Who's the daddy? Cores > IPC & Frequency ?

Holst1981 · 25 Feb 2017 at 10:07

SiDeards73 said:
The point i was trying to make above, badly it seems, was say Intel IPC is worth 100 units per 1ghz, and AMD Ryzen is worth 90 units per 1ghz, Intels HT is worth 50 units per 1ghz, but AMDs SMT is worth 70 units per ghz

Intel 7700k @ 5ghz all cores would be worth = 750 units
Intel 6900k @ 4ghz all cores would be worth = 1200 units
AMD Ryzen @ 4ghz all cores would be worth = 1280 units

all totally theoretical of course and probably a very bad example

You could work that out in theory
But it's completely application dependent.

Fundamentally even games that do benefit from more 8 cores will still run fine on a 4 core cpu.
Likewise games that don't even use 4 cores will be fine on a 8 core chip, especially if you overclock.
For gaming it won't matter very much if you go for 4 or 8 cores.

As for ryzen if it overclocks well then it might be a good option.

I wait a bit and see how the guys who pre-ordered get on.

If you have loads of money buy x99 and a 1080.
If you don't you will do better to buy a faster gpu and cheaper cpu.

Glanza · 25 Feb 2017 at 10:18

SiDeards73 said:
Do you understand why they test at 1080p? it is to make the game CPU bound, higher resolutions mean the game becomes GPU bound.... do some research before making posts like this, you will get ridiculed

Oh no not ridiculed :eek:

please save me oh great SiDeards73!

geordiejc · 25 Feb 2017 at 10:23

Holst1981 said:
On a range of games they are very similar.

If you play a particular game that needs a strong single core 7700k is better.
If you play a particular game that's mulithreaded a x99 chip is better.

Fundamentally they are all going to be ok.
Then it's down to cost / performance.

X99 really shows its advantage in other uses, if your encoding video or doing anything that scales with cores x99 will be better.

I would wait for ryzen reviews, it may be a good balance of cost cores and performance.

Agreed, x99 is better for encoding and other HT tasks as physical cores > logical cores but idk about gaming. 99.9% of people won't buy a CPU for a single game and with the Ryzen 8 cores to the masses strategy, i'm not convinced there's a gaming market to be had, hence why i want to understand across a range of popular titles if there's any significant benefit to be had of adding extra cores at the same frequency & IPC.

Bliddy tough subject to crack this one due to the variance being the software.

8 Pack · 25 Feb 2017 at 10:30

Depends on game. Most games need ipc. Some need 4 cores. Very few need more or HT

Glanza · 25 Feb 2017 at 10:32

geordiejc said:
Agreed, x99 is better for encoding and other HT tasks as physical cores > logical cores but idk about gaming. 99.9% of people won't buy a CPU for a single game and with the Ryzen 8 cores to the masses strategy, i'm not convinced there's a gaming market to be had, hence why i want to understand across a range of popular titles if there's any significant benefit to be had of adding extra cores at the same frequency & IPC.

Bliddy tough subject to crack this one due to the variance being the software.

Wouldn't it be down to the engine the game has been built on? Frostbite will make use of more cores but the engine Bethesda love to use seems to prefer using less cores with more speed.

geordiejc · 25 Feb 2017 at 10:33

What we need is an OC'd 7700k vs 6900k to the same frequency minus the IPC of broadwell/kaby to get a true reflection across those games tested, therefore comparing apples with apples.

It is my hypothesis is that the gains had of an 8c vs 4c @ same IPC/Frequency is minimal due to current game coding/engine practices when it comes to multi-core. I'm not sold on the Ryzen 8 core 16 thread CPU to the masses for gaming as i believe game engine/software optimisation for 4+ cores isn't mainstream enough to warrant an 8c CPU.

DONOHUE07 · 25 Feb 2017 at 10:48

hornetstinger said:
No such thing as "future proofing"

Also I'm still using a i7 950. And I'm not even overclocking it, still running at stock, games are fine. And will continue to use it till games are unplayable.

Can't see why your fx8350 is a badchip

Because it gets smoked by my friends older 2700k in pretty much all the games we play, same 8gb of ram and same 4gb r9 290, thus his older cpu is more future proof, he averages more FPS and more importantly for me a higher minimum FPS

Edit: as another poster rightly said, it's engine specific

StarShock · 25 Feb 2017 at 12:30

For gaming its also worth remembering that there are 2 pretty big bottlenecks ahead of the CPU. GPU and Monitor. Better CPU is better, but in a lot of real world situations anything from a 2500k and up is going to have fairly similar actual performance.

KompuKare · 25 Feb 2017 at 12:40

Well didn't do the calculations, but I put in the 1080P overall figures into a spreadsheet and uploaded it to ethercalc.org
https://ethercalc.org/ybhaa42m8n
Ethercalc is not to great but unlike googlesheets it doesn't need registration.
Put in the columns for all the individual games but didn't enter the data.
Anyway, put in a rough FPS per GHz calculation but with base and turbo even that's a bit hard to calculate. On the other hand the Broadwell-E scores do come up fairly similar per GHz of Turbo.
Thing is, unlike PCGH or gamegpu.ru they don't actually show CPU loads (although to be meaningful those would have be a timegraph).
From the article, it looks like they intend to use this review for the Ryzen results so look forward to that. Their also promising frametimes then so maybe CPU utilisation graphs too.
Their article is already at 679 comments, so I might have a look later if anyone has asked for that.

