[BUG] Only 40 FPS with an overclocked GTX970... REALLY????

Hey, I've a question about planetside 2 performance : is it affected by RAM ?
I've currently 8GB RAM and PS2 use about 3,5GB.

As far as I know, I just need enough RAM to run PS2, so having 8GB is around 5GB overkill. But a random guy ingame wololoed me to trust that 16GB RAM will improve my performances.
What do you think about it ?

Additionnal info witch are useless for my question :
CPU : AMD fx 6300 3.5GHz
GPU : radeon r9 270x

The discussion is that passmark is trash. Showing you exactly what makes it trash is relevant... it IS the discussion. So, naturally you don't want to discuss it.

Specifically: while memory speed does have a minor impact on single-threaded performance; memory capacity does not. The same guy posted two different scores who's only change was going from 8gb to 16gb. That's a fundamental problem with passmark. If you were to use a more reliable single-threaded benchmark, like Cinebench, you will find that increasing capacity does not alter the benchmark's single-threaded performance score. Pointing that out should have cleared things up for you. Why didn't it?

Only in passmark. Everywhere else, reviewed by everyone else, the natural order is restored. Skylake ipc is faster than all prior gens.

Apparently you don't know what "faster ipc" means? The 6600k isn't midrange. Perhaps you should browse the Skylake catalog?

This is not how boost works. Perhaps you should look that up too? Boost does not kick in with respect to what you're doing; boost kicks in depending on how many cores are working when you're doing what you're doing.

The results you seek are, literally, everywhere. Go read a review that isn't passmark. Literally thousands? Is that like the 3 xeons? Hyperbole much?

Xeons are amazing cpu's, but they weren't designed to run games/single-threaded apps. Xeon's smoke in multi-threaded environments, but that isn't what were talking about. Besides, I started out showing you what a load the passmark score was with respect to Xeon's. Try as you might, you have not addressed the glaring flaws with the Xeon submission: 3 samples, all running 1600MHz ddr3, radically different scores. Oh, and there's still that capacity thing...

Ram speed can increase performance in cpu limited games. So while more ram may or may not help you, faster ram will certainly make the game run better.

With SLI 2x GTX970 & GTX470 I run at 80% on all 6 cores in PS2... this would mean a 4 core machine would never in a million years handle PS2. I want to point out that PS2 now uses all 6 cores.... not 5 cores. I wanted to check this stupid conversatation you talking about. Yup you read it here PS2 now uses all 6 cores.... 80 % of all 6 cores.

Oh, that's what you meant? Lol, I didn't even consider that, since those "Two different scores" you refer to differ by a whopping THREE PERCENT. Wow. Totally not just normal margin of error, that.. /s
Regardless, though, you really don't get it, do you?
Even if there were a significant difference (which there isn't), when there are thousands upon thousands of samples taken, you will get results from systems with all sorts of RAM capacities. It all 'evens out in the wash', so to speak, as do other variables that can affect score.
This is how statistical samples work.

Except that ipc is not the single and only factor determining CPU performance.

Have you heard of the i7? That's top-range. i5 is mid-range.
ipc? Instructions Per Clock?
Even if that were the only determining factor in CPU performance, almost every chip you listed runs at a higher clock speed than the 6600k.

This is exactly how boost works. I have an i7-4770K. Under 100% load, all cores clock up to the boost frequency until thermals require otherwise. Can it clock up individual cores? Yes. Does it have to? Noo.

Got it yet? Or just purposely being obtuse now?

Xeons share the same fundamental core architecture as all the rest of the lineup (i5, i7, etc.). There's no architectural reason their individual cores would be any slower. (unless you can provide one)

Anyway, you seem to be under the delusion that the i5-6700k, since it's newer, or since Skylake "has higher ipc!!", must outperform every CPU that came before it.
Which is a silly notion, frankly. I mean, are you one of those people who thinks the GTX 960 is better than the GTX 780 because it's a newer generation with a superior architecture?
Here are some real-world 'heavy workload' tests of the 6600k: As you can see, the 6600k is, in every test, outperformed by a number of Haswell and even Ivy Bridge i7s, even as well as some Haswell i5s.
http://www.tomshardware.com/reviews/skylake-intel-core-i7-6700k-core-i5-6600k,4252-7.html

"But Skylake has more ipc!!!"

