Phenom II 965 seems faster in some games than I7 920

[Timman_24]

Why not just get the Intel 750? It is just as fast in games but is much faster in the "number crunching." Hell, the 750 hits 3.6Ghz without a voltage increase and is stable at 4.2Ghz all on air (http://www.tomshardware.com/reviews/...ck,2438-3.html.) It costs 199 at newegg and you get the rock solid stability of intel chipsets.

I'm with nater.

[anubis44]

I don't even understand what you're saying here. Are you saying you can't upgrade an Intel processor? Are you trying to say that AMD doesn't change sockets? What is your point exactly?

Don't understand? Well, I'll explain it very slowly. It's called an 'example.' I was giving you an example of a situation where it might NOT make sense to buy an Intel-based system right now. I'll break it down into itty, bitty steps. Suppose you already have a socket AM2 or AM2+ system with an A64 or a Phenom I chip. You want to speed up the CPU portion of your system, so you buy a Phenom II chip and just put it into your existing system. THAT'S my point. Yes, you can buy faster Intel chips, but going from, say, a Core 2 Duo or Core 2 Quad to any i5 or i7 chip means a new motherboard and ram. That's my point exactly.

[anubis44]

Why not just get the Intel 750? It is just as fast in games but is much faster in the "number crunching." Hell, the 750 hits 3.6Ghz without a voltage increase and is stable at 4.2Ghz all on air (http://www.tomshardware.com/reviews/...ck,2438-3.html.) It costs 199 at newegg and you get the rock solid stability of intel chipsets.

I'm with nater.

Since I'm probably going to get a Radeon 5870 card soon anyhow, and it's faster than the i7 for video transcoding

http://www.guru3d.com/article/radeon...review-test/25

I don't really see the point of switching platforms. Fact is, the 'number' cruching stuff has already started moving to GPUs from CPUs, so Intel's only real temporary advantage over AMD is becoming irrelevant. The only issue for me is gaming speed, and for the games I play, the i5 and i7 either aren't any faster, or are faster by such a small margin that it doesn't make sense for me to switch. And by the way, I'm actually pretty darn happy with the 'rock solid stability' (lol) of my AMD chipset, thank you very much.

[Nater]

Since I'm probably going to get a Radeon 5870 card soon anyhow, and it's faster than the i7 for video transcoding

...if you want your encoded videos to look like ***, yes it is faster. AMD's AVIVO video converter produces absolutely terrible artifacting. The only decent GPU-based video encoding is for nVidia GPUs and even then doesn't have any real advantages over CPUs. At least in the consumer realm, professional stuff developed by Elemental for nVidia GPUs (re Quadros) is supposedly really good. Do yourself a favor, get Handbrake (front-end for x.264) and wait on that Phenom II.

Video encoding and transcoding isn't a terribly good application for a GPU, at least not anything that isn't named 'Fermi' (potentially, maybe not even that, I'm not a software engineer). Right now, all steps of the process can't happen on the GPU so data has to be shuffled back and forth between the main system RAM and the GPU's RAM, then between global and private VRAM and the other way around. The PCIe bus is a huge bottleneck when it comes to this. Fermi is going to be a big step forward as far as GPU compute goes, but massively parallel processors are probably going to need a faster way to access system memory if they're going to be really effective. Some sort of standard slot that can interface with HyperTransport and QPI would be the way to go. But for now, GPUs are only really good if you can cram all of your code into it's on-board memory and you don't have a need to address other system resources.

Not to mention, they didn't even say what software they were using in that test. Guru3D is usually pretty damn good, it surprises me that they left out such an obvious detail. I'm guessing that it was Cyberlink's Espresso setup, which produces alright quality on CPUs and nVidia GPUs with less terrible than AVIVO quality on AMD GPUs. Don't believe me? Have a look for yourself.

