It's a sad day for AMD when...

[pelov]

I think that its more that they just don't do their reviews in a professional manner that is repeatable by others, or leave out important caveats about their testing setup.

Tom's data might be interesting but isn't as informative, perhaps, as testing done elsewhere is.

Absolutely. They're not biased they just tend to be a bit clueless when it comes to gaming benchmarks and have turned their site into an ad whore where "news articles" look like documents straight from the mouths of PR folk at hardware companies =P

They would recommend AMD CPUs all the time for the low end of the spectrum and Intel at the high end, what's changed now is that at stock settings no AMD CPU is really competitive in that respect anymore. You've got to invest in both cooling and OC it heavily to see the performance you should be getting at stock. That's not Tom's fault, that's just reality.

Microcenter shouldn't be the be-all end-all when it comes to suggesting builds or proving a point when it comes to value. Microcenter is the exception, thus bringing to light any superduperawesomeamazing deals from MC isn't a valid point. Most people don't live anywhere near a Microcenter and for some who do, like me, I still wouldn't go there. If you want to make a point with respect to value then you should be quoting NE or Amazon or MSRP.

 

[Chiropteran]

Another note:

Street price of the FX-4100 is about the same as the i3-2100.
At stock, i3-2100 beats the FX-4100.

Overclockers claim that after overclocking the FX-4100 the FX-4100 will match or exceed the i3-2100.

Lets say we stick with stock cooling on the 65w i3-2100 and 95w FX-4100.

An i3-2100 paired with a sub $50 bare bones H61 board can beat an FX-4100.

If we were to introduce current Microcenter pricing,
$99 i3-2100
+
$55 biostar board
- $40 MC bundle discount on intel boards

$114

Are you talking about this deal?

http://www.microcenter.com/specials/promotions/expired/0414BUNDLEpromo.html

I think it is expired. Is there a new current active deal? I'm asking out of curiosity, because as far as I know there isn't, but I could be wrong- Microcenter's website is a maze and conflicting information is all over the place.

 

[bononos]

...... Not even 1%. We are talking about gaming here when most of the time we are GPU limited.
And both memory’s cost the same.

You can OC the FX6100 and the 8120 much like I have done with the FX4100 in my review, all the FX series CPUs are multiplier unlocked.

Actually, the FX6100 when OCed is currently($139,99) the best performance/price CPU both for gaming and other tasks.

Ok.

The 6100 and 8120 at stock are clocked lower than the 4100 thus making it harder for them to equal or surpass the 4100's performance by overclocking. So it places them at a disadvantage of being more expensive but somewhat slower than the 4100 at gaming. There are enough benchmarks around to confirm what I said.

I'll just disagree about the overclocking gambit. Its not officially supported, components fail earlier, introduces odd hard to diagnose problems and may require more expensive aftermarket cooling which might negate the initial cost savings. With the FX 4100 the tdp is already 1/3 more than the i3 and overclocking will drive up the heat even furthur. The low power i3 Sandy's otoh are great for marginal budget class mbs with few phases and no vrm cooling.

 

[Chiropteran]

Ok.

The 6100 and 8120 at stock are clocked lower than the 4100 thus making it harder for them to equal or surpass the 4100's performance by overclocking.

It's not exactly that simple.

The 4100 is 3.6ghz, 3.8ghz with 2 core turbo
The 6100 is 3.3ghz, 3.9ghz with 3 core turbo
The 8120 is 3.1ghz, 4.0ghz with 4 core turbo

So an 8120 limited to 4 cores is going to run at 4.0ghz stock, which is in every way superior to a 4100. A 6100 limited to 3 cores is similarly always going to be faster than a 4100 for any application that uses 3 or fewer cores, since it runs at 3.9ghz turbo as opposed to the 3.8ghz max potential of the 4100.

The differences are not huge, and for most gaming the GPU is the bottleneck anyway so the performance is going to be about the same between all the CPU options, but saying the 4100 is the fastest of the 3 would not be true.