I guess you forget this part?

I honestly don't know one manages to look at that link and not immediately realize something is very wrong? You can complain "I said consumer" all day... except you didn't. If Xeons were as good at gaming as you and your link suggest, they'd be very popular.

So the k series are mid-range huh? lol, ok... I guess you also don't realize people buy the i5's because they're just as fast as the i7's in games because games, except for a handful, don't benefit from hyper-threading. What? Passmark not showing you this? Here:

http://www.ocaholic.ch/modules/smartsection/item.php?itemid=3948&page=3

http://www.anandtech.com/show/9320/intel-broadwell-review-i7-5775c-i5-5675c/9

You can find the same results literally anywhere you look... again, aside from passmark.

I think you're now being terribly disingenuous. "thousands of samples" is actually 3. "3%, within the margin of error" is actually closer to 10%.... and, that ridiculous image of Abode media encoder, lol.. What on earth are you posting that crap for? It's a multi-threaded benchmark.

You have said, literally, that the i3 4170 is a better gaming cpu than the 6600k... among other asinine things...

Listen, you're actively dodging my questions and contorting my statements. I grow tired of this, but:
- "avoidance"
I don't even... wat?
- Xeons will never be popular for gaming because, even though you WILL get the same performance as with an i5 or i7, you'll pay 5x as much for the extra 10 cores you won't be using.
- The consumer CPUs all have thousands of samples. The Xeons do not. I don't know why this is hard for you to grasp.
- I find no evidence that Passmark artificially inflates single-threaded scores on hyper-threaded CPUs, as you seem to believe it does.
- The "different scores for 8 and 16gb RAM" scores you linked are exactly 3% different. Not 10%.
- If you bothered to actually click the link, you'd see that the i5-6600k was beaten by older processors in all of the algorithmic tests, both single and multithreaded. Not that it matters when you're comparing processors with the same number of cores and threads.
- Yes, there are even Pentiums which will match the 6600k in performance for most games, simply because most games rely heavily on a single thread, and rarely need three or more extra cores to handle their parellel operations. And the Pentiums are best at this since they have the same core architecture as an i5, but with half the thermal constraint, allowing them to clock very high. Planetside 2 is an exception in that it needs a quad-core to run well. But I have an i7-4770k (as i mentioned previously) and a Pentium G3258. It is the same generation as the i7, but OUTPERFORMS it playing the Witcher 3 at 4k with an r9 Fury Nano. Now, someone like you would call this conclusion "asinine". But I'd bet that it would best the 6600k as well, since it can run at a steady 4.8ghz without overheating, compared to my i7-4770k's 4.2ghz max and the 6600k's measly 3.9ghz boost clock.

Which part of this statement of yours aren't you getting?

- Xeons will never be popular for gaming because, even though you WILL get the same performance as with an i5 or i7, you'll pay 5x as much for the extra 10 cores you won't be using.

No, you don't get the same performance. Show me otherwise. keeping in mind, that it has to be real, exactly like I've linked for you. Yes, if Xeon's did provide the same performance people would be buying them. I would if they were "the best gaming cpus". You think someone who drops several grand on sli'd titans give a hoot about a $500.00 cpu? I used to run Opterons back in the Athlon64 days expressly because they could run hotter. That they were more expensive did not matter one bit.

- The consumer CPUs all have thousands of samples. The Xeons do not. I don't know why this is hard for you to grasp.

You don't know that. What part about that aren't you getting? You have no clue how many submissions were made, none.

- I find no evidence that Passmark artificially inflates single-threaded scores on hyper-threaded CPUs, as you seem to believe it does.

Nobody has any idea what passmark does; this is why no legitimate reviewer uses it. I never said passmark artificially inlfates single-threaded scores, what I doubted was that it's a single-threaded test at all. A cpu's single-threaded performance does not increase with capacity. It barely increases when increasing ram speed. A cpu's single-threaded performance increases with clock speed.

- The "different scores for 8 and 16gb RAM" scores you linked are exactly 3% different. Not 10%.

Yes, now how about answering the question? The discrepancy isn't between the guy who submitted two tests with 8 and 16gb, the discrepancy is between him and the other two guys who submitted Xeon scores far below. The difference between 9514 and 10,718 is not 3%. Explain this. Cpu and ram speed are identical for all samples.

