-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
[ARTICLE] Every Core i5 generation benchmarked from Lynnfield to Skylake
- Thread starter Random Guy
- Start date
Soldato
- Joined
- 9 Jun 2011
- Posts
- 3,635
Looking at this thread I'm glad I went 5820k over 6700k
Did you post that article intending it to prove or disprove my point????
I don't speak Polish which is probably true for pretty much everyone else on this UK based forum? I'm sure if there was a valid point to be made an article could be found in English to support it given the amount of reviews and comparisons that come from either the UK or America?
As far as I can tell the article was mainly pitched at asking whether it was worth buying DDR3 or DDR4 for Skylake
The simple answer appears to be clock for clock that there is no significant difference say between DDR3 2400 and DDR4 2400
'Since DDR3 and DDR4 the same settings for Skylake offer identical performance'
(courtesy of google translate)
I notice as well that you link to an article that compares DDR4 2400 to DDR4 3200 with the 6700K when I already linked to articles that said that DDR4 2666 was optimal for price/performance!
'We saw some nice performance gains from DDR4-2133 to DDR4-2666'
DDR2400 is pretty slow to be using with Skylake I would agree... you may as well use DDR4 2666, 2800 or 3000 with Z170/ Skylake because there's not much of a price differential up to DDR4 3000Mhz modules and performance shows distinctly diminishing returns to the point of flat lining above 3000Mhz sure slap some real slow 2133Mhz or 2400Mhz memory onto an overclocked 6700K and you'll suffer to a degree....
Conversely go from say a 3000MHz Corsair Kit to a 4000Mhz Corsair Kit and you'll be spending an extra 100 notes for 8gb of memory alone with no tangible performance gain whatsoever! My post previous post was in response to someone pining to run 3400Mhz ran as a minimum or preferably 3800Mhz ram! = Total waste of money
But even slightly gimping Skylake with slow memory did not make a whole lot of difference overall and anyway it would be a silly false economy to save a whole six pounds or so from going from 8GB DDR4 3000Mhz to 8GB of DDR4 2400Mhz
But still rather pointlessly gimped results ahoy!
BF4 1920x1080 Ultra/HBAO [email protected]
DDR4 3200 163.5/155
DDR4 2400 156.4/148
Crysis 3 1920x1080 Very High AFx16 [email protected]
DDR4 3200 75/71
DDR4 2400 73.5/70
FC4 1920x1080 Ultra HBAO+ [email protected]
DDR4 3200 82.3/71
DDR4 2400 77.5/68
Watch Dogs 1920x1080 Ultra HBAO+ High [email protected]
DDR4 3200 78.4/76
DDR4 2400 74/71
ARMA III 1920x1080 Ultra SSAO FXAA [email protected]
DDR4 3200 44.9/40
DDR4 2400 43.3/39
Skyrim 1920x1080 Very High AFx16 AAx8 [email protected]
DDR4 3200 69.4/68
DDR4 2400 67.9/66
StarCraft II 1920x1080 Ultra & Extreme [email protected]
DDR4 3200 96.8/73
DDR4 2400 93.4/70
GTA V 1920x1080 Very High AFx16 FXAA [email protected]
DDR4 3200 67.3/61
DDR4 2400 64.3/58
So in summary its probably a false economy to buy the cheapest 2133 or 2400Mhz ram to run with Z170/Skylake as it wont cost you much more to just get 2666-3000MHz or so RAM but even if you do buy 2400Mhz ram you wont suffer that much in the grand scheme of things gaming wise. Conversely spending $$$$ to get 'fast' DDR4 (i.e. 3866+Mhz or so) wont net you any worthwhile improvements for a far larger jump in price!
Last edited:
this tells us how much Intel have held back innovation across the IT industry for the last 7 years. And now they layoff 11% of workers because of the stagnant PC sector.
Maybe if they released faster CPU's, people would buy PC's again.
Maybe if they released faster CPU's, people would buy PC's again.
Its the physics of using silicon that's holding back chip progress... Intel are hardly alone in suffering from limited gains in progress in recent years...
In total agreement.
For what gain? Either they support a 3rd uarch for mass prod (atom, current mainstream and server, and this hypothetical performance uarch) or update all their mainstream/server lineup to get better roi on R&d but produce larger more expensive to manufacture dies. Either way Intel would charge a fair bit to make it worth their while to produce larger dies and to preserve their margins. Who would pay, we are a fairly insignificant segment of the general consumer market - which comprises the masses with fast enough hardware and no need to upgrade from anything since core2 except for formfactor/functionality reasons. Which would be undermined with more power hungry and higher heat generating cpu's.
Its still around 6% IPC.Its more than you get from tick-tock strategy.
And 3200Mhz ist not that fast.You can go with 4000Mhz DDR4 with skylake.
That will add 10-12% IPC vs 2400DDR4.
its pretty big actually.
Also look like skylake is bottleneck by 1600Mhz ram.Compare 1600mhz vs 3200Mhz.
Last edited:
Permabanned
- Joined
- 9 Aug 2009
- Posts
- 12,234
- Location
- UK
I wont repeat my original post again. All the tests are being run (including the ones you posted) with max details. Re-run the tests in 1080p with low detail.
Lynnfield still shows strong numbers even 8 years later. Lynnfield CPUs in 2016 are about the same age as the early Willamette Pentium 4s in 2008.
In real world gaming my 2500k smoked my old lynnfield i7 860. Can't explain it in light of these benches.
quad core upgrade from 4ghz exposedI KNEW IT ! pats trusty i5 750 @ 4.2 ghz and also 60hz IPS screen.
I guess one day i'll use that z97p-d3 though, maybe when I can bag a 4790k for £100 or less.
i saw a i5-750 on sale for $50, i thought wow thats crazy value
thousands and thousands, maybe hundreds of thousands of players who play SC2 and HOS. As the engines use two cores, running in ultra gives big lag spikes and it is hard to see what is going on.
In HOS for every 800mhz on the ram I get a 10% frame rate boost regardless of CPU clock.
I built a rig specifically for these two games as they are CPU/RAM clock bound.
for 3200mhz DDR4 ram vs 1600mhz ram i get a 20% performance boost (my rig runs at 3700-DDR4 C18).
The difference on the average between my Sandybridge-E rig and skylake is around 90fps vs 125fps. I need to do some fraps captures to get the full story because the biggest difference is minimum frame rates which are roughly half the average fps. I play with heroes where you need to press a specific key during team damage spikes, being able to land this key-press at the right time is essential to winning and so 45fps vs 70fps+ is a really big deal.
Last edited:
do u mean HOTS? i duno what HOS is >.<
if so yeh but its not too bad, mostly its when they mess with effects or add new stuff that isnt optimized
i cap 120fps but sometimes yes there is crazy dips
i think a more extreme example would be ARK, survival of the fittest which i got into recently seems to only use a single core, in fight with a decent view distance its not much above 60fps in the wait room can get 140+
these badly coded games shouldnt be about still but yeh
bliz has no excuse with all the money they have
if so yeh but its not too bad, mostly its when they mess with effects or add new stuff that isnt optimized
i cap 120fps but sometimes yes there is crazy dips
i think a more extreme example would be ARK, survival of the fittest which i got into recently seems to only use a single core, in fight with a decent view distance its not much above 60fps in the wait room can get 140+
these badly coded games shouldnt be about still but yeh
bliz has no excuse with all the money they have