 

[pauldun170]

Are you talking about this deal?

http://www.microcenter.com/specials/promotions/expired/0414BUNDLEpromo.html

I think it is expired. Is there a new current active deal? I'm asking out of curiosity, because as far as I know there isn't, but I could be wrong- Microcenter's website is a maze and conflicting information is all over the place.

Holy crap...good catch.
The deal I referenced expired last year.

 

[pauldun170]

I have an answer for you, but first your numbers are wrong.
There is no sub $50 H61 board at Microcenter now.
There is also no $40 off bundle for an i3. You must be thinking of the $40 off bundle deal for AMD.
There is an i3 2100 + MSI H61 combo but it is $120 total. That is $99 for the i3 and $20 for the board.

Now to the FX-4100 side of things. The FX-4100 is $99 also, and you can get the ASUS
M5A78L-M LX PLUS for $10 in a combo, so the total is $110. You save $10 vs the i3.

The board supports 125W cpus, and even the FX-8150.
It can overclock a 95W FX-4100 easy. So there you go.

edit:

link for the board being $10 with an FX-4100 or FX-6100

http://viewer.zmags.com/publication/bd4b92bc#/bd4b92bc/3

If that doesn't work, just look at page 3 of the new April ad.
top left corner

My mistake, I pulled up the microcenter site and they have old deals still up.

 

[BallaTheFeared]

I have used the FX4100 both at default and OC in my review.

But dont blame me, blame Intel for not letting us OC the i3

Blame you for what?

Pointing out the 4100 is slower even at lower fps?

Now do some cpu limited titles like SC2 and WoW instead of doing a review that reads like an AMD PR campaign of Gaming Evolved quad core+ titles.

 

[pelov]

Or throw that 6950 into the rig with the lowest resolution you've been charting instead of the GTS450 =P

Tests showing CPU limitation aren't CPU gaming benchmarks but rather a test whether certain CPUs being tested are good enough.

 

[AtenRa]

Now do some cpu limited titles like SC2 and WoW instead of doing a review that reads like an AMD PR campaign of Gaming Evolved quad core+ titles.

Or throw that 6950 into the rig with the lowest resolution you've been charting instead of the GTS450 =P

Tests showing CPU limitation aren't CPU gaming benchmarks but rather a test whether certain CPUs being tested are good enough.

First thing when you read a review is to read the TITLE, let me quote it for you once again in order to realize what you actually reading.

AMD FX4100: DX-11 budget gaming evaluation (a gamers perspective)

Now let me translate it to you,
DX-11 gaming and Gamers perspective, meaning we test at highest resolutions possible with highest settings playable in each game. Why ??

Because THAT IS THE WAY WE GAME and not at 800x600 with lowest image quality.

 

[pelov]

Because you're not actually testing the CPUs capacity to drive games but rather showing a strictly limited scenario that although you claim to be representative of actual real life scenario would fall quickly when stretched across various resolutions/GPUs.

Is that the right answer?


Need I remind you, there's a reason people test low res and extremely high res multi-GPU when determining a CPU's ability to game. If we followed your logic a core i5 would be equal to a fx4100, an athlon x4 and a 3930K because in your set of tests they'd probably score the same.

That my friend is why nobody would take your reviews seriously even if you did have a review site and that's why Tom's has been ignored and steadily slipping in its credibility. If you stick to GPU limited scenarios than you're not testing the CPU but rather the same GPU twice.

 

[peonyu]

I agree with Toms there, the I3 might be weaker in some areas vs Bulldozer but on the average it is more solid. Single threaded performance it is much better, and single thread is still extremely important. On a budget decision id say I3 is all around a safer bet.

 

[BallaTheFeared]

First thing when you read a review is to read the TITLE, let me quote it for you once again in order to realize what you actually reading.

Your title should have read: AMD Gaming: A marketing spin on gpu bound testing that tells you nothing.

Let's do a quick recap...

AvP? GPU bound, what's next... Metro 2033? I hear this title was all the rave back in 2010, no not really, but I guess from a gamers perspective you should pick obsolete titles that nobody actually cares about.