- If you bothered to actually click the link, you'd see that the i5-6600k was beaten by older processors in all of the algorithmic tests, both single and multithreaded. Not that it matters when you're comparing processors with the same number of cores and threads.

Scroll up, I provided you with real benchmarks from ocaholics and anands showing that game performance... which is what you were talking about, remember... is identical.

- Yes, there are even Pentiums which will match the 6600k in performance for most games, simply because most games rely heavily on a single thread, and rarely need three or more extra cores to handle their parellel operations. And the Pentiums are best at this since they have the same core architecture as an i5, but with half the thermal constraint, allowing them to clock very high.

/sigh... again... that's the problem with passmark. It does not tell you if the submission is overclocked or not. Case in point: the 1 Xeon in question scoring far above what the other two Xeon's scores, presumably all running at the same clock speed, at least according to passmark.


Planetside 2 is an exception in that it needs a quad-core to run well. But I have an i7-4770k (as i mentioned previously) and a Pentium G3258. It is the same generation as the i7, but OUTPERFORMS it playing the Witcher 3 at 4k with an r9 Fury Nano. Now, someone like you would call this conclusion "asinine". But I'd bet that it would best the 6600k as well, since it can run at a steady 4.8ghz without overheating, compared to my i7-4770k's 4.2ghz max and the 6600k's measly 3.9ghz boost clock.

You can oc any cpu... that isn't the point at all. Again, clock for clock, in gaming, the 6700k and 6600k are identical. You can, if you want, look at real world performance displayed as fps in the links I gave, or you can stick with meaningless random numbers passmark throws at you.

DDR3 1600 Mhz. Nothing special.

With Firefox open, CPU use bounces between 1 - 4%. It's pretty much idle. I dont have problems in any other game and there is nothing anywhere to imply there is something wrong with the computer or OS.

The vast majority of the benchmarks I've seen show that RAM speed basically makes next to no difference in any practical application whatsoever. It only makes a difference in synthetic benchmakrks. As RAM speed goes up, so does latency so in the end you're just increasing one type of speed to decrease another.

I dont know why everyone is so hung up on single threaded applications. That's a totally irrelevant discussion as single threaded programs literately no longer exist with WIndows 10. Windows 10 will force a program across multiple cores, even if it's not coded to run multiple threads. Windows 10 has the ability to do that. So does 8.1 and 8 if I recall right. That's one of the flagship things they advertise about the OS "can more effectively make use of multiple CPU cores, even if the program is not written for it." In the last few years, I have "never" opened Task Manager to see one CPU core maxed out and the others sitting idle. That's a thing of the past. All CPU core will do work on all programs all the time with modern day OS', even if the program was written for WIndows 95 era.

While it's true that the "vast majority" of benchmarks show that ram speed makes no difference, it's also true that the vast majority of games are gpu limited, not cpu limited. In cpu limited games ram speed does make a difference.



http://www.eurogamer.net/articles/digitalfoundry-2015-intel-core-i3-6100-review
http://www.overclock.net/t/1487162/...affect-fps-during-high-cpu-overhead-scenarios
http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/

PS2 performance suffers because the primary gameplay thread is so large/costly. Dgb can't break it up without breaking the game. This is probably the most cpu bound fps ever made. Windows, no matter what version, isn't going to change that. That Windows 10 is better at spreading load doesn't change the fact that games are still single-threaded applications. This is precisely why a cpu's single-threaded performance is key to gaming performance. If everything you said were true, then Windows 10 plus an AMD 8 core would the thing to buy... it's not.

Well, seeing as I never said "avoidance", and that you edited my quote to say that...
None?

Easy.



As you can see, the Xeons are on-par with even the most powerful i7s from the same generation.
But even the "big spenders" won't be stupid enough to buy one for gaming when they can get an i7 for $1000 less.

ROFLMAO, you serious?

Unfortunately for you, there's no evidence that Passmark is doing anything wrong, or that it "isn't a single-threaded benchmark at all"

Really? A 10% difference in score between two completely different systems? One with faster memory than the other? One with SSDs in RAID and the other with a single physical disk? It's inconceivable to you that these systems could produce slightly different scores?

Those tests compared the i7-6700k and the i5-6600k. What exactly is your point there?

