Why is AMD so shit

>>60372461
http://www.guru3d.com/news-story/new-cpu-z-upgrade-lowers-ryzen-performance.html lol.

>>60372461
new cpuz seems to gimp everything but kaby
even skylake looks shitty now
i5 7600k is +7% sc and only -18% in multi core
compared to an i7 6700k that has the same 4.2 turbo sc and 200mhz higher clock speed + HT for mc. Taking the 5% difference in speed, ht only scales 13%? why are i7 shilled?

>>60372644
good thing I'm on 17.01

File: cpu-z.png (69KB, 810x404px) Image search: [Google]

69KB, 810x404px

>>60372686
>1.79.0.x64 is literally right there on the screenshot
>17.01.64 is the benchmark version
You're either a really bad liar, or a retard, and I don't know which is worse at this point.

>>60372800
I ment the bench version clearly

File: newcpuz.png (48KB, 803x398px) Image search: [Google]

48KB, 803x398px

I don't think newer intel looks much better
look at an overclocked haswell-e 8c vs a stock raisin 1800x
+68pts single thread pts +16%
+200pts multi thread +4.4%
+$500 retail +50%

>>60372646
what if intel pays benchmark software to promote themselves? oh thank god intel never does shady marketing :)

>>60372461
>>60372644
>>60372646
>we modified our algorithm to make Intel look better

File: CPU-Z benchmark.png (15KB, 403x402px) Image search: [Google]

15KB, 403x402px

>>60372461
>two CPUs
>both OC'd to (estimating your score/my score * my clock speed) 4.5GHz
Of course it's better m8. You kill the 5960x too.

File: fx gimped.png (59KB, 1203x402px) Image search: [Google]

59KB, 1203x402px

>>60373231
>>60373254
Reason for update was blah cache something. But many non Ryzen cpus were affected as well. The old FX chiips and well, everything that isn't the newest Intel® CPU is scoring lower compared to the hedt chips that are in both versions. You can test it yourself. I am interested to see if kaby scores higher than it did in 1.78

>>60374460
Wow it's almost like CPU-Z's benchmark is a joke and we should all just ignore it exists. I didn't think someone would make a benchmark that makes pissmark look good, but there you go.

File: kek.png (289KB, 596x628px) Image search: [Google]

289KB, 596x628px

>>60374460
>Ryzen is performing well because their benchmark is cache intensive and Ryzen has a really huge, fast cache
>Better make cache less important

Intel outperforms Ryzen in AVX instructions, next step is to make the benchmark entirely AVX based

File: for science.png (74KB, 1205x397px) Image search: [Google]

74KB, 1205x397px

>>60372461

>>60374822
they gimped the 6700k as well

BUY OUR NEW STUFF ALREADY.We changed sockets, We took away limited oc whith haswell, we changed sockets, we removed bclk with skylake, we paid off cpuid WHAT DO WE HAVE TO DO?

>>60374822
Holy fuck. Does Intel's jewery know no bounds?

>>60375970
At least the actual performance of the legacy cpus is the same, only the perceived performance is lower.
>Looking at you Nvidia

File: jews.png (116KB, 830x866px) Image search: [Google]

116KB, 830x866px

>>60372644
>>60372461
>>60372686

File: THEY.png (66KB, 819x415px) Image search: [Google]

66KB, 819x415px

>>60372461
what have (((THEY))) done

>>60372461
>24 threads have faster multithreaded performance than 16
Who could have guessed

>>60378161
>2.144
B-B-B-B-B-B-B
BONFIRE!

>>60378372
yet it still runs cooler than 7600k housefires

>>60378401
And slower.

>>60374460
that's housefire tier

>>60374106
Mine are running at 4.5 as well.
What's your voltage at?

>5960x
you mean 6950x?

File: cpuz-5680-bench.png (16KB, 417x414px) Image search: [Google]

16KB, 417x414px

>>60378161
Benchmark used to be mostly for Intel, so cashe played a big part because the difference between the cpus was small.
Then AMD throws a fuckload of cashe on their chips and it greatly affects the benchmark.