BF3 - Hey great job showcasing a GPU bottleneck, pitty that overclock didn't do anything for you. I guess we can presume that bulldozer doesn't scale with clock speeds

What happened in Civ 5? I dunno that's a mystery..

Deus - Hey look, our first Gaming Evolved title. I don't sense any bias in these "picks'. Once again virtually no OC scaling, why not rename your test "gpu bound - a foray into pointless comparisons"?

Dirt 3 - Hey look I'm noticing a pattern here!

F1 2011 - LoL you're not even trying to hide it are you?

Shogun2- And thus completes our tour of Gaming Evolved titles and this hand picked review from AMD.



In conclusion we have 4/7 AMD's own titles, followed by two GPU bound titles, and one moderately cpu intensive program where the 4100 get's throttled.

Here's what I learned from your review, creating a gpu bottleneck works out in all but one title. Keeping FPS below 60 is a must not to expose the 4100, unless we're looking at quad optimized titles, often from the same Gaming Evolved Dev, then we can break 60 without fear. The 4100 has no OC scaling, probably shouldn't even overclock it.

Did I get out of your review what you had intended?

 

[Lepton87]

It's not exactly that simple.

The 4100 is 3.6ghz, 3.8ghz with 2 core turbo
The 6100 is 3.3ghz, 3.9ghz with 3 core turbo
The 8120 is 3.1ghz, 4.0ghz with 4 core turbo

So an 8120 limited to 4 cores is going to run at 4.0ghz stock, which is in every way superior to a 4100.

I don't agree, 4GHz is the max frequency, average frequency is much less even in single threaded benchmarks, so it's hard to say which CPU has higher average frequency in single threaded or dual threaded benchmarks.

 

[pelov]

I don't agree, 4GHz is the max frequency, average frequency is much less even in single threaded benchmarks, so it's hard to say which CPU has higher average frequency in single threaded or dual threaded benchmarks.

That gets more complicated when you take into account the way the turbo works depends on whether the module is tasked with a single thread or two threads and how many modules are filled/not filled.

There's max frequency turbo, which is the advertised turbo, and the regular turbo which is the non-advertised turbo.

And then you have to throw the scheduler in there as well... It gets really messy =P

The thing that we should remember, though, is that AMD's Turbo in BD is actually very very good (better than Intel) and its one drawback is that it just doesn't clock high enough in either of the two modes.

Intel's chip is the one throttling here, and apparently needlessly at that. Although Anand wasn't completely sold on the Asynch clocking feature I think it's a great idea if the scheduler is made aware of how the chip is designed. It could potentially save a significant amount of power for a chip this size -- and if you look at the idle power consumption figures it's quite clear that it has too. All's not horrible with Bulldozer, there are definitely some great ideas AMD implemented. It just sucks that the performance wasn't there.

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/4

 

[SunnyD]

THG has ALWAYS accepted sponsored payments from vendors/manufacturers. This shouldn't surprise you. They've been called out before for being paid reviewers for both NVidia and Intel in the past.

 

[AtenRa]

Because you're not actually testing the CPUs capacity to drive games but rather showing a strictly limited scenario that although you claim to be representative of actual real life scenario would fall quickly when stretched across various resolutions/GPUs.

Is that the right answer?


Need I remind you, there's a reason people test low res and extremely high res multi-GPU when determining a CPU's ability to game. If we followed your logic a core i5 would be equal to a fx4100, an athlon x4 and a 3930K because in your set of tests they'd probably score the same.

That my friend is why nobody would take your reviews seriously even if you did have a review site and that's why Tom's has been ignored and steadily slipping in its credibility. If you stick to GPU limited scenarios than you're not testing the CPU but rather the same GPU twice.

Not everything is black and white in real life. Not everyone has $2-3K to spend for a gaming PC system. In real life most of the CPUs sold are low budget models like the Core i3 2100 and FX4100. In real life most of the Graphics cards sold are at the sub $100 segment. People still play games with this low budget hardware.

What my review trying to communicate is that, when you are on a low budget for a gaming system spend more for the GPU. Simple because you will be GPU limited when you actually play a DX-11 game with a low budget GPU like the GTX450.

