CPU impact on gaming (in Off-topic)

http://techreport.com/articles.x/23246/3

summary of good parts for a gaming pc? :o

Looks more like Intel propaganda to me.

I noticed a lack of buttons for AMD high and extreme or intel budget.

because they are no good for gaming :P

I still wouldn't buy an intel machine ever in my life. Those results are very skewed. Sorry milliseconds worth of difference in frame latency that the eyes can barely detect when the game is running smooth as butter to begin with isn't going to make me hop ship and decided to spend an extra chunk of change to go intel then be locked into whatever socket type I chose. Sorry there are no easy upgrade paths for intel due to the way they set up their sockets and I'm not a fan of that at all.

Intel fan boys go ahead and cream yourself at these results, and I'll just laugh as you get excited over how big you think you're e-peen is, while still running my games maxed out.

dude, I Have a custom build Intel PC, built by me, and I did tons and tons of real world bench marks reviews, and as much research as I could.

I have an X9650 (core 2 duo quad 3.0ghz) I can overclock it to 3.8-4.0ghz stably. I have 4gb 1066 ram, forget what I have my lacentecy let to.

Asus P5K overclockers board. I had a 4870HD before my ex trashed my house.. and it was overclocked to the limit too. I could play any game on max settings and it would be smooth as hell.

I now have a 6790hd becuase i had to replace it, and was short on cash.

But regardless, I can still play any game on the market with settings maxed.

FYI that x9650 is $1000 chip and well worth the money. huge cache, fast as hell, and unlocked.

its obviously a few years old now.. but still runs better then any p.o.s dell compaq avio or hp they mass produce.

Your chip cost much the same as an entire(screen,keyboard,printer,the works)computer sold by those listed. I'd hope so. =/
This tech report is inevitably propaganda all the same. Don't have to be AMD or Intel fan boy to critically read.

Lol I love my i7 No complaints here, but I am amd at heart.

Xeno:

http://www.tomshardware.com/reviews/guild-wars-2-performance-benchmark,3268-7.html

I skipped to the relevant page of the long article. I'm not an Intel or AMD fanboy, but right now, Intel's the way to go.

Of all CPUs being sold right now on newegg, looking at the prices, you get...

AMD - 29 CPUs (the ones with graphics on chip being slightly weaker)

40 USD for single-core @ 2.8 GHz
55 USD for dual-core @ 2.5
...
160 USD for eight-core @ 3.1
190 USD for eight-core @ 3.6


Intel - 60 CPUs being offered (almost all with Intel graphics on chip)

50 USD for dual-core @ 2.4
53 USD for dual-core @ 2.5
...
120 USD for dual-core @ 3.1 (cheapest i3 SandyBridge)
...
180 USD for quad-core @ 2.9 (cheapest i5 SandyBridge)
...
200 USD for quad-core @ 3.1 (cheapest IvyBridge)
...
340 USD for quad-core @ 3.5
570 USD for six-core @ 3.2 (dual memory channel)
1030 USD for six-core @ 3.3 (quad memory channel)

...

The only shocking lack of buttons would be "intel budget".

P.S. The clear winners of the test in the OP as far as price/performance ratios go seem to be:

a) the Phenom II X4 850, quad-core @ 3.3 GHz, listed at a bit over 100 USD on the conclusions page, which is no longer being actively sold, but the Phenom II X4 965, quad-core @ 3.4 GHz is available at 110 USD nowadays

and

b) the Core i5-3470, quad-core @ 3.2 GHz, listed at 184 USD in the article and at 200 USD on newegg right now

P.P.S. Also, quite a few of those intel CPUs actually have hyperthreading.
That does NOT double their speed, but it boosts it from anywhere between 15% to 30% depending on application (it can also be absolutely no benefit whatsoever for single-core-only apps, or it can be even more than 50% in special circumstances).
For practical intents and purposes, a 4-core HT CPU is equivalent to a 4.6-5.2 core no-HT CPU, and a 6-core HT CPU is equivalent to a a 6.9-7.8 core no-HT CPU, of similar frequency, give or take, depending on application.

I should had expect this kind of answer, the reason i post this review was not for the score since they are in line with all the others review out there, its the way they test GPU/CPU under a game. All review back in the day were about average FPS. In the last couple of year we start to saw review that include min FPS. This review push its futhers. Frame latency is more complicated and viewer would need adjustement since its wont just be a couple of bar in a graph.

The idea goes like this there 1000 ms in a Sec and for a game to run smooth you need min of 60 frame per sec. 1000/60=16.6 ms so any frame that take longer that 16.6 is slower that the Vsync. The bar for playability is around 32 ms, its choppy but playable.

So what you want to know from your systemes or the 1 you will be buying is will its be able to run the game smooth at ALL time, aka under the worse condition. The review make a average of the 1% worse latency time or 99% percentile of the highest frame latency.

Well, two things.

First, most older games were far more uniform in graphics load across most of the gameplay time, so the distinction between min FPS, average FPS and max FPS was minimal, since FPS didn't fluctuate all that much back in the day.
Nowadays, you can easily encounter sections of the same game running at over 120, while barely eking a paltry 15 FPS in other sections, under the same graphics settings. When explosions and such happen, OUCH, watch out, that's a real FPS killer... and you know when a decent FPS is usually the most important ? EXACTLY when stuff is exploding like mad all around you.