Statistics, man.
Statistics.

Well, most Pentiums and i3s run at higher base clock speeds than i5s even without overclocking.
But have fun trying to run your 6600k at 4.8ghz.

Your susperstition about how Xeons should be worse for games because they aren't "designed for gaming" (whatever the hell that means in regards to designing a RISC x86 CPU) tells me that you don't have an adequate understanding of how processors are built and how that affects performance in different applications. Sorry.

Statistics indeed. The Xeon in question has 4 entries, 2 from the same guy... so 50% are the same cpu. What's the probability that the rest of the entries are also second, or third submissions, etc. etc?

Yes. Look up, work that problem, then conclude that you really don't know how many samples there are; you only know how many submissions have been made. So again, you have no clue, and neither does anyone else.

A cpu single-threaded performance test is just that: a cpu test. It is not a system test. If the score is changing because of the system, then something is wrong with the test. Ram speed does not matter: http://www.hardocp.com/article/2015/08/05/intel_skylake_core_i76700k_ipc_overclocking_review/5 <-- See? From 2133 to 3600MHz the cpu score is the same... because a reliable cpu test tests the cpu... and the suggestion that a cpu test would be impacted by an ssd or raid is... well, lol. Maybe it's his fancy headphones and gaming mouse?

A reliable test is one you can reproduce at home. That's why everyone uses Cinebench and nobody uses passmark. For example, the 6700k is going to get ~ 183 give or take a point or two no matter who tests it, see: http://hothardware.com/reviews/inte...chipset-review-skylake-for-enthusiasts?page=7
There it goes again, 181: http://techreport.com/review/28751/intel-core-i7-6700k-skylake-processor-reviewed/11
What's this, someone else produced the same result? http://www.techspot.com/review/1041-intel-core-i7-6700k-skylake/page5.html Fascinating how these reproducible reliable tests work!
Oh noess..... not again?! http://www.kitguru.net/components/c...600k-skylake-cpu-review/5/?PageSpeed=noscript
Now this is just getting embarrassing... http://www.legitreviews.com/intel-core-i7-6700k-skylake-processor-review_169935/11 183...

Sure, there's no evidence passmark is doing anything wrong... because nobody knows what it's actually doing. On the other hand, as far as cpu single-threaded performance goes, there's ample evidence that it isn't doing anything right... again, which is why nobody uses it.

Passmark seems to make no distinction between overclocked and stock; it reports all submissions as stock. Before you moan, just go back to that Xeon.

Did you forget why we're here already? You said your link was "in order the best gaming cpus". I said nonsense, look how far down the list the 6600k was. You said, blah... it's a midrange cpu! I then linked you to gaming benchmarks showing you that the 6600k performs identically to the 6700k. Actually a smidgen faster in some cases. That was exactly my point... That passmark is garbage; thus, your claims were garbage.

If your G3258 is beating your i7, then you're doing something very wrong... both the i5 and i7 blow it away.
What's that? You meant in Witcher 3? Ok ok, I got you... .... looks like an oc'd G3258 can't keep up with the stock i7''s there either. The i7's minimum was the G3258's average, and the G3258's minimum was terrible.

Yes, there are a few games where Xeon's can look good. However, not that many games benefit from more than 4 cores. You found a few exceptions, bravo, but it's still being beaten by the i7's. Is PS2 such a game? Not by a long shot. How much performance could one expect in PS2 with a Xeon? Well... you can look at passmark I suppose with it's meaningless random numbers that change with ssds ... or... you look at reliable cpu single-threaded benchmarks so you won't be disappointed you listened or orange on the forums:
https://us.rebusfarm.net/en/tempbench?view=benchmark Not so hot.
If that link errors out, just google xeon cinebench... it'll still look just as bad no matter when u get there.

Are you really still on that 4 sample Xeon?
Wow.

Yes, I'm sure there's some hooligan out there who submitted 100 samples on his own. But does it matter in a pool of 10,000? Nope.

Even if every user submitted 10 results (which is most certainly not the case), that's still 1000 different systems sampled.

Nobody here said Passmark is the most perfect, 100% consistent benchmark out there. Rather, it's got a convenient and useful database that marks like Cinebench do not. Making it useful in many situations.

And Cinebench is certainly not a perfect benchmark either. (More on that later.)

