[Tom's] CPU Bottlenecking with 7970 CF - 3770k vs. 8350

[Phynaz]

You know very well that not everyone paying 3.5K USD for software. Also, there are freeware applications.

And just to clarify this, im not talking about professionals and workstations, im talking about every day desktops, people doing both work and gaming in their machines. Not everyone is budget unlimited

You didn't answer the question.

 

[AtenRa]

You didn't answer the question.

Free 3D Render applications

25 (Free) 3D Modeling Applications You Should Not Miss

16 Free 3D modeling software to create your own 3D models at no cost

Autodesk 123D

Im sure there are more

Non free.

It seams to me that FX8350 is faster and cost less than Core i5 3570K even in free 3D Render applications.

 

[amenx]

You forgot the 7-zip bench.

 

[SlowSpyder]

The FX didn't do all that bad (if you toss out the Skyrim results, anyway). Though there is zero doubt that the Intel chips are the faster parts, overall. Intel's chips are more flexible, that is you get great single and lightly threaded performance as well as solid multi-threaded performance. With AMD you get solid mulithreaded performance, but you would certainly be slower in single and lightly threaded games... nothing really new here.

If I had two 7970's, I would not play at 1080P. But if you have a 120Hz 1080P monitor and want to stick with just a single monitor, then I think Intel gives you the better experience. But based on what I'm seeing in the charts, if I'm gaming on three monitors I wouldn't have any problem using an AMD FX part to do so. Sure, the Intel chips are often faster, but the AMD parts are often very close and above or very near 60FPS.

Really, it is mostly the same story here. A few posters on both sides of the fence making a mountain out of a molehill... some of the PC gaming Gods here wouldn't touch an AMD part because the 58FPS AMD's chip can push would make their eyes bleed because it is so pathetic. Then on the flip side, others here will somehow try and frame AMD as a better part when that's clearly not the case.

To me, Intel is the clear choice for a higher end gaming system using two 7970's and Eyefinity. If you're dropping $1k+ in monitors and video cards, I would suggest you get the pricier Intel CPU too (vs. a cheaper AMD FX part). With that being said, AMD is nipping at Intel's heels here and there with a 32nm chip. I think that is fairly impressive, especially when you factor in R&D expenditure.

 

[AtenRa]

You forgot the 7-zip bench.

First you have to read post 69, then you will understand how stupid you look trying to be a smart ass.

 

[Phynaz]

First you have to read post 69, then you will understand how stupid you look trying to be a smart ass.

And here I was going to say I learned something today.

Now I'll just say you derailed the thread because this was about gaming, not rendering. Show AMD in a bad light though and you always seem to want to change the measurement.

 

[LOL_Wut_Axel]

I recently purchased a Phenom II X4 965 because it seemed like a steal for $85. Would it bottleneck a GTX 460@850MHz at 720p or 1080p resolution?

 

[SlowSpyder]

I recently purchased a Phenom II X4 965 because it seemed like a steal for $85. Would it bottleneck a GTX 460@850MHz at 720p or 1080p resolution?

It is probably a good match for that GPU. Do you overclock? If so, don't simply take the easy route and adjust the multiplier. Adding NB/L3 speed as the cores get faster can give you a tangable bump in performance.

 

[AtenRa]

And here I was going to say I learned something today.

Now I'll just say you derailed the thread because this was about gaming, not rendering. Show AMD in a bad light though and you always seem to want to change the measurement.

It was you that asked about the Rendering software, not me

 

[spartan805]

Wat do you guys make of TEK SYNDICATEs 8350 video?

http://www.youtube.com/embed/eu8Sekdb-IE

 

[SlowSpyder]

Wat do you guys make of TEK SYNDICATEs 8350 video?

http://www.youtube.com/embed/eu8Sekdb-IE

It'll be dismissed.

I remember a loooong while back, someone bringing up Overclocker Club's review that shows a stock clocked Phenom I sometimes leading the pack when resoution increased, depending on the game of course. But, by then it was too late for AMD, PhI already had such a negative perception compared to C2D. My guess is any current review that shows AMD's FX in a good light will not be taken seriously either.

I think sometimes there is more to the whole story than just some low-ish res benchmarks when it comes to gaming CPU's. The whole subsystem is stressed when benched at resolutions and setting we actually play at. So it would be possible for what appears to be a slower CPU with to hold its ground as resolution and settings increase, if it truly has a better chipset, I/O, etc., in my mind. But once a part launches to negative press, often times minds are already made up.

And I am not saying the FX is better than Intel's line up. But what I do believe is that often times real world experiences may not exactly reflect tech website benches.

 

[chimaxi83]

I'm confused, why are 3D rendering benchmarks in this gaming benchmarks thread?

 

[AtenRa]

I have just uploaded a video of FX8350 running F1 2012 build in benchmark with same settings as in Toms CF review.

AMD FX8350 + ASUS M5A97 R2.0

AMD FX8350 - F1 2012 1080p Benchmark - CPU utilization and power usage

System setup.

AMD FX8350
ASUS M5A97 R2.0
Kingston 2x 4GB 1866MHz 10-11-10-24
ASUS HD7950 CU II 1000MHz core 1500MHz Memory
Win 8 Pro 64bit
AMD Cat 12.10

F1 2012 build in benchmark Ultra.

Average 82fps, Minimum 63fps

No way HD7970 CF will output less fps than my single HD7950. It seams that CF is not working in F1 2012. Anyone with CF to test it ??

 

[SPBHM]

going by the GPU usage on your test, I think a CF wouldn't help much!?

