FX-8370 beats i5-2500k in new games due to multithreading

[daerragh]

An interesting CPU test has been published by computerbase.de. AMD's CPU beats Intel's by 11% in new games that rely on multithreading more than older games.

I think that a new era of true 6 and 8 core will have begun in mainstream CPUs market with the release of new Ryzen CPUs with their good prices and Intel's i7 prices will lower afterwards, especially after the release of 6 core Ryzen in 2 half of the year.

https://www.computerbase.de/2017-02...-test/#abschnitt_performancerating_in_full_hd

 

[maddogmcgee]

Lol. Overclock both and see who wins. My old 2500k hits 4,5 ghz.

 

[ShintaiDK]

They decided to nerf the 2500K with 1333Mhz memory while giving the FX 1866Mhz memory.

21GB/sec on SB, 30GB/sec on FX.

"So unterstützt Sandy Bridge zum Beispiel maximal DDR3-1.333 und nur diese Geschwindigkeit wurde auch genutzt."

 

[NTMBK]

Meanwhile the i7-2600k beats both.

 

[Insomniator]

I'm a little suspicious of how bad the 2500k looks there, but meh I kinda hope its true -- I want an excuse to upgrade!

Of course its impossible to compare results when no one in history has ran a 2500k stock.

 

[RoarTiger]

They decided to nerf the 2500K with 1333Mhz memory while giving the FX 1866Mhz memory.

21GB/sec on SB, 30GB/sec on FX.

"So unterstützt Sandy Bridge zum Beispiel maximal DDR3-1.333 und nur diese Geschwindigkeit wurde auch genutzt."

Negligible performance difference in gaming due to memory speeds with SB arch. Anyone who has been a forum participant for that long knows better.

http://www.anandtech.com/show/4503/sandy-bridge-memory-scaling-choosing-the-best-ddr3/6

 

[ShintaiDK]

Negligible performance difference in gaming due to memory speeds with SB arch. Anyone who has been a forum participant for that long knows better.

http://www.anandtech.com/show/4503/sandy-bridge-memory-scaling-choosing-the-best-ddr3/6

Try move time forward 5 years. Not the same 5 year old games tested is it?

 

[SPBHM]

the strange thing is the FX matching the i7 4770k on total war warhammer and shadow warrior 2.

but yes, as others have said, pair the 2500K with DDR3 2133 and it's going to look different, also the 2500K has a lot more to gain with OC, personally I would easily pick the 2500K over the 8370.

Negligible performance difference in gaming due to memory speeds with SB arch. Anyone who has been a forum participant for that long knows better.

http://www.anandtech.com/show/4503/sandy-bridge-memory-scaling-choosing-the-best-ddr3/6

http://www.eurogamer.net/articles/d...it-finally-time-to-upgrade-your-core-i5-2500k

 

[ashetos]

Who the fudge overclocks anyway, nobody. Literally everybody I know do not overclock.

This result is significant, because it compares the 2 processors at stock settings, officially supported configurations and without violating warranty for both chips.

Official memory support for i5-2500K is 1066/1333 and for FX-8370 is 1866.

This is a result worth discussing.

 

[SPBHM]

Who the fudge overclocks anyway, nobody. Literally everybody I know do not overclock.

This result is significant, because it compares the 2 processors at stock settings, officially supported configurations and without violating warranty for both chips.

Official memory support for i5-2500K is 1066/1333 and for FX-8370 is 1866.

This is a result worth discussing.

I would think most 2500Ks were overclocked, since it was $20 more expensive than a CPU which performed exactly the same for non OC users.

 

[JimmiG]

Try move time forward 5 years. Not the same 5 year old games tested is it?

Memory bandwidth requirements are mostly a function of the CPU architecture, though. New software won't modify the SB execution pipeline to make it more bandwidth hungry.

The results still aren't really valid since everyone and their pet hamster has overclocked their 2500K's. It shows AMD were on the right path with Bulldozer, but the execution was a total disaster.

 

[ashetos]

I would think most 2500Ks were overclocked, since it was $20 more expensive than a CPU which performed exactly the same for non OC users.

20$ was worth it for moving from HD2000 to HD3000. I own the 2500 and I regret not having HD3000.

 

[Flapdrol1337]

any similar tests with an amd gpu?

 

[SPBHM]

20$ was worth it for moving from HD2000 to HD3000. I own the 2500 and I regret not having HD3000.

HD 3000 was terrible even when new, so I wouldn't pay $20 for that, if you were gaming back then you would buy discrete graphics, if you were not HD 2000 was the same thing, pretty sure people bought the K for the OC not for a less slow DX10 IGP which was still far from useable for gaming,

also, let's not forget the 2500K is like over half an year older than an 8150,
8370 is from what? Haswell days?

