Are AMD Current CPUs Not Enough For Solid 60 FPS Anymore?

[theonedub]

I believe the root of this issue is AMDs desire to maintain backwards compatibility. Hopefully AM3+ will really contribute the the performance gain of bulldozer. Also for those in doubt; massive game difference - http://www.anandtech.com/bench/Product/109?vs=147

Are those benches really accurate? I thought the performance gap was smaller- those results look pretty one sided.

 

[cadaveca]

Are those benches really accurate? I thought the performance gap was smaller- those results look pretty one sided.

So far my results seem to mimick those anadtech ones...in cpu-intensive apps, AMD wins, but Intel wins on 3D. Even better, alot of times 8x8 PCI-E Crossfire gives better performance than 16x16 PCI-E... and I do not understand why.

Also, there is no mention of drivers used, etc...too many unknown variables to call those metrics "fact", so I agree they are questionable.

 

[Dent1]

Robert-The-Rambler,

The problem is on your side, believe me. Ghostbusters isnt my idea of an intensive game and if you can not maintain 60FPS its on your side.

Download the Final Fantasy XIV amd 3D Mark Vantage Benchmark and lets see if there is a bottleneck or if its driver related.

 

[erocker]

lets see if there is a bottleneck or if its driver related.

That's what cadaveca is trying to point out. The bottleneck with ATi's latest drivers are the drivers! Crazy stuff, but through some little testing I did, it indeed has merit... plus it's dissapointing.

 

[Dent1]

That's what cadaveca is trying to point out. The bottleneck with ATi's latest drivers are the drivers! Crazy stuff, but through some little testing I did, it indeed has merit... plus it's dissapointing.

To me this thread seems flame baity. I can not remember the last time I couldn’t sustain 60FPS average on my crusty rig (by today's standard). Robert-The-Rambler has posted once and once only in this thread so far. He seems like the type of guy that posts stuff to get people worked up then disappears and sits back and watches us fools argue.

Where can I download the Dragon Age benchmark from? I have a Athlon II X4 which I would like to test.

 

[erocker]

To me this thread seems flame baity. I can not remember the last time I couldn’t sustain 60FPS average on my crusty rig (by today's standard). Robert-The-Rambler has posted once and once only. He seems like the type of guy that posts stuff to get people worked up then disappears and sits back and watches us fools argue.

Where can I download the Dragon Age benchmark from? I have a Athlon II X4 which I would like to test.

Well, I can attest with a 5850 mildly overclocked and my 965be at 4ghz I easily get 60fps with both Dragon Age and Ghostbusters. There are no benchmarks for them unfortunately, you need to use Fraps. Tonight I should do a test of these games with my CPU at stock and overclocked.

 

[cadaveca]

Dragonage is cpu-limited(i think cache-limited, unsure of mem, drive, or what...)...I mentioned this a long time ago, using my 4890 rig as an example..I was getting over 60FPS...barely...but vgas were only running @ about 66% utilization. Cpu wasn't maxed out...vgas weren't either...but something was keeping the framerate down.


I look forward to seeing what you find, e.

I'm working through my steam games, then I'll move onto others...Mass Effect has proven interesting, but my plan was to hit up Dragon Age afterwords.

As an aside, afterburner also reports FPS, and can then write it to a log file....

That's what cadaveca is trying to point out. The bottleneck with ATi's latest drivers are the drivers! Crazy stuff, but through some little testing I did, it indeed has merit... plus it's dissapointing.

yes, exactly. I also think that AMD is unaware of the problem, as I think most of thier driver team is working on i7-based machines, and i7 doesn't run into these limitations as quickly...but is still a victim of the same behavior.

 

[dir_d]

Cadaveca did you read that Anandtech article i gave a link to. It talked about the HT link bandwidth and how the stock 2Ghz is not enough to saturate the whole HT pci-e 4x bandwidth. 2.8ghz+ is where we start seeing the pci-e 4x bandwidth being saturated fully, thus resulting in better performance. Read it...its interesting.

 

[erocker]

Cadaveca did you read that Anandtech article i gave a link to. It talked about the HT link bandwidth and how the stock 2Ghz is not enough to saturate the whole HT pci-e 4x bandwidth. 2.8ghz+ is where we start seeing the pci-e 4x bandwidth being saturated fully, thus resulting in better performance. Read it...its interesting.

Thing is, it's not the HT bandwith. It's the north bridge or more specifically the IMC frequency that is causing the problems. It is also stock at 2ghz like HT and it's not enough.

 

[Robert-The-Rambler]

Here is what I'm looking for

This is not flame bait thread. I've been a message board moderator before and I know what that is. I'm not even trying to start a debate about the difference in technologies. Everybody knows the I7 Socket 1366 platform is the best performing gaming platform. I just want to know if anybody else out there knows what other games I might have trouble trying to play with my AMD setups.

