CPU Shootout: i5-760 vs i5-4670K - (updated w/760@3.5GHz) Where's the progress?

[wand3r3r]

3DMark looks good, Deus ex ok, then everything else is pretty surprising (to me).

http://techbuyersguru.com/i5CPUshootout.php

Am I the only one shocked by the (lack of) progress of processors since 2010, particularly for gaming?

I can imagine there are scenarios which are optimized for new instructions, but current games sure aren't. It's almost like a one horse race but the lead horse has stopped to graze midway and the loser is somewhere in the dust cloud with no certainty about it's whereabouts.

E* they added the 760 @ 3.5 GHz

 

[SithSolo1]

I don't think I've seen a comparison that in-depth in years.

Love this part:
"Well, take a look at that! Except in our two outlier cases, the CPU-bound Deux Ex and the CPU-intensive BF3 multi-player, the combination of an i3-3220 dual-core and a GTX 670 gets the best of the ultra-modern i5-4670K paired with the midrange HD 7870. Sure, the ideal scenario would be a 4670K paired with a GTX 670 (or heck, how about an i7-3960K and a Titan as long as we're dreaming!), but for real people spending real hard-earned cash, this is a vivid illustration of why CPU power really takes a back seat to video card prowess when it comes to games."

Except where it actually matters, it doesn't. Who knew?

Most people with a 4670k aren't going to be pairing it with a 7870.

 

[crashtech]

Interesting, I was trying to argue a similar point in another thread a while back. Tests of this nature are not common, so it's interesting to see.

 

[tential]

I'm surprised by the lack of CPU power needed for modern gaming.

 

[crashtech]

Most people with a 4670k aren't going to be pairing it with a 7870.

Yeah, but the takeaway is that for gaming, it's actually best to heavily skew towards the GPU side. In the aforementioned thread, the OP in that instance is planning on buying an Ivy i5 or similar to mate with an existing 7770. I thought it might not be the best plan.

 

[Silverforce11]

You could get a Q9400 (or Phenom II) and OC to around 3.4ghz and it will run BF3 64 player MP 60 + fps consistently, I know, I've used such a setup.

As long as its not a crippled dual core, any quad will do and unless you game at 720p, the bottleneck at 1080p is solely on the GPU. This is true even if you put in a Titan, once you crank AA to 8x and details to the max, there's plenty of games where even Titan struggles to hit 60 fps.

The only situation where this isn't true is in CF/SLI setups and trying to achieve 120 fps for special monitor setups. But there are games where even i7s cannot hit 120 fps (single threaded games for example), such is the poorly optimized code (or lack of thread scaling), these are always the odd ones out and not the norm. Im playing two of these games right now, World of Tanks (1 thread) and Fallen Enchantress (1 thread), FE runs like crap even on high end i7s with SLI titans, because its so badly coded.

Gaming has always been about the GPU, the CPU plays a secondary role as long as its got enough cores.

 

[SPBHM]

I'm surprised by the lack of CPU power needed for modern gaming.

well, testing like this is always limited, I recommend searching for more, there are other games far more demanding then the ones tested, and tests like this normally don't capture the worst situations during the game... but sure, the old i5 should still be good enough for gaming in most cases.

 

[Silverforce11]

well, testing like this is always limited, I recommend searching for more, there are other games far more demanding then the ones tested, and tests like this normally don't capture the worst situations during the game... but sure, the old i5 should still be good enough for gaming in most cases.

That test has many of the top recent games. The oldest is Deus Ex.

The "problem" with that testing is that its running maxed settings at 1080p, the GPU is the bottleneck here pretty much always. In essence its not a good test to show CPU limitations, for those, 720p or less will reveal that indeed, newer CPU give much higher numbers. But its not relevant, the numbers are a) too high and b) not obtained even by top GPUs at 1080p. It MAY be informative for 3 years down the road for a GPU upgrade, but even then, that's iffy.

I have an old Q9400 rig (3.4ghz) and if I plug in this gtx670, crank up graphics details at 1080p, its not gonna matter squat compared to my i5... and that is a very ancient CPU by todays standards.

In recent titles, I've only seen CoH2 that is extremely CPU bottlenecked. Other games at 1080p maxed, even top GPUs struggle.

 

[boxleitnerb]

High graphics settings, relatively slow GPU -> (at least a partial) GPU bottleneck. For example the difference in BF3 multiplayer is much much larger if properly tested:
http://www.pcgameshardware.de/Battl...ld-3-Multiplayer-Tipps-CPU-Benchmark-1039293/
About 50% for the 4670K (it's 10% faster than the 3570K)

 

[AtenRa]

High graphics settings, relatively slow GPU -> (at least a partial) GPU bottleneck. For example the difference in BF3 multiplayer is much much larger if properly tested:
http://www.pcgameshardware.de/Battl...ld-3-Multiplayer-Tipps-CPU-Benchmark-1039293/
About 50% for the 4670K (it's 10% faster than the 3570K)

That test was run at 1280x720 (720p) not 1920x1080 (1080p).

But for BF3 MP its all about consistency at 60fps and time frames. I would prefer to have a Core i7/FX8350 with HD7850 than the Core i3 with GTX670.

 

[boxleitnerb]

That test was run at 1280x720 (720p) not 1920x1080 (1080p).

But for BF3 MP its all about consistency at 60fps and time frames. I would prefer to have a Core i7/FX8350 with HD7850 than the Core i3 with GTX670.