I dont test to see if the CPU is faster or slower at lower resolutions when the GPU is not a limiting factor. Im evaluating the CPU in a real life gaming environment when the low budget GTX450 will definitely be bottlenecking the system and not the CPU. This is the environment that lots of lower budget gamers will come across when they will play those DX-11 games.

So again, you fail to understand or you dont want to understand what is the purpose of that review. Learn to read and understand what each review is trying to communicate first.

It is easy to criticize someones work when you are out of the loop. I would like to see your evaluation some time too

 

[pelov]

While what you're saying is true, the person should always spend more on the GPU so long as the CPU is good enough, it doesn't represent itself as a CPU benchmark. Like I said before and all the better review sites seem to agree: You don't test a CPU's gaming ability in GPU bottlenecked scenarios. If the scores are the same or there's only a slight difference between them then chances are the test was done poorly and you can't really reach a conclusion. Stating they're all the same in X scenario only provides info for X scenario while all the others are ignored. That's not how you bench a CPU...

I understand the purpose of the test and I'm telling you it didn't state anything worthwhile, but I could have told you before you started taking numbers. Tests at GPU bottlenecked resolutions and settings don't show differences in CPU performance. What you should have done (and still should, imo) is use that 6950 and the lowest res possible (800x600 or 1376x768) with the same CPUs and show just why that's the case by working your way up in resolution from lower res tests where there is a difference. Intel chips are no better than AMD chips at GPU bottlenecked scenarios, therefore the reader should just buy the cheaper chip if they're going to play at GPU bottlenecked resolutions.

You're only telling half the story and provide no explanation for why that's the case. You could easily have done that if you showed that there is a difference but how that difference quickly disappears as the GPU runs out of room, regardless of how much you spent on the CPU.

 

[AtenRa]

Did I get out of your review what you had intended?

Completely,

Now take off those blue glasses and see the review as it should be, from a gamers point of view.

I'm just wondering,
Why do you have GTX470 SLI in your system ??? Is it because a single GTX470 would be a bottleneck when you were actually gaming ??? You spend more for the two GPUs than your CPU for gaming. Why ??? did you trying to eliminate the GPU bottleneck ???

It is funny when gamers know that the bottleneck is the GPU most of the time and not the CPU in todays DX-11 games. But when you show them the graphs to verify this they criticize you of being bias.

 

[BallaTheFeared]

Because I had two at 1080p for a long time and my 4.5GHz 1090T was bottlenecking the one card already

I had to go SB because after trying a 245, 555, 555, 965, 1090T I was left feeling unimpressed by the best AMD had to offer and had a strong desire to go SLI which would have been nothing short of awful with AMD because of their poorly performing processors.

I could probably get by with an AMD cpu is most titles with my current setup, 2 470's at 5800x1080 isn't winning any FPS awards, but it would still choke in the same cpu intensive games so what's the point of buying AMD if it's producing lackluster results no matter what?

Sure you can find titles where Intel vs AMD with a single 6970 doesn't matter, and I can find titles with a single 460 where the AMD cpu gets trounced. I'd rather have cpu performance at all times, in every title, rather than select gaming evolved titles that even older C2D and probably that gawd awful Phenom I would do fine in as well.

 

[AtenRa]

You don't test a CPU's gaming ability in GPU bottlenecked scenarios.

But yet those are the actual real life gaming scenarios. People say CPU A is faster than CPU B in gaming when most of the time DX-11 games are GPU bound.

If the scores are the same or there's only a slight difference between them then chances are the test was done poorly and you can't really reach a conclusion.

So, when two CPUs score the same the test was done poorly and we cant reach in a conclusion that both CPUs are equal ??? We can have two or three or more CPUs or GPUs have the same performance, it dosnt mean the test was done poorly.

When both CPUs have the same performance with the GTX450 it means that both cpus have enough performance and the bottleneck is in the GPU. What is the next step ??? upgrade the GPU not the CPU. Every information is helpful when you know how to read it.

Intel chips are no better than AMD chips at GPU bottlenecked scenarios, therefore the reader should just buy the cheaper chip if they're going to play at GPU bottlenecked resolutions.