I'm looking for games that AMD CPUs, even the latest ones like the X6s, just simply can't do 60 FPS for the majority of the time with Vsync enabled. It might even be a platform limitation but in order to use AMD CPUs you have to accept the limitations of the given platforms whether they have DDR2/DDR3. I stated an observation that I noticed when utilizing FRAPs. I was hoping some of you guys might also try just playing the actual games for a while just watching the FRAPS counter. I'm more interested in the real world gameplay then benchmark results.

I also want to mention that the AMD setups I have are running 4 gigs of DDR2 800 RAM and one of them is running 4 gigs of DDR 1066 RAM.

P.S 3 4850 1 gig cards in triple crossfire can play Metro 2033 at very high DirectX 10 at reasonable FPS. (Not 60 solid but playable) They are enough for most things. Nobody even mention Crysis. I may launch a cruise missile your way.

 

[cadaveca]

Read it...its interesting.

I did. And as erocker says, it's IMC speed(CPU-NB) that they are testing, not HTT. I get very minimal gains from HTT clocking...there are some, but CPU-NB gains are far more drastic.


Now, again, increasing CPU-NB increases L3 speed, and Intel's L3 speed, as well as memory bandwidth, are both higher than AMD, and when I consider that Intel doesn't run into this problem as quickly, it's easy to just say it's those things that make the difference.

However, I am not 100% sure on this, so I am doing this testing, in hopes that maybe the truth will be exposed, so to speak.

In the end, it's not really THAT important...the majority of us are going to overclock, and when overclocked, this isn't such a big issue. But because we have to void our warranties to avoid this problem , I do not feel that the OEMs are really doing us any good having technology on the market that requires overclocking..specifically SLi and Crossfire.

Of course, Sli seems to not be affected by this sort of issue so quickly...so that points a finger @ Ati drivers...which we kinda looked at in the 5850Crossfire thread.

Of course, I've already said all of this, as currently, that's all there is to say. Sli sclaing is better than Crossfire, and it seems a big part of that is CPU bottlenecks...not real bottlenecks, but ones created by how the AMD driver works currently.

And because of it looking to be drivers, I think it's even less important..as those drivers could change at any time...that's up to AMD. However, without knowing about the issue, it's gonna be hard for AMD to fix it...so in the end, my goal is to see if there is any real merit to this, and if there is, to put it in the public's eye, in hopes that it can be rectified.

And to you, RObert, clearly I see some of the same sort of behavior you describe...and I'm currently looking further into this. Who knows..in the end, I could be wrong...adnd that's fine...but noone will truly know until the numbers speak for themselves.

 

[Robert-The-Rambler]

But why does crossfire kick ass on the I7 Platform?

I did. And as erocker says, it's IMC speed(CPU-NB) that they are testing, not HTT. I get very minimal gains from HTT clocking...there are some, but CPU-NB gains are far more drastic.


Now, again, increasing CPU-NB increases L3 speed, and Intel's L3 speed, as well as memory bandwidth, are both higher than AMD, and when I consider that Intel doesn't run into this problem as quickly, it's easy to just say it's those things that make the difference.

However, I am not 100% sure on this, so I am doing this testing, in hopes that maybe the truth will be exposed, so to speak.

In the end, it's not really THAT important...the majority of us are going to overclock, and when overclocked, this isn't such a big issue. But because we have to void our warranties to avoid this problem , I do not feel that the OEMs are really doing us any good having technology on the market that requires overclocking..specifically SLi and Crossfire.

Of course, Sli seems to not be affected by this sort of issue so quickly...so that points a finger @ Ati drivers...which we kinda looked at in the 5850Crossfire thread.

Of course, I've already said all of this, as currently, that's all there is to say. Sli sclaing is better than Crossfire, and it seems a big part of that is CPU bottlenecks...not real bottlenecks, but ones created by how the AMD driver works currently.

And because of it looking to be drivers, I think it's even less important..as those drivers could change at any time...that's up to AMD. However, without knowing about the issue, it's gonna be hard for AMD to fix it...so in the end, my goal is to see if there is any real merit to this, and if there is, to put it in the public's eye, in hopes that it can be rectified.

Don't you think the performance issues could be hardware limitations? My two 4890s are crazy even at 2560 * 1600 is most games with an I7 920 stock on the accelerator. For example part of what inspired this thread is that they simply stayed stuck 60 FPS at max detail in Dragon Age Origins. Now unfortunately I could not read much of the text. Screw Bioware and EA for not handling the 2560 * 1600 resolution properly. You literally need a magnifying glass to read the text.

But again I want to say that I am looking for the other games that hate AMD and simply will not perform "perfectly" at 60 solid FPS with Vsync on.

 

[Reventon]

A 1090T is also $600 cheaper than an i7-980X. Just sayin.

 

[erocker]

But again I want to say that I am looking for the other games that hate AMD and simply will not perform "perfectly" at 60 solid FPS with Vsync on.

I don't know of any. Are you reffering to an AMD CPU at stock speed or overclocked?

 

[CDdude55]

A 1090T is also $600 cheaper than an i7-980X. Just sayin.

