Ryzen Sandy Bridge IPC

[humbug]

@humbug you ignored the entire article page I linked and you extrapolated IPC in Cinebench as being the norm, is it really that difficult to read to the bottom of the page where they actually give you a % jump in IPC based on several workloads?

Literally at the bottom of the page:
Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up
Haswell to Skylake (DDR3): Average ~5.7% Up

Its a discussion forum, in future if you want to point someone at something specific do that instead of simply posting an entire article without the necessary context and then complain because persons didn't spend 2 hours reading the whole thing, like a ####.

Cinebench is a good indicator which is why its what i used scrolling through it trying to find something like that, the numbers you wanted me to look at are not wildly different to the numbers i used from Cinebench, you're loosing you're cool over semantics.

Anyone would use the first relatable thing they came across instead of reading the whole thing. pepole do have lives outside of this forum

Its just a forum, relax.

Typical humbug response

Same goes for you Gavin, your constant condescending tone with everyone on this forum is completely unnecessary and getting very tiresome.

 

[gavinh87]

Not everyone, just a select few who refuse to listen to anything negative towards something they have no experience with.

 

[Scougar]

Looked up some figures for the 7700k ad saw that at stock it appears to get around 200 in single thread R15.

Now... if we take actual gaming into account, I wonder how that i5 2500k actually looks compared to them?

Being realistic.. most people only got to about 4.5Ghz with an SB. 5Ghz upwards was for those dedicated and trying to push things.

 

[jigger]

Looked up some figures for the 7700k ad saw that at stock it appears to get around 200 in single thread R15.

Now... if we take actual gaming into account, I wonder how that i5 2500k actually looks compared to them?

Being realistic.. most people only got to about 4.5Ghz with an SB. 5Ghz upwards was for those dedicated and trying to push things.

I think most people don't bother pushing for more TBH. Seem the master race started a myth that DDR3 1600Mhz was all you needed and that's going to make big overclocks a little pointless. Well for playing games anyway.

 

[AndreiD]

@Scougar My old 4.5Ghz 2600K scored about 158 in Cinebench R15 single, it's pretty far behind Kaby Lake (stock 7700K turbos to 4.5Ghz so it should be somewhat comparable).

But don't be fooled by humbug in thinking Cinebench alone is a good representative of IPC, you need lots of different workloads to determine a precise IPC difference between architectures.

And you're right, 4.5Ghz was usually what most people ran their Sandy chips at, 4.5 to 4.8 was usually possible at acceptable voltages depending on chip, but higher required some pretty extreme voltages and implicitly extreme cooling.

 

[Martini1991]

Cinebench is very good at determining per core performance as far as games as concerned. Cinebench is FPU intensive.
You're never going to get a precise IPC measurement.

 

[AndreiD]

Doesn't really seem to be the case given how Zen & Skylake-X perform in Cinebench vs Games.

 

[Martini1991]

Doesn't really seem to be the case given how Zen & Skylake-X perform in Cinebench vs Games.

Gaming is fpu, cinebench is a measurement of fpu.
Obviously running the MT benchmark isn't the whole story as games won't be using the cores that intensively. Which is why the single thread benchmark is the one that matters, that'll then scale up depending on how many cores the game can use etc.

 

[AndreiD]

A lot more things can impact gaming performance than pure FPU performance though.
Cinebench is good for what it is, a Cinema 4D benchmark that measures rendering performance.

 

[Martini1991]

A lot more things can impact gaming performance than pure FPU performance though.
Cinebench is good for what it is, a Cinema 4D benchmark that measures rendering performance.

I don't know what you're arguing against.

Are you saying the single thread benchmark of zen in Cinebench makes it seem too good or not very good? I don't think the single bench results are particularly aspiring and frankly fits in line with what I'd expect.

 

[AndreiD]

Yes it does, and it also does with Skylake-X in a more extreme fashion. Gaming isn't just about pure FPU performance, which Cinebench isn't even intended to be a good measure of, it's just supposed to be a Cinema 4D benchmark.

Look up some Skylake-X ST results in Cinebench for example and compare them with gaming reviews.

Skylake-X has the mesh latency & cache changes and Zen the infinity fabric latency as penalties which will inevitably bring their performance down in some workloads such as gaming.

 

[humbug]

Yes it does, and it also does with Skylake-X in a more extreme fashion. Gaming isn't just about pure FPU performance, which Cinebench isn't even intended to be a good measure of, it's just supposed to be a Cinema 4D benchmark.

Look up some Skylake-X ST results in Cinebench for example and compare them with gaming reviews.

Skylake-X has the mesh latency & cache changes and Zen the infinity fabric latency as penalties which will inevitably bring their performance down in some workloads such as gaming.