 

[ashetos]

HD 3000 was terrible even when new, so I wouldn't pay $20 for that, if you were gaming back then you would buy discrete graphics, if you were not HD 2000 was the same thing, pretty sure people bought the K for the OC not for a less slow DX10 IGP which was still far from useable for gaming,

also, let's not forget the 2500K is like over half an year older than an 8150,
8370 is from what? Haswell days?

The discussion we're having is off topic, it does not matter if HD 3000 is worth 20$ and it does not matter if 8370 is newer than 2500K.

What matters is that a slow 8-core CPU beats a fast 4-core CPU. And those are real chips at default settings, not some theoretical result. And this shows something.

 

[beginner99]

It's sad they did not include the 5775c because the HEDT chips could also be profiting from larger caches or quad-channel memory and not only from more cores. The 5775c could at least rule out the cache/bandwidth advantages or confirm them. Piledriver also has larger L3 cache so theoretical it could be due to caches and not cores.

EDIT:

And the 2600k also has more cache. So might not only be the HT.

 

[nathanddrews]

Most interesting to me is the abrupt fps gains moving from Sandy Bridge to Haswell and then again to Kaby Lake/Broadwell-E. Granted, they don't have Ivy Bridge on the chart so the gap seems more extreme. Definite "line in the sand" moments, though. Hopefully after the Ryzen review, they can add those results to these charts. My hype train is out of control at this point.

 

[antihelten]

the strange thing is the FX matching the i7 4770k on total war warhammer and shadow warrior 2.

Regarding TW: Warhammer, I also found it kind of peculiar that the 7700K was so far ahead of the 4770K (62%), I wonder what is going on in that game?

 

[SPBHM]

The discussion we're having is off topic, it does not matter if HD 3000 is worth 20$ and it does not matter if 8370 is newer than 2500K.

What matters is that a slow 8-core CPU beats a fast 4-core CPU. And those are real chips at default settings, not some theoretical result. And this shows something.

sure but it's relevant enough to mention that the 2500K without OC is the same as the 2500, the 2500K is significantly older and running significantly slower ram, if we had an 8150 it would be clearer.

Most interesting to me is the abrupt fps gains moving from Sandy Bridge to Haswell and then again to Kaby Lake/Broadwell-E. Granted, they don't have Ivy Bridge on the chart so the gap seems more extreme. Definite "line in the sand" moments, though. Hopefully after the Ryzen review, they can add those results to these charts. My hype train is out of control at this point.

this test included all the gens from sandy to skylake i5 and the 8350
https://www.youtube.com/watch?v=WZ_5p9wd2dk

 

[arandomguy]

Are there actually people who think there are no newer games that scale beyond 4c/4t?

 

[richierich1212]

Are there actually people who think there are no newer games that scale beyond 4c/4t?

The 2500K was a legendary CPU. People will still think it is the best performer today, even 6 years later.

 

[nathanddrews]

Are there actually people who think there are no newer games that scale beyond 4c/4t?

I'm sure that there are those people that say that literally, but I think people are more prone to say "it doesn't make a big difference" and that is true on a case-by-case basis. A year from now, I think it will be true in the majority of cases.

 

[ashetos]

sure but it's relevant enough to mention that the 2500K without OC is the same as the 2500, the 2500K is significantly older and running significantly slower ram, if we had an 8150 it would be clearer.

Agreed that memory speed should be similar for better comparison. Cache size is also an unknown variable.

 

[arandomguy]

The 2500K was a legendary CPU. People will still think it is the best performer today, even 6 years later.

A large part of that was the rather easy 20%+ (really more around 30%) performance uplift from overclocking. It also has a strong reputation for still be very a serviceable a gaming CPU after all these years which these tests also show since for the most part the FPS numbers are certainly within playable. So whether or not it loses out to an almost 4 year newer product stock to stock doesn't really affect whether or not someone should've been happy with getting a 2500k back when you would have been buying one at retail (certainly not in late 2014).

I'm sure that there are those people that say that literally, but I think people are more prone to say "it doesn't make a big difference" and that is true on a case-by-case basis. A year from now, I think it will be true in the majority of cases.

It really should be looked at case by case which is why I think these overall results can be misleading. You can pick a list of newer games that scale to varying degrees, and I think at least in this case they certainly picked some of the titles have been known to scale much better.

But I don't like this attitude either way of one side is irrelevant. So that was also a bit of cheeky comment directed at both sides of the argument.

 

[ashetos]

About RAM speeds though, we should be reminded that AMD had pretty poor performance with faster RAM cause intel had better memory controllers:

https://www.overclock3d.net/reviews/cpu_mainboard/amd_vishera_fx8350_piledriver_review/4

I also remember that piledriver had pretty bad cache performance. I think the number of cores is the most important factor for the original post results.