Am I reading this right? 2500 vs newer CPUs

[gusfring22]

My 5 year old 2500k which i have been consiering upgrading (along with new mobo and memory, sigh) scores a 6449 and a 6600k scores 7804 in an aggregate benchmark tool (cpubenchmark). In ~5 years ~20% is the only CPU gains we are seeing? I must be interpreting this incorrectly. Saw another benchmark at futuremark where it is a similar gain.

http://www.cpubenchmark.net/common_cpus.html

http://www.futuremark.com/hardware/cpu?_ga=1.256454860.635404493.1464981816


I am comparing this to my 3 year old 7700 GPU, and the newly released 1070 is more than double the performance. Obviously not apples to apples but dont see the value in upgrading cpu/mobo/ram based on these #s and the hassle involved, particularly for gaming. If most of what I am seeing is accurate, is going from 2500 to 6600 or 6600k worth all the trouble? I feel like if i am going to upgrade i want to get atleast a 50% bump in performance, atleast that is how i treat GPU upgrades

 

[teejee]

My 5 year old 2500k which i have been consiering upgrading (along with new mobo and memory, sigh) scores a 6449 and a 6600k scores 7804 in an aggregate benchmark tool (cpubenchmark). In ~5 years ~20% is the only CPU gains we are seeing? I must be interpreting this incorrectly. Saw another benchmark at futuremark where it is a similar gain.

http://www.cpubenchmark.net/common_cpus.html

http://www.futuremark.com/hardware/cpu?_ga=1.256454860.635404493.1464981816


I am comparing this to my 3 year old 7700 GPU, and the newly released 1070 is more than double the performance. Obviously not apples to apples but dont see the value in upgrading cpu/mobo/ram based on these #s and the hassle involved, particularly for gaming. If most of what I am seeing is accurate, is going from 2500 to 6600 or 6600k worth all the trouble? I feel like if i am going to upgrade i want to get atleast a 50% bump in performance, atleast that is how i treat GPU upgrades

Yep, the CPU progress on desktop is sad...
I remember the 90's though, started with 20Mhz 486 and ended with 1 Ghz Pentium 3...

Sent from my LG-D855 using Tapatalk

 

[Ken g6]

There's another reason you're having trouble deciding to upgrade. The 2500K is probably the best chip advance to come out this decade.

But Ian Cutress thinks Skylake is a good upgrade.

Sandy Bridge was notable because it inferred a large performance gain at stock speeds, and with a good processor anyone could reach 4.7 GHz and even higher using a good high end cooler. With that, Intel has had a problem enticing these users to upgrade because their performance has been constantly matched by Ivy Bridge, Haswell and Broadwell – for every 5% IPC increase from the CPU, an average 200 MHz was lost on the good overclock and they would have to find a good overclocking CPU again. There was no great reason, apart from chipset functionality to upgrade.

That changes with Skylake.

From a clock-to-clock performance perspective, Skylake gives an average ~25% better performance in CPU based benchmarks, and when running both generations of processors at their stock speeds that increase jumps up to 37%. In specific tests, it is even higher. When you scale up to a 4.5 GHz Skylake against a 4.7 GHz Sandy Bridge, the 4% frequency difference is only a tiny portion of that. There are other added benefits, such as the move to a DDR4 memory topology that has denser memory modules, as well as PCIe storage and even PCIe 3.0 graphics connectivity.

 

[BonzaiDuck]

There's another reason you're having trouble deciding to upgrade. The 2500K is probably the best chip advance to come out this decade.

But Ian Cutress thinks Skylake is a good upgrade.

I won't discourage that thought.

But this shouldn't be so difficult to understand, even if I'm making an educated guess from coming mostly from the software side of "serious work."

Every generation is an attempt to reduce two things and increase another.

First of all, they're trying to reduce the power requirement because they've seen the laptop and tablet writing on the wall. Lower power translates into more things people can do with their laptops between stopovers to an AC outlet.

Second, they've attempted to reduce the size or the lithography -- reflecting the proximity of circuits. My paraplegic friend, who had worked in the "real world" and the classroom as a physicist, had predicted they would have more problems involving quantum physics -- perhaps he meant "leakage" of current -- with smaller lithography. I had also met another gentleman -- an Australian engineer -- when attending a church social in Seattle. We were discussing the trends after he revealed he was leaving Intel. He kept saying wide-eyed: "It's getting down the mol-ec-u-lar level!" And that was back in 2004, but if you worked at Intel, you'd know something of what was on the drawing board, and Intel believes in extensive plans of a Future.