Those Ryzen Fabric latencies are overcome by not using slow RAM, Ryzen easily matches its Cinebench MT performance in games where more than just a couple of threads are used. in some games a 4Ghz 1600 easily matches a 4.7Ghz i9 7800X, i wouldn't say that's in any way lacking gaming performance.

 

[humbug]

Those Ryzen Fabric latencies are overcome by not using slow RAM, Ryzen easily matches its Cinebench MT performance in games where more than just a couple of threads are used. in some games a 4Ghz 1600 easily matches a 4.7Ghz i9 7800X, i wouldn't say that's in any way lacking gaming performance.

A slice of fairly new games with fast low latency ram

Its not as fast as the mighty 7700K but that ain't bad at all, its not bulldozer, not by a long way.

This one is the worst result.

 

[CAT-THE-FIFTH]

I would argue gaming performance of Ryzen is being masked by some issues:
1.)The need for higher speed RAM to overcome inter-CCX latency which will be compounded by games not allocating threads taking into account the different arrangement. Many games will be coded for the Intel way of doing things,so this is why AMD has sent out so many Ryzen kits to devs.
2.)SMT can lead to regressions in performance in certain games. Intel HT is better understood quantity and even that sometimes has caused problems.
3.)Clockspeeds.

These are all fixable issues with the basic CPU implementation either via later revisions of Ryzen or devs better understanding the design of the CPU.

I mean SKL-X has well characterised cores and using cores which are a known quantity,and with a different cache arrangement and memory controller lead to performance not looking as good as people expected in many games,since it moved away from what games were being coded for in the past.

The Ryzen cores themselves have a decent level of IPC as indicated by software like Cinebench which was long considered the gold standard especially during the Bulldozer days,and has tended to perform better on Intel even before that.

 

[Zeed]

IPC's again ??
There You go all cpus on max oc they could handle 24/7

2500k @ 5ghz with tuned ddr3 @ 2xxxmhz cant remember number

 

[humbug]

IPC's again ??
There You go all cpus on max oc they could handle 24/7

2500k @ 5ghz with tuned ddr3 @ 2xxxmhz cant remember number

Thats pretty bloody good. especially when you consider 158 Points at 4.5Ghz on AndreiD machine, You're scoring 23% higher with an 11% higher clock rate, one of these must be wrong and i have no idea which.

@Scougar My old 4.5Ghz 2600K scored about 158 in Cinebench R15 single, it's pretty far behind Kaby Lake (stock 7700K turbos to 4.5Ghz so it should be somewhat comparable).

But don't be fooled by humbug in thinking Cinebench alone is a good representative of IPC, you need lots of different workloads to determine a precise IPC difference between architectures.

And you're right, 4.5Ghz was usually what most people ran their Sandy chips at, 4.5 to 4.8 was usually possible at acceptable voltages depending on chip, but higher required some pretty extreme voltages and implicitly extreme cooling.

 

[Zeed]

Mine are all good. Same Cinebench folder since Phenom II time. Got allot of 2500k scores ALL ARE ON 5hz only difference is DDR3 bottom is stock 1600mhz if iremember ofc legendary DDR3 Samsung green with added radiators for looks mostly. Basically overclocking ddr3 and fine tunning timings to maximum stable settings gave as you see 0.05 ipc gain

My 5820k was **** clocker so tahts on 4450mhz with 2666 ddr4 max tight timings and CL14 could not get anything better out of CPU or DDR's no matter what Volts i'w put on it :/


All in all if one got 2600k that will do 5ghz ( like all of them if u got good cooling) Ryzen is waste of money for gaming FACT.

 

[gavinh87]

Thats pretty bloody good. especially when you consider 158 Points at 4.5Ghz on AndreiD machine, You're scoring 23% higher with an 11% higher clock rate, one of these must be wrong and i have no idea which.

There are loads of ways of increasing a cb score. Setting the priority to real time is one.
But yeah it is a good result, the 1700x getting mullered by a CPU released in 2011.

@Zeed I completely agree, Ryzen as 60hz is fine. 144+ requires more horses.

 

[Scougar]

I don't know how the scores compare/translate between versions btw.

Also.. don't forget that you need to compare these SB's in games as well. Instruction sets are added as time goes by that increases performance. As mentioned, SB gets it's ass kicked in games like PS2 by slower clocked 3xxx series chips, regardless of the speed shown above.

SB was and still is impressive, regardless.

 

[Zeed]

Thats pretty bloody good. especially when you consider 158 Points at 4.5Ghz on AndreiD machine, You're scoring 23% higher with an 11% higher clock rate, one of these must be wrong and i have no idea which.

You missed that its CB 11.5 not cb 15. Cant remember if ui got Sandy on cb15.

I had a look and I dont have CB15 screenshot uploaded anywhere. But ill do 3 runs with my courent Overclock to see how that looks and post it.
Atm i got 3950 with 3333cl14 and max tuned timings on my 1700x.