Calling Passmark "trash" for a small fluctuation between systems with different hardware and different operating systems is a "trash" assessment, frankly.

Your insistence that the 6600k is the absolute best gaming CPU out there is the real garbage claim here, sorry. :O

Looked to me like the G3258 was getting two more frames per second in TW3 than the i5 and i7 at the beginning there.
Slowdown later on was likely due to the 3258 being dual-core rather than quad-core.
But the faster performance in the beginning just confirms that the G3258 has better single-threaded performance than the i5 and i7.

Oh, do you mean Cinebench Multi-Core, where the top-scoring CPU is a Xeon?
http://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r15_multi_core-8
Or did you mean Cinebench Single-Core, where the top-scoring CPU is a Xeon?
http://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r15_single_core-7

Your beloved Cinebench seems to be directly contradicting your assertions. Ouch.

Finally, if Cinebench is such a great single-threaded benchmark, the 6600k and 6700k should score the same in the single-threaded test, right? Because, as you've said yourself (and which I agree with), HT does not affect single-thread performance.
Spoilers: they don't. The 6700k scores 180, while the 6600k scores 166. How embarrassing! That's a remarkably large difference where there should be none. How do you explain that?

At the end of the day, no benchmark is perfect.
But calling one "trash" for its inconsistencies and then proposing an alternative that has just as prevalent inconsistencies makes very little sense.

Statistics man, statistics.

1, 100, 1,000, or 10,000, it makes no difference. I said you have no idea how many, you can't admit that's true. You also seem to ignore that passmark doesn't make distinctions between clock speed. That "minor issue" has a major impact on results. For some reason, this isn't a problem for you.

Anything that doesn't account for clock speed isn't useful in any situation.

Your definition of small has changed. It used to be 3%, now it's 10...

If you have to lie and put words in people's mouths, why bother? 6600k best gaming cpu? No where did I say that.

lol, yeah, I'm sure that's what you tell everyone, "hey, this is an amazing cpu at the beginning of the game... just restart it and you won't notice the slowdowns."

Oh, well yeah, for 2 grand.

There you go again being dishonest. What I actually said was that it was reliable and widely used.

No, when something is trash it's trash. Passmark is trash, which is why reviewers don't cite it.

Clock speed.

But, since you attributed to me statements I didn't make about the 6600k, I can see why you're confused.[/quote]

Lol!
You invoke statistics, and then in the very next paragraph prove that you don't understand statistics. Niiice.
There is no major impact on results.

Like any large sample, differences in clock speed will even out in the wash. Is it good that Passmark allows different clock speeds? No. Does it affect the final scores? Also no. Even though you seem to think it causes "major impact" on the final result.

The original point was simply that the G3258 has better single-thread performance than the i5. It does.

So.. exactly what I've been saying this whole time?

I'm being dishonest here?

Here's what you said:

Cinebench directly disproves you here, as well as your nonsensical theory that Xeons are somehow inherently "bad at gamez".
And I am being "dishonest" for pointing that out? Seriously?

I'm noticing that the only recurring argument you're making is that "Passmark is trash because it's trash, and nobody uses it!"
That circular logic does not hold water, sorry.
Is Passmark the best tool for one-off tests? No. That's why many (not all) review sites don't use it.
Is it wildly inaccurate in its results as you claim? No, thanks to the large sample database.

6% difference in clock speed = 11 % difference in score?
Nope, try again.


I appreciate you calling me 'dishonest' for disproving your points.

Really, I do.

Not really sure what's left to say. The diff between the 6600 and 6700k in games is non-existent, for the most part, and I showed you that in links. The diff between the 6600 and 6700 is clock speed and hyper-threading; yet, when reminded of that, you go into denial mode. It's that same strange denial mode you went into when I linked you the videos of your cpu getting spanked: your cpu in a lot of the videos was getting half the fps and you managed to look at that and say "it was ok at the beginning", lol.

Did you really already forget that the 10% difference in score from the same cpu on passmark's test was something you brushed aside as "within the margin of error?" Yet somehow, clock speed now... oh never mind...

The list you said was accurate isn't. Sorry.

I won't go into what's wrong with your most recent post, because everything I'd need to say is already in previous posts I've made.

With that I agree.
With your opinions and superstitions, I do not.