Finally, there's the matter of a crude measure of speed in Megahertz. This also becomes a less important objective just for refining the CPU and adding more to the instruction set. If a point is reached where the technology's limits exceed a broad set of consumer needs, you should be able to see where all of this is going.

Even so, with the K chips that include the 2500K and the 6700K, they haven't completely left enthusiasts with a "need for speed" out of their production plans.

But of course, whether the figures come from Ian Cutress or somebody else, 25% to 37% is quite an improvement -- maybe not yet enough to give up a venerable "perfect computer build" and spend more money again. Depends on "where you're comin' from," how enthusiastic is your enthusiast enthusiasm and your spending priorities.

But ol' Ian there is simply adding confirmation to my own thoughts. The raw measure of speed in Ghz/Mhz is only part of a performance picture.

 

[ehume]

But then to get the most out of a 6700k, you have to have DDR4-4000 RAM, and that's not cheap yet.

 

[MajinCry]

In games, isn't Skylake around 20% faster than Sandybridge? There was a good graph kickin' around that should the generational improvements.

Basically, if ya got Sandybridge, there's no reason to upgrade. And the i5 2500ks are going for £80 on Ebay, which is pretty swanky.

Trick is ta find a well-priced motherboard. The p67 and z68 motherboards are around a hundred bob, fer cryin' out loud.

 

[guachi]

Keep your CPU; invest in a video card by whichever GPU maker floats your boat.

But that's really just reiterating what you already started.

 

[Burpo]

Complete i5 2500 midtower PC for $129 :sneaky:
http://www.ebay.com/itm/381499785756?_trksid=p2055119.m1438.l2649&ssPageName=STRK:MEBIDX:IT

 

[Roland00Address]

I do like Passmark for some things, mainly because it is one of the few benchmarks you can easily get lots of processor data on for there is more processors than most people test. A lot of cpus in the mobile world do not get benchmarked and passmark is one of the few games in town. That said in some things it is fundamentally flawed and is not a good example of real world usage. This is both the pros and cons of a "canned benchmark." When we are comparing desktop cpus do not use passmark for your choices for better benchmarks and metrics are orders of magnitude greater number and many of these benchmarks are better than passmark.

----

Anandtech in its bench has these two cpus compared and a lot more benchmarks and they are closer to real world usage.

http://www.anandtech.com/bench/product/288?vs=1543

That said Sandy to Skylake is only a 20 to 40% single threaded improvement and this can decrease if you overclock.

Yet at stock speeds in multi thread it is 50 to 70% faster in most tests (with the median being near 50 to 60%).

Now if you compare an i7 2600k Sandybridge to an i7 6600k the difference is less pronounced for you do not gain the benefit of hyperthreading comparing the old i5 to the new i7 since they are both i7 hyperthreaded quad cores.

-----

Now upgrading has added benefits of allowing you to get faster storage due to advances in SSD technology and us moving to pci-e based SSDs but if you are going to keep the same storage there will be barely any improvements here. If you game upgrading graphics can be easily double to triple the performance instead of 20 to 40% faster in ST and 50 to 70% faster in multithread.

 

[Gonad the Barbarian]

Complete i5 2500 midtower PC for $129 :sneaky:
http://www.ebay.com/itm/381499785756?_trksid=p2055119.m1438.l2649&ssPageName=STRK:MEBIDX:IT

Tempted. But do I need a second i5 2500 machine?

 

[BonzaiDuck]

Tempted. But do I need a second i5 2500 machine?

The HP bargain uses a non-K chip.

The Anandtech benchmark comparison posted by Roland00Address reveals a lot. In some four benchmarks, the 2500K Sandy actually outperforms the Skylake.

So I actually wonder what would result from pushing my own SB-K chips by another 100Mhz -- because . . . like the millionaire in "8 MM" -- I can . . .
But with six-year-old technology, I suppose you're welcome to twist up the voltages on your motherboard as you might dare. Then, when something "pops," you can always say "Well -- it was old, and lasted a long, long time . . . "

 

[JujuFish]

The Anandtech benchmark comparison posted by Roland00Address reveals a lot. In some four benchmarks, the 2500K Sandy actually outperforms the Skylake.

I think you're misreading something, because I don't see a single benchmark where the 2500K wins.