Here's mine on a slower clock on the old version.
It has 24mb of L3 and something like 3mb of L2

>>60380864
Benchmarks shouldn't be changing at all

You make one, and that's that. It stays how it is forever. If you want a new benchmark you make a new one instead of changing the old one.

>>60378372
its a bug you dumbfuck look at the older version for the correct voltage.

>>60381143
how are we going to entice the goyim into getting a new socket if his archaic 6700k with a clock is still on par with the newest stuff?

>>60374576
>Intel outperforms Ryzen in AVX instructions, next step is to make the benchmark entirely AVX based
No anon, when Skylake-X drops the next step it's AVX512, anyone who doesn't support it gets 0 points

File: cpuz.jpg (346KB, 1062x698px) Image search: [Google]

346KB, 1062x698px

What does this mean?

>>60381690
Your chip it's shit now get a 7700K, then get a 7740X on launch day

>>60381690
The 7700K turbos to 4.5GHz for single threaded and stays at 4.2GHz for multi threaded. Your 6700K is overclocked to 4.6GHz for both single threaded and multi threaded.

>>60381143
So they should just leave it broken?
Fine, install the old version and have a fundamentally broken benchmark.

When I can beat a 10 core with my 12 core by 25% because I have more cashe, something is wrong.

>>60381907
That's absolutely incorrect, defaults to 4.5 turbo on all threads tho this is generally user customizable.

>>60381923
>So they should just leave it broken?
>Fine, install the old version and have a fundamentally broken benchmark.
It wasn't broken, the benchmark was like that
>When I can beat a 10 core with my 12 core by 25% because I have more cashe, something is wrong.
It's not, if the 12 core gets a shitload more cache misses than the 10 core it's obvious who will be running faster, execution resources will go unused in the 12 core while it waits to retrieve data
>>60381942
The default it's 4.2, but most mobo manufacturers set their own default at 4.5

>>60381942
It turbos to 4.4 for multi threaded, not 4.5
https://en.wikipedia.org/wiki/Kaby_Lake#Desktop_processors

>>60381971
No you idiot. Look at my post of my dual cpu desktop from 2011 beating a 6950x by 25%.
What the fuck cashe misses, it's the benchmark we are talking about, and there's your result.
>>60380864

>>60381976
Then my hardware monitor is fucking lying. The only change after a Bios update and reset was undervolt. Hwinfo or w/e showing me 4499 max values right now across all cores.

>>60382173
maybe because you have 2 cpus with better cache latency than a single 10c?

>>60382194
its an unlocked cpu, your mobo prob has a higher default setting. if you put it on a b250 it will run a t 4.2 mc and 4.4 sc

File: WP_20170513_22_32_10_Pro.jpg (2MB, 3264x2448px) Image search: [Google]

2MB, 3264x2448px

>>60381976
Shit i was wrong

Default settings, some retard browser bench in background

>>60372461
The Ryzen chip is killing it in multi-threaded and is cheaper overall, why are you shitposting?

>>60382249
>takes picture of monitor
whatever you say itt is now filed under T.

>>60382204
Actually yeah, maybe.
I just checked and the 6950x has about the same cache size.
fug i dont know

>>60382254
Hint: the higher score on multi is not the Ryzen
also these cpus are like $100 each on ebay. For cpu + motherboard, I think it was for $600

>>60382309
While sipping 300+W just for the mobo+cpu and getting 2P sheanigans
They're still great value, though the R7 1700 it's quite competitive with them

>>60382336
Well yes. this isnt for everyone, its more of a statement that a fucking 5 year old pc is on par with Ryzen core for core.
I haven't had issue with dual processors, except I have to give P1 more voltage then P0, same for the Dimm banks. Not sure why, but they are about 60mv out.