So? It shows the potential of the CPUs. Not everyone wants to play with 40fps minimums in multiplayer. And people have faster graphics cards than a 7870. Or are willing to turn down AA for better fps. As I always say: Maximum amount of information, else don't bother benchmarking at all. People should be able to combine CPU and GPU benchmarks for a complete picture. If they can't, they are stupid. If they don't want to, they are lazy. Either way, that's the only sensible way to do these things and please everyone.

 

[AtenRa]

So? It shows the potential of the CPUs. Not everyone wants to play with 40fps minimums in multiplayer. And people have faster graphics cards than a 7870. Or are willing to turn down AA for better fps.

Or, if you dont play BF3 MP you will be better with Core i3 and GTX670. It all depends on the games you play.

Also, if you turn down the AA you will get more than 60fps even with the Core i3.

 

[boxleitnerb]

Or, if you dont play BF3 MP you will be better with Core i3 and GTX670. It all depends on the games you play.

Also, if you turn down the AA you will get more than 60fps even with the Core i3.

AA doesn't have any influence on CPU load. None at all. Besides, the PCGH benchmark I posted clearly contradicts you. For constant 60+ fps on the CPU side you will need a true quadcore Sandy Bridge or better.
Nowadays I would never recommend an i3. There are just too many games out there that already profit from 3+ (true) threads. I'm not an advocate for speculating and futureproofing too much, but the evidence that an i3 is not enough for many games today is already here.

 

[ShintaiDK]

I went from a i7 860 to a 3570. And it was night and day difference in games.

760 is simply too slow unless you play games with very light CPU load. And using scripted runtime benchmarks is not exactly the best to showoff the CPU.

 

[BallaTheFeared]

I used a i3-540 @ 4-4.3GHz with a 7950, it did fine in some games, was pretty bad in others.

My thoughts as a gamer are a bit conflicted, but in the end I would say get the i5 over the faster GPU /w i3..

I know this isn't intuitive, but you can change graphical settings, often with minimal loss in IQ for a fairly decent gain in FPS. There are very few settings, mostly view distance and shadow related that have any effect on boosting CPU performance. There is no setting to make up for a lack of cores, currently 2 is ok in most, but 4 is nearly required for some.

So my thought is you can use settings to get framerates that are playable with a vast array of GPU options however there is little that can offset cpu needs, dual core is ok in some not so much in others - unlike GPUs though there is virtually nothing you can do to help it.

Add to that GPU's nearly double each generation in performance, it's hard for me to suggest gimping your platform in pursuit of a higher performance gpu which will be outdated, and half as fast as next years product at the same price point.

 

[AtenRa]

AA doesn't have any influence on CPU load. None at all. Besides, the PCGH benchmark I posted clearly contradicts you. For constant 60+ fps on the CPU side you will need a true quadcore Sandy Bridge or better.
Nowadays I would never recommend an i3. There are just too many games out there that already profit from 3+ (true) threads. I'm not an advocate for speculating and futureproofing too much, but the evidence that an i3 is not enough for many games today is already here.

Im not talking about the BF3 MP, as i have already said, BF3 MP needs more than 4 Threads and Core i3 is not recommended.

 

[boxleitnerb]

Im not talking about the BF3 MP, as i have already said, BF3 MP needs more than 4 Threads and Core i3 is not recommended.

There are lots of other games where the i3 fails compared to an i5. It just doesn't make sense today.

 

[mikk]

3DMark looks good, Deus ex ok, then everything else is pretty surprising (to me).

http://techbuyersguru.com/i5CPUshootout.php

Am I the only one shocked by the (lack of) progress of processors since 2010, particularly for gaming?

I can imagine there are scenarios which are optimized for new instructions, but current games sure aren't. It's almost like a one horse race but the lead horse has stopped to graze midway and the loser is somewhere in the dust cloud with no certainty about it's whereabouts.

GPU limited gaming tests as usual. Nothing to see here, move on.

 

[Vesku]

Shame they didn't also include the 760 OCed, a pretty good showing for its age and stock 2.8GHz.

 

[Silverforce11]

An AMD A10 is smoother in BF3 MP than i3, and that's because BF3 is one of those games that have been shown to love a minimum of 4 cores (real cores, not 2 + 2HT) and scaling up to 6 cores.

Newer games tend to be like this as well, although there will always be exceptions and games come out using 1 thread. The point is, you CAN skimp on the CPU, but it needs to be a quad core minimum.

 

[frozentundra123456]

Changing the CPU in gpu bound games has little effect. Wow, I would have never er guessed that.

 

[BrightCandle]

The problem with this (and many other tests) is they are extremely limited in their scope but they are used to conclude much more broadly. The essence of the problem is that some games are clearly need more resources than others. Any random selection of games is going to potentially show the full gamut of the range of results or it isn't, and in this case it clearly does not.

This is why I like what gamegpu.ru is doing, they are testing a lot of CPUs at realistic settings along with testing lots of GPUs. Testing the combination of both however is what we really need, and we need it for every game. Its just an impractical number of data points.

 

[SiliconWars]

Most people with a 4670k aren't going to be pairing it with a 7870.

They probably will be pairing their CPU with a card of that level, or likely lower. The percentage of buyers who pay more than $300 for a graphics card is tiny.

 

[Quantos]

I don't get the point of this test. Isn't simple logic respected here? GPU limited games will fare fine with an older CPU, whereas CPU limited games will not.

 

[NTMBK]

I'm surprised by the lack of CPU power needed for modern gaming.

And yet people still complain about the consoles using Jaguar...