 

[Nhirlathothep]

My 5 year old 2500k which i have been consiering upgrading (along with new mobo and memory, sigh) scores a 6449 and a 6600k scores 7804 in an aggregate benchmark tool (cpubenchmark). In ~5 years ~20% is the only CPU gains we are seeing? I must be interpreting this incorrectly. Saw another benchmark at futuremark where it is a similar gain.

http://www.cpubenchmark.net/common_cpus.html

http://www.futuremark.com/hardware/cpu?_ga=1.256454860.635404493.1464981816


I am comparing this to my 3 year old 7700 GPU, and the newly released 1070 is more than double the performance. Obviously not apples to apples but dont see the value in upgrading cpu/mobo/ram based on these #s and the hassle involved, particularly for gaming. If most of what I am seeing is accurate, is going from 2500 to 6600 or 6600k worth all the trouble? I feel like if i am going to upgrade i want to get atleast a 50% bump in performance, atleast that is how i treat GPU upgrades

less than 20% (2500k average oc is >>>)
and the price of the correspondent skylake i5 is 100-150$$ more than 2500k at release

consumer cpus are about the same in performance, but <<< TDP

 

[BonzaiDuck]

I think you're misreading something, because I don't see a single benchmark where the 2500K wins.

I feel so stoo-pid! I read through so much material in a day's time, I often read too fast. I overlooked the "Lower is better" small print, if it was all that small. If I spent more time just LOOKING at benchmark comparisons, I would've known better.

Thanks for bringing this to my attention, as humiliated as I am by the experience.

ANOTHER THOUGHT: I think this is "going to be the year" for me to catch up. But that leaves me either an entire motherboard-RAM-&-processor bundle to sell, or I have one too many computers.

Even the first of these SB-K's -- built in mid-2011 -- inspires great fondness. And it is also my HTPC, business and record-keeping system. I don't even play games on it anymore, choosing the 2700K instead.

But that's OK . . . Ah'm doin' gurr-eat thingks! Gurr-eat thingks!

 

[wingman04]

Yep, the CPU progress on desktop is sad...
I remember the 90's though, started with 20Mhz 486 and ended with 1 Ghz Pentium 3...

Sent from my LG-D855 using Tapatalk

How true It is a sad time with technology for intel CPUs is about 5% or 10% with each new product.:\

Anandtech said that going from sandybridge to skylake is averaging about 25% more performance, there just hitting a wall with clock speed and efficiency.

 

[JimmiG]

Yep, the CPU progress on desktop is sad...
I remember the 90's though, started with 20Mhz 486 and ended with 1 Ghz Pentium 3...

Sent from my LG-D855 using Tapatalk

CPUs have basically stagnated. The only reason to upgrade is if your current CPU breaks. We aren't even getting more cores for the money any more. You can't only blame manufacturing tech and physics, because GPUs still manage to evolve.

BTW I started the 90's with a C64 running at 1 MHz and ended with an Athlon @ 900 MHz.

 

[Nhirlathothep]

xeons market and development is ok

 

[seitur]

Inablity (?) to increase CPU performance more than at slow pace is the reason why DX12/Vulkan are created and used even though it comes at high cost of developement in gaming studios.

Because CPUs in DX11 just cannot feed newest more powerful GPUs anymore.

So even though it is costly, AAA graphic games in future will use either DX12 or Vulkan (or other low-overhead API).

 

[Ken g6]

The HP bargain uses a non-K chip.

There's also [thread=2452438]some benchmarks[/thread] that say a 2500 non-K is very similar in performance to a modern i3-6100. Which is one reason I got a modern i3-6100.

I feel so stoo-pid! I read through so much material in a day's time, I often read too fast. I overlooked the "Lower is better" small print, if it was all that small. If I spent more time just LOOKING at benchmark comparisons, I would've known better.

Thanks for bringing this to my attention, as humiliated as I am by the experience.

I have a User Script that can help you with that problem. I can see at a glance that Sandy doesn't win any benchmarks.

 

[RussianSensation]

But then to get the most out of a 6700k, you have to have DDR4-4000 RAM, and that's not cheap yet.

That's a great point. None of the launch 6700K reviews online showed the true potential of the Skylake architecture since they tested it under a memory bandwith bottleneck.

Another point is AT never tested any games where HT shows a massive difference. For instance, Crysis 3, Total War Warhammer, etc.

Some of you aren't comparing the data correctly. I5 2500K OC is a huge bottleneck for 980Ti/1070/1080 level cards. It's an issue because you are buying a high-end card. The last thing you want is going from a 1070 to a 1080 and seeing almost no difference. For those who intend to run 1070 SLI, it's even worse. That means it's not just a matter of Skylake beating Sandy by 25-30%. Look at Digital Foundry where in CPU demanding games, the 4.4Ghz 2600K trailed the 4.2Ghz 6700K by 30%.