 

[Termie]

I have just uploaded a video of FX8350 running F1 2012 build in benchmark with same settings as in Toms CF review.

AMD FX8350 + ASUS M5A97 R2.0

AMD FX8350 - F1 2012 1080p Benchmark - CPU utilization and power usage

...

F1 2012 build in benchmark Ultra.

Average 82fps, Minimum 63fps

No way HD7970 CF will output less fps than my single HD7950. It seams that CF is not working in F1 2012. Anyone with CF to test it ??

The irony here, of course, is that both the 8350 and the 3770k systems clearly had issues with the 7970s in F1 2012, just to different extremes. You are correct that CF isn't "working". It's working poorly on the 3770k as well. The results still say a lot.

going by the GPU usage on your test, I think a CF wouldn't help much!?

The problem here is a common one - you can't judge CPU bottlenecking by the CPU usage metric in Windows. Games aren't perfectly threaded. So you have to look at the GPU usage metric, which in the case of the youtube benchmark above is 80%.

 

[BrightCandle]

The problem with F1 was there in F1 2011 as well. It does not work with AFR well at all and scales negatively. It used to scale so badly that it was unplayable )sub 20fps.

One of the few times I got a refund for a game on steam as it explicitly said it supported the 5970 whereas clearly it does not. Its confirmed by codemasters and they don't intend to fix it.

 

[SPBHM]

The problem with F1 was there in F1 2011 as well. It does not work with AFR well at all and scales negatively.

the thing is, if the he is already limited by the CPU with that card (GPU usage around 80%), would adding more GPU power really help?

it may well be the case, both the i7 and the FX are the "bottleneck" for the cards (and CF scaling will be poor), but since the i7 is faster, the limit is higher.

 

[boxleitnerb]

http://www.youtube.com/watch?feature=player_embedded&v=4et7kDGSRfc

Again with this TEK guy? His results make no sense whatsoever, just take a detailed look at the numbers and think about if that is possible or not. I wish people would stop posting this nonsense.

 

[Abwx]

So whatever show an FX being good is automaticaly non sensical..?..

He provides numbers wich you only contradict with words
based on an subjective perception.

Where are your own numbers to compare with .?..

 

[boxleitnerb]

There are differences where there shouldn't be any (due to GPU bottlenecks, large fps drop from 1080p to 1440p), OC at 1440p often yields more than at 1080p which also is not possible or it yields an fps boost on one CPU and barely anything on the other despite a suggested CPU bottleneck. In general, his results are in complete contrast to all of the properly done benchmarks on the web.
He also suggests at the very beginning of the video that 30fps are fine, which disqualifies him right there. He goes on about saying that GPU bottlenecks behave differently with different systems which is wrong, as can be seen on countless reviews over the last 10-20 years.

This applies to almost all of his "measurements" (I have looked at each and every one of them). If you knew me, you would also know that I have a detailed understanding about how to conduct benchmarks and that if I say something I have a good reason for it.

You don't need to test something yourself to point out obvious errors. I would suggest you take a look at those results yourself and critically think about if they are plausible, because judging by your words, you clearly haven't.

And aside from all that, how would we be able to compare numbers if he doesn't exactly say what and how he benchmarked? Documentation is key. Here is what a good and complete documentation should look like:
http://translate.google.de/translat...Tests/Benchmark-Test-CPU-Grafikkarten-744422/

 

[Abwx]

I didnt pay attention to the scores , i just checked if it was
not rigged.

That said , AMD themselves say that the FX is largely on par
with the 3570K for games , so why dont reviewers get the same
conclusions than AMD.?

Are people at AMD unable to competently bench their own products.?.

Are the following chart irrelevants.?..

http://www.lesnumeriques.com/cpu-processeur/amd-fx-8350-p14680/test.html

 

[boxleitnerb]

Yes, they are irrelevant.
Why? Because obviously AMD, Intel, Nvidia etc. benchmark in such a way as to put their products in the best light. Usually, those benchmarks are massively cherrypicked, thus not representing reality. I would never ever be so naive and trust IHV benchmarks. If you want to get benchmarks, go to independent sites without an agenda.

 

[Abwx]

That s 12 popular games , yet it s still cherrypicking...

Btw , there s not a single site that did use 12 games for their benchs ,
so who is really cherry picking , actually..?....

 

[boxleitnerb]

You don't understand.
It's not only about the games themselves. Depending on the hardware and software configuration and the actual benchmark scene, results can differ tremendously. For example in Starcraft 2 with an AMD GPU with MSAA and few units on the screen you are GPU bottlenecked. But with lots of units fighting and/or an Nvidia GPU you become completely CPU bottlenecked. Or Skyrim. Inside you are GPU bottlenecked, outside you are often CPU bottlenecked, especially in cities. Shogun 2 GPU benchmark is GPU bottlenecked. Actual battles can be CPU bottlenecked. Etc.

I never look at just one review anyway, but like 5. That way I get numbers for 15-20 games.

 

[Abwx]

Agree with your sayings but then what you say means that the usual
tests protocols used by sites and wich consist mainly in integrated benchmarks are not relevant as well.

No surprise that Hardware.fr do not use them and prefer to check
the things using very demanding scenes.

I find it odd that Atenra claim better results than THG despite
inferior hardware , just like the bench posted earlier that show the same
score in F1 2012 for a 8150 and a 8350...

All this point to non normalized methods , and in this respect
if you consider the ones showing the FX under a good light as
irrelevant then the ones that are not favourable should be
taken with as much grains of salt if not more...