Jumper118 · 25 Feb 2017 at 12:53

Price to performance is king.

kaku · 25 Feb 2017 at 12:59

The problem is compounded because whilst early tests may point to Ryzen/Skylake/Kabylake having roughly the same IPC, it's also been shown that Ryzen struggles to clock much higher than 4GHz (see: Gibbo's comments on overclocking the 1700).

'Effective' raw power of a CPU is determined by cores, IPC and operating frequency. However, the calculation becomes more complicated when developers don't use the resources correctly.

ljt · 25 Feb 2017 at 13:31

Jumper118 said:
Price to performance is king.

+1

If something offers 80% of the performance for 50% of the cost then I'll take the 50% cheaper product any time

geordiejc · 25 Feb 2017 at 13:40

KompuKare said:
Well didn't do the calculations, but I put in the 1080P overall figures into a spreadsheet and uploaded it to ethercalc.org
https://ethercalc.org/ybhaa42m8n
Ethercalc is not to great but unlike googlesheets it doesn't need registration.
Put in the columns for all the individual games but didn't enter the data.
Anyway, put in a rough FPS per GHz calculation but with base and turbo even that's a bit hard to calculate. On the other hand the Broadwell-E scores do come up fairly similar per GHz of Turbo.
Thing is, unlike PCGH or gamegpu.ru they don't actually show CPU loads (although to be meaningful those would have be a timegraph).
From the article, it looks like they intend to use this review for the Ryzen results so look forward to that. Their also promising frametimes then so maybe CPU utilisation graphs too.
Their article is already at 679 comments, so I might have a look later if anyone has asked for that.

Brill i hope they supply the data! thanks for the efforts so far!

geordiejc · 25 Feb 2017 at 13:47

ljt said:
+1

If something offers 80% of the performance for 50% of the cost then I'll take the 50% cheaper product any time

Guess you both missed this ;

"I'm not looking for a value for money comparison, that segways into another subject"

Killer7xx · 25 Feb 2017 at 14:09

This is basically HT vs cores again, and here again having more cores is superior. The difference in IPC between a i7 7700k and a R7 1700 is not significant enough to make a large difference for the majority of users out there even at 1080p. Having 2/4 extra cores will always result in more extra processing power than even a heavy overclock and will be more power efficient since each core can run a lot slower to provide the same amount of overall performance.

Panos · 25 Feb 2017 at 14:19

Gents, just google search some recent reviews using top of the range cards like the GTX1080. Not that difficult

15 games benchmark when you scroll the images under the header Number crunching with i7-7700K
http://www.pcgamer.com/intel-core-i7-7700k-review/

Another 5 game benchmark using GTX1080 and various CPUs

http://www.guru3d.com/articles_pages/core_i7_7700k_processor_review_desktop_kaby_lake,14.html

So the result is if you play current games cores with a good IPC is better.

As you will see on some benchmarks, 7700K struggles to keep ahead from multi core CPUs running 600Mhz+ lower
And if you plan to use anything less than top of the range graphic card, forget it. Buy what's better value for money.

Here is a benchmark using 290X and GTX980
http://www.anandtech.com/show/10968...w-review-the-new-stock-performance-champion/6

Yes the 7700K @4.5Ghz is beaten in many of them by the 5775C (Broadwel) running at 3.7Ghz!

CAT-THE-FIFTH · 25 Feb 2017 at 14:22

I still remember the debates between the E8400 and the Q6600,etc. But don't worry when Coffee Lake is out with 6C mainstream CPUs,MOAR cores will be essential for gaming.

geordiejc · 25 Feb 2017 at 15:33

A test which will settle this debate is getting a 7700k and a 6900k at the same frequency (overclock) and testing 15 or so popular 2016/17 titles for average FPS. I believe more cores for gaming, curretly makes little difference i.e. 2-7 fps but if the frequency & IPC of each CPU is equal this could show some substantial fps gains e.g. 4.5ghz.

If OCUK would simply send me 1 x x99 motherboard & 6900k and 1x z270 & 7700k cpu i can get to work... :rolleyes:

CAT-THE-FIFTH · 25 Feb 2017 at 15:39

I don't see how hard it is to read the OP - all those 6C and 8C CPUs have lower IPC and much lower core clockspeed than a Core i7 7700K.

Thats the issue - those 6C and 8C HEDT CPUs will not be boosting that high. The problem is that a Core i7 7700K overclocked to 5GHZ is only 10% higher than its maximum boost clockspeed. A Core i7 6800k which is overclocked to 4.3GHZ will be running at 20% higher than its maximum boost clockspeed.

geordiejc · 25 Feb 2017 at 15:45

CAT-THE-FIFTH said:
I don't see how hard it is to read the OP - all those 6C and 8C CPUs have lower IPC and much lower core clockspeed than a Core i7 7700K.

Thats the issue - those 6C and 8C HEDT CPUs will not be boosting that high. The problem is that a Core i7 7700K overclocked to 5GHZ is only 10% higher than its maximum boost clockspeed. A Core i7 6800k which is overclocked to 4.3GHZ will be running at 20% higher than its maximum boost clockspeed.

Hey, mind if ask you to clarify what you mean ? I'm not following. i'm suggesting if the 6900k and 7700k's frequency was the same, what would be the performance gain in a range of games due to the extra cores of the 6900k? (as broadwell vs kaby IPC gives ~1 fps)