswell's (2 of which are Xeons) for free.