And it's also significantly slower then a 980x. Just sayin.

 

[Robert-The-Rambler]

I don't think I ever enabled it

Just curious, did you try disabling Cool 'n' Quiet and running all the chips at full speed? If the frames drop drastically that might be the culprit.

Also if you can put those 4890s in CF on the AMD 920 machine and put the 4850 CF in the I7 machine and see if the issue is still there.

I only enabled cool and quiet with the X3s to get them to run cool at idle since they are used mostly for surfing the web. My little ASRock motherboard even tells you how many watts you are consuming at idle. I think it was 5 or some ridiculous number.

Unfortunately I can't move any parts around. I did that so often in the past I am burned out on doing that any more. A recent nasty injury that required stitches has me apprehensive about doing much handy work inside my PCs.

 

[Reventon]

And it's also significantly slower then a 980x. Just sayin.

Operating Frequency
3.2GHz

Hyper Transports
4000MHz

L2 Cache
6 x 512KB

L3 Cache
6MB

Manufacturing Tech
45 nm

It's still a good processor. Just sayin.

 

[Robert-The-Rambler]

I mean stock

I don't know of any. Are you reffering to an AMD CPU at stock speed or overclocked?

If you have to overclock then something is amiss. Overclocking is supposed to be for fun not a necessity!

Anyhow I am going to try some games out and report back.

I will be using a rig with these specs.

Windows Vista 64 bit Home Premium

MSI K9A2 Platinum

4 gigs of DDR 2 800 RAM

3 Radeon 4850 1 gig in triple crossfire

1 Asus Xonar DS

2 500 gig hard drives not in RAID

2 DVD ROMs and 1 DVD Burner

I'll see if I can illustrate this with FRAPS.

 

[CDdude55]

It's still a good processor. Just sayin.

I agree. Just sayin.

 

[crazyeyesreaper]

okay first things first

scaling with Tri Fire on 4850s on an older motherboard in Vista isnt going to be stellar trifire dosent scale well unless your pushing your CPU to the limits in order to feed those gpus and thats even with older drivers. before they changed the way they did things.

also anything thats not Phenom II aka Athlon II has no L3 this is a 10% blow on average no matter what when equal cores are used at same clocks.

but yea with 3 gpus overclocking is a Necessity.

oh and as far as GPU usage etc i think Cadaveca is onto something here aka how the ATi driver uses cpu time ive always noticed that at lower resolutions Nvidias GPU can walk all over the ATi cards depending on games but as the resolution goes up they even out etc. I think now i have a reason for why that is

 

[TheMailMan78]

I am stil ltesting, but it seems the i7 overclocked still beats out AMD. i7 costs more, so to me, it's no big deal...you get what you pay for.



Again, MM, i'll say this, you are trolling. Either post data that refutes it, or get out. Even Mussels and erocker agree cpu is the bottleneck.

Crossfired 5850's Issue

Oh I'm trolling? I'm not the one that came on to this forum and claimed to have been the creator of the TWKR chip. You walk right of Mordor with that bullshit.

Second I get 60+ FPS in ALL of my games with my current rig and....(gasp) its all AMD?!

If you are not with a 5850 or a 5870 then you have a driver issue or something is wrong with your rig. Thats the bottom line. It has NOTHING to do with the CPU itself.

Here I built you a sand castle in honor of your BS. I call it Mt. Cadaveca.

 

[cadaveca]

Oh I'm trolling? I'm not the one that came on to this forum and claimed to have been the creator of the TWKR chip. You walk right of Mordor with that bullshit.

That was a joke, MM. I guess you've never seen the Windows "It was MY idea" commercials?

If you are not with a 5850 or a 5870 then you have a driver issue or something is wrong with your rig. Thats the bottom line. It has NOTHING to do with the CPU itself.

Um, pay attention dude..I'm saying it's all about the driver. Post reported 4 trolling and flamebaiting.

That's what cadaveca is trying to point out. The bottleneck with ATi's latest drivers are the drivers! Crazy stuff, but through some little testing I did, it indeed has merit... plus it's dissapointing.

 

[TheMailMan78]

That was a joke, MM. I guess you've never seen the Windows "It was MY idea" commercials?








Um, pay attention dude..I'm saying it's all about the driver. Post reported 4 trolling and flamebaiting.

Be sure to include the screen I posted showing a solid 60 FPS in your complaint.

 

[cadaveca]

Just because you don't have any issues, doesn't mean there isn't one. You haven't read the thread, clearly, as this issue only presents itself @ stock cpu, with more than 1600 shaders to push. Again, you are trolling, trying to start a fight here, and I'm not falling for it.

Post yet again reported.

 

[TheMailMan78]

Just because you don't have any issues, doesn't mean there isn't one. You haven't read the thread, clearly, as this issue only present itself @ stock cpu. Again, you are trolling, trying to start a fight here, and I'm not falling for it.

Post yet again reported.

Mine is at stock. Explain that.