File: Ryzen Scores.png (27KB, 802x399px) Image search: [Google]

27KB, 802x399px

>>60372461
>>60382309
Here's my R7 1700 at 3750mhz

my singled threaded is higher than that 420 ish something just closed it..., lots of stuff in the background for multi-threaded I guess

RAM is 2400mhz 13-13-13-13-31

There is a huge difference though in terms of the ratio to a 5960X between the versions though

>>60382401
>Well yes. this isnt for everyone, its more of a statement that a fucking 5 year old pc is on par with Ryzen core for core.
It's not, Zen's the fastest architecture IPC wise ever made for x86

>>60382457
Ipc doesn't matter when you cant clock it up.

>>60382457
You people really believe that shit don't you?

>>60382486
1800X's 4.1 Ghz it's pretty fucking fast, just 400 mhz below your Westmere at 4.5, and 400 mhz below the 7700K turbo's
>b-but the 7700K can reach 5 Ghz on this golden samples!
>>60382524
It's the only explanation to STH benchmarks where the 1800X beats 6950X's and 16c 2P SB-E Xeon machines

>>60382548
Oh it doesn't stop at 4.5, I'm right at 80c on my cheap AIO liquid coolers. If I sort that out, I can push up the voltage a lot more and see how far my wallet goes. Should make 5.0 unstable but not too crazy high on the voltage.

>>60382548
Golden? Literally the vast majority hit it with 50bucks of cooling. All can be undervolted for high gains in effiency. Two cpus for two workloads. 7700k best choice for most tho.

>>60382660
>Golden? Literally the vast majority hit it with 50bucks of cooling
Yeah, that's why less than half of the 7700K on valid.x86.fr hit 4.8 Ghz, and now Intel it's telling it's consumers to not OC it
Keep shilling though, you will eventually get paid

>>60382681
CPU lottery has the numbers. Most people don't OC or they one click it via mobo and get stuck on 1.4V and low oc's

>>60382681
4.8GHz - top 100%
4.9GHz - top 91%
5.0GHz - top 62%
5.1GHz - top 24%
5.2GHz - top 4%

>>60382703
>Passed the ROG RealBench stress test for one hour
Totally stable systems

>>60382723

Serious question (not the other guy, just a newbie at overclocking). Is there a real difference between running a stress test for 30 minutes vs running it for a few hours? What would change in the CPU that would make it unstable suddenly after a couple hours?

>>60382772
Not necessarily. After a bit of time under stress it becomes impossible to quantify the failure points. Longer times are better but variety of stressors is even better. An hour is good, a few is too. 24 is likely overkill.

>>60382772
I don't even bother. I'll stress it with a benchmark a couple times, and call it good. If. It gives me issues, I'll throw more voltage or back it out a bit.
It will become evident if I have a problem quickly. But I run LLC, so my voltage is the same at idle and full, if it was going to crash, I don't think utilization matters, and it would just start crashing stuff randomly.

>>60381923
It's not broken though. It's a benchmark.

Here, let me provide you some clarification:

You want to benchmark some cars. How quickly can they reach a distance of 1000m in a straightaway? Some cars will do that in 15 seconds, maybe some in 20 seconds, or 10 seconds.

Then you decide that this benchmark isn't relevant for the real world and change the distance to 5000m; past data is no longer relevant but more importantly you're testing for something entirely different than before. It would make no sense at all to keep calling this the same benchmark as before, the scoring parameters fundamentally changed.

What groups who aren't being shady and underhanded would do is create a new benchmark and leave the old one how it is. People can run the old benchmark, or they can run the new one.

But that is not what happened with CPU-Z. They arbitrarily decided to change the metrics of their benchmark and only did so because of ryzen - had ryzen not come along and outperformed intel at this they wouldn't have changed anything. Their excuse was that ryzen had inflated scores because of their larger l3 cache; well you know what, maybe their larger l3 cache shouldn't be "accounted" for!

AMD, to take the previous example, designed a car with faster acceleration than Intel but a lower top speed, and CPU-Z then changed the parameters of their benchmark to favor higher top speed over higher acceleration and wham suddenly intel's in the lead again.

That is underhanded and shady as fuck.

>>60382457
no IPC is easy to code that's why and it has nothing to do with x86 code set.for its not the best for over all compute dumbass.

Look Your games might be revant now for you but other do more than that try to think about that for a tick.