" But there are still some noticeable boosts - GTA 5 on the 6700K is 17 per cent faster clock for clock than the 4790K, and 29 per cent faster than both Ivy and Sandy Bridge. Far Cry 4 - an eight-core aware title that demands high per-core performance sees Skylake move 17 points clear of the 4790K, and a mammoth 32 per cent ahead of the second and third-gen i7s."

That's not even testing 6700K with fast DDR4.

To answer the OP, if you are going to upgrade, go 6700K, not 6600K. If you only have a GPU with R9 390/980 level of performance, then of course there is a smaller bottleneck than if you desire to get a 1070/1080.

Once 1080Ti and Vega 11 HBM2 come out, it will be the final coffin into i5 2500K-3770K. All of those will be too slow for such a fast graphics card.

In the real world it also makes it tough to answer since some gamers are OK with dips into 30-40 FPS. For them, even if 2500K is way slower than the 6700K, the overall performance may be satisfactory anyway. Mathematically speaking though, i5 2500K already bottlenecks R9 390 level card. Maximum overclock to 4.5-4.8Ghz is required to feed a 390/970 card with a 2500K:

http://www.eurogamer.net/articles/digitalfoundry-2016-is-it-finally-time-to-upgrade-your-core-i5-2500

As I said earlier, even that isn't enough.

3770K OC vs. 2500K OC with a Titan X:

"Elsewhere, we see The Witcher 3's minimum performance level gain 17 per cent, Far Cry 4 rises by 20 per cent while Battlefield 4 and Crysis 3 gain 31 per cent and 45 per cent respectively. That's not bad at all considering that we have actually lost 200MHz in raw clock-speed compared our maxed Core i5 2500K."

Sorry, but 2500K OC is MUCH slower than 25-30% against a 4.6-4.8Ghz 6700K + DDR4 3466 or higher. The difference in CPU demanding sections of games will rise to 40-50% easily. In Fallout 4, a game highly memory bandwith and CPU bottlenecked, 6700K 4.6-4.8Ghz + DDR 4000 will probably end up 75-100% faster than an i5 2500K 4.8Ghz + DDR3 1600.

--

This actually brings out another point. The benchmarks OP used tell us nothing about how i5 2500K compares to the 6700K in the real world. That's why generally all synthetic benchmarks that don't use real world programs/apps are worthless for extrapolating real world performance differences.

 

[wingman04]

Memory scaling is vary little with skylake. There is a 25% average improvement from sandybridge to skylake according to Anandtech.

Our Skylake does not look too good here compared to our Haswell (1% increase). 9% faster than Ivy Bridge. 13% faster than Sandy Bridge. When we take our Skylake up to its stock 2133MHz speed we see an increase of 5% over Haswell, 13% over Ivy Bridge, and 17% over Sandy Bridge. 2666MHz memory gives us a 10% increase over Haswell.

http://www.hardocp.com/article/2015...76700k_ipc_overclocking_review/6#.V3vjko-cGUk

 

[frozentundra123456]

Memory scaling is vary little with skylake. There is a 25% average improvement from sandybridge to skylake according to Anandtech.

http://www.hardocp.com/article/2015...76700k_ipc_overclocking_review/6#.V3vjko-cGUk

Huh? The very charts you linked showed a 13 and 11 percent gain respectively in the two games tested going from 2133 to 3600 ram.

Personally, I feel the data is somewhat conflicting about the memory scaling with Skylake, but certainly there is enough data to indicate that saying "scaling is very little" is a very doubtful conclusion.

 

[guachi]

I guess if you have $600 for a high end GPU the cost of a 6700k + motherboard + RAM for 30% more FPS over a 2500k is worth it.

If I had a 2500k I'd just get a 390 or equivalent (or whatever can push my CPU) and call it a day. That much money can buy a lot of games. Or lots of other stuff. I mean, the older I get the more things I can find to spend money on.

5.5 years for 30-50% improvement just seems kind of sad.

 

[Roland00Address]

Huh? The very charts you linked showed a 13 and 11 percent gain respectively in the two games tested going from 2133 to 3600 ram.

Running at 640x480 resolution on low settings, at 300+ fps for all tests are using the literal best case scenario where ram bandwidth will translate into fps gain. Once you increase the resolution other bottlenecks will appear and ram will have a mucher lower gain.

 

[lenjack]

An oc'd 2500k is an awesome animal!