Second, this review does not push the "minimum FPS"//"maximum frame latency" thing further, it actually dials it back a little bit.
Instead of accounting for the lowest encountered frame rate, they actually EXCLUDE THE WORST 1% of the frames, getting you a noticeably higher equivalent "usually//almost lowest" FPS number because of that (since the tail-end of frame latency looks pretty bad for most CPUs listed there, both AMD and Intel, not just AMD).

ok so you are telling me there a 100th percentile ?

Luckily I'm not playing GW2 at the moment ;P

I generally expect Intel to bench higher, that's almost a given after reading so many bench comparisons over the years. However I'm extremely opposed to the way Intel sets up their sockets. It forces you to lock yourself into one, which severely limits upgrade paths and if they decide to release a new CPU under a new socket, well then you got to get a new motherboard to go with that socket if you want to upgrade (That's at the bare minimum, there could always be other parts you have to replace as well depending on the MOBO you are upgrading to). This in the long run ends up costing more then I'd like.

The way AMD sets up their socket types, they have forward/backward compatibility in mind, which means they are far more flexible with their upgrade options. Motherboard dying on me? (Happened once, and its slowly happening again.. Never getting an MSI again.) Get a newer one and just drop my current CPU in with plans to upgrade in the future. Now if Intel switched over to this mindset, then I might not have nearly as much of a problem with them as I currently do, but as it stands I still find Intel to be too expensive for my tastes in the long run.

Also the fact that their best (unlocked) processor costs nearly $1000 leaves a sour taste in my mouth (It might not be that price right now, but I know the best unlocked i7 was when it came out). Before you say anything, yes I know you can still OC locked processors, but when AMD offers their best unlocked at nearly 1/4 of that price that's something to take into consideration.

Though to bring me to my final point: Actual Gameplay, If I can run the game at max or even near max settings without hiccups (or so few they are barely noticeable) then I'm happy. If I'm able to accomplish that on a relatively low budget with an open upgrade path for the future, then that's even better.

With all that being said.. I start working for IBM on the 10th.. They use Intel machines >_> Free Intel laptop for me ;P

Rofl! Can't let you use that socket, Dave.

Yeah. It's called "minimum FPS". The thing most other reviews use.

i suggest you google percentile

So what exactly else so you think MINIMUM FPS actually means other than "the absolute lowest FPS you can possibly get" ?

Take for instance a graph similar to those depicted there, with the percentile frame latencies graph going from, say, 10 ms to 20 ms almost linearly for most of the graph (within the 99 percentile), and then near the end, a handful of frames start spiking towards 25 ms (when you reach 100 percentile).

20ms latency translates into 50 FPS, 25 ms latency translates into 40 FPS.

By definition, the MINIMUM FPS on that graph is 40 FPS.
However, the 99 percentile minimum FPS is 50 FPS.
That's a less bad result.

Percentage is not percentile, there is no 100% percentile as its mean infinite.

That's true only when you're talking about something like a theoretical normal distribution, which can still have a tiny percentage of the potential measurements at no matter how many standard deviations away from the average you go, so, yes, in that case, it's "infinity".

Percentile is defined as "the value of a variable below which a certain percent of observations fall".
In case of finite and measured hard data, "100 percentile (value)" would simply be a synonym for "highest value recorded".

So according to you the 100th percentile would be the highest record value and the 99th percentile would be what again everything but the worse 1% ?

More or less, yes, something like that.

When you say "lowest FPS", you say "the absolute worst frame latency" and calculate equivalent FPS from it.
You can call it "100th percentile frame time" if you like, but you are right that it does sounds ridiculous.

When you say "99th percentile frame time", you EXCLUDE the 1% worst frames, and calculate the equivalent FPS from the worst of the remaining ones.
The number you get is always better than the lowest FPS.

99th percentile you take the lowest 99% value and you discard them.

when in percentile you dont use 0 or 100 or you ending with 101 posibility

Dude, how hard is it to realize you're basically saying the same thing (even if getting hung up on terminology) but somehow seeming to imply you are reaching a different conclusion ?

Fact remains, THOSE guys in THAT test sort all frames by time needed to compute them and line them up on a nice graph side by side, each frame time being the vertical value of the point on the line.
Then they pick the frame time that's 99% away from the best (lowest) time and 1% away from the worst (highest) time on the horizontal line.
That gives you a FPS number that's better (HIGHER) than the absolutely lowest FPS experienced, and they use that better number in their assessment.

Take a second to think about this one : explain to me how exactly could you EVER get anything WORSE than the lowest possible FPS ?
Obviously, you can't get anything worse than the lowest FPS you encounter, it's impossible.
The FPS they use for their conclusions is a bit better than the lowest (worst) FPS encountered.
It's as simple as that.

Or what exactly about that are you disagreeing with anyway ?

Where at any point they mention FPS in there graph ?

The point where they say:


...and of course, the table to the right of that text.
FPS = 1000 ms / frame time in milliseconds.

And there's this one, which matches what I pointed out on the other graph:



(emphasis added).

I actually have something useful to contribute to this conversation. Nerds.

Nerds designed the machine you're using now and the games you're playing on it, so I'll take that as a compliment :p

^I agree with that statement.