Well, finally you have come to the point i was trying to make. When you actually game at high resolutions with high image quality settings most of DX-11 games are GPU limited.

You're only telling half the story and provide no explanation for why that's the case. You could easily have done that if you showed that there is a difference but how that difference quickly disappears as the GPU runs out of room, regardless of how much you spent on the CPU.

This is an interesting proposition, you could tell me that from the begging. I will have a look at it

 

[pauldun170]

I tend to look at it this way, you buy a GPU for games of today. You buy CPU for games of tomorrow.

When it comes to gaming on low end "budget processors" such Pentium\i3-2100\AMD FX's...all I can think of saying is "Save up some more money"
At least when it comes to home built boxes.

For low end OEM stuff, your kinda screwed out of OC options so go intel and find a nice vid card.

 

[AtenRa]

Because I had two at 1080p for a long time and my 4.5GHz 1090T was bottlenecking the one card already

I find it very difficult to believe a 4.5GHz 1090T (actually i dont even believe that OC) would be a bottleneck for the GTX470 in 1080p gaming, except a game or two.

Any way, SLI/CF is a different situation altogether. Something worth of another review maybe

 

[pelov]

This is an interesting proposition, you could tell me that from the begging. I will have a look at it

The point is you have no reference in your article. In DX11 games CPUs matter less because... why? There's no DX9 there as a reference to show there's a difference and you never delve into why there's a difference in the first place (threads is the biggie but it's still limited). There's actually a difference between the two chips and it's smaller in one scenario than it is in the other, DX9 vs. DX11, and it should affect the scores at lower resolutions drastically and while roughly the same at DX11 GPU bottlenecks it would still show itself at DX9 GPU bottlenecks.

Do you see what I'm getting at here? Pointing to a DX11 benchmark that's GPU bottlenecked and stating it's GPU bottlenecked therefore CPUs dont matter does nothing but state the obvious and furthermore it doesn't explain why.

It would actually be a really good test to carry out and would show quite a bit of information but for some reason you ran a couple of benchmarks, reached an obvious conclusion that could have (and should have) easily been predicted from the beginning instead of giving the reader something to think about.

I'm not giving you flak because I hate you, I'm giving you flak because I expected more

 

[AtenRa]

The point is you have no reference in your article. In DX11 games CPUs matter less because... why? There's no DX9 there as a reference to show there's a difference and you never delve into why there's a difference in the first place (threads is the biggie but it's still limited). There's actually a difference between the two chips and it's smaller in one scenario than it is in the other, DX9 vs. DX11, and it should affect the scores at lower resolutions drastically and while roughly the same at DX11 GPU bottlenecks it would still show itself at DX9 GPU bottlenecks.

Do you see what I'm getting at here? Pointing to a DX11 benchmark that's GPU bottlenecked and stating it's GPU bottlenecked therefore CPUs dont matter does nothing but state the obvious and furthermore it doesn't explain why.

It would actually be a really good test to carry out and would show quite a bit of information but for some reason you ran a couple of benchmarks, reached an obvious conclusion that could have (and should have) easily been predicted from the beginning instead of giving the reader something to think about.

I'm not giving you flak because I hate you, I'm giving you flak because I expected more

Well the review is not an evaluation of DX-9 performance vs DX-11 performance with the same hardware. The review is about gaming performance at DX-11 games with real life actual image quality settings. It was meant to show that with those settings the GPU is most of the times the limiting factor and no matter what CPU or how much you OC your CPU it doesnt matter because your GPU will hold you back.

Now, i would like to make another review and evaluate the performance scaling from 1024x768 all the way up to 1080/1200p or even higher with the same hardware but that is a different review for a later date.

Thanks for the recommendations

 

[BallaTheFeared]

(actually i dont even believe that OC)

Oh ye of little faith...

You think the guy who runs everything he owns to the bleeding edge wouldn't crack out some decent numbers with a 1090T on water?

4,515MHz, 3,225 CPUNB... I'd take Phenom II over Bulldozer in a heartbeat.