[Soul Assassin]

Don't understand? Well, I'll explain it very slowly. It's called an 'example.' I was giving you an example of a situation where it might NOT make sense to buy an Intel-based system right now. I'll break it down into itty, bitty steps. Suppose you already have a socket AM2 or AM2+ system with an A64 or a Phenom I chip. You want to speed up the CPU portion of your system, so you buy a Phenom II chip and just put it into your existing system. THAT'S my point. Yes, you can buy faster Intel chips, but going from, say, a Core 2 Duo or Core 2 Quad to any i5 or i7 chip means a new motherboard and ram. That's my point exactly.

So saying that buying AMD represents a better solution when you're upgrading a CPU with the same motherboard represents an advantage over Intel? Seriously? You actually use that as an "example"? How is it even relevant to a discussion comparing AMD to Intel?

If you're upgrading a motherboard than you're locked in to whatever socket your motherboard has.

[Timman_24]

Since I'm probably going to get a Radeon 5870 card soon anyhow, and it's faster than the i7 for video transcoding

http://www.guru3d.com/article/radeon...review-test/25

I don't really see the point of switching platforms. Fact is, the 'number' cruching stuff has already started moving to GPUs from CPUs, so Intel's only real temporary advantage over AMD is becoming irrelevant. The only issue for me is gaming speed, and for the games I play, the i5 and i7 either aren't any faster, or are faster by such a small margin that it doesn't make sense for me to switch. And by the way, I'm actually pretty darn happy with the 'rock solid stability' (lol) of my AMD chipset, thank you very much.

If you are okay with spending the same amount of money on an inferior chip then go for it. I still don't understand why you choose to do that unless you are just a brand fan. So what you are saying is that you are going to go with the chip that costs the same yet is inferior because you have another device that will offset it only in video encoding?

[anubis44]

If you are okay with spending the same amount of money on an inferior chip then go for it. I still don't understand why you choose to do that unless you are just a brand fan. So what you are saying is that you are going to go with the chip that costs the same yet is inferior because you have another device that will offset it only in video encoding?

That's the point, it's not clearly inferior to the i5 750. They're pretty much equal - one loses to other in a similar number of benchmarks, and not by very much. It gets even sillier when you consider going with a Phenom II 550BE. I paid $120 Canadian for my 550BE, so they're not equal in price, and I didn't have to upgrade my mobo or ram, so I saved even more.

The discussion was ORIGINALLY about why the i7 didn't seem to beat the Phenom II 965 in several games at 2560 resolution with an ATI 5970 card, and then it evolved into a discussion about when it would make sense to buy a Phenom II vs. an Intel i5/i7 cpu. I gave an example of when the AMD upgrade makes more sense, and you guys just can't seem to wrap your minds around that being possible. I don't really see what's so complicated. If I ALREADY have an AM2/AM2+ motherboard, I can either spend $195 for JUST a CPU, or I can spend $199+mobo+possibly DDR-3 ram.

If somebody DOESN'T already have an AM2/AM2+ mobo, and wants to do a new build, then sure, the i5 is an attractive (but by no means a no brainer) option.

[anubis44]

Video encoding and transcoding isn't a terribly good application for a GPU, at least not anything that isn't named 'Fermi'... ...Fermi is going to be a big step forward as far as GPU compute goes

Ah, so now the truth comes out. Not merely an Intel fanboy, but an NVidia one as well. You must be really pissed that ATI is eating NVidia's lunch right now.

The fact is, ATI and NVidia GPUs both produce single or double precision floating point calculations. In the end, that's all they do, and that's why they're faster than x86 floating point architectures right now - because they're optimized from the ground up to do only that. They use all sorts of slightly different architectural tweaks from one another to arrive at this, but ultimately they produce EXACTLY the same kind of calculations. Saying NVidia's double precision floating point is somehow 'better' floating point than ATI's is ludicrous and you know it. If there's any variation in the quality of the transcoding produced by either one of them (not the speed, the QUALITY), that is PURELY a software issue at this point, since they both do double-precision.

By the sounds of what you seem to know, you should know better than to argue this kind of tripe. Shame on you.

[Soul Assassin]

I don't think buying the best option available makes someone an Intel fanboy. I've owned 3 different AMD systems during the period when they were better option. And I only buy AMD video cards. But I doubt any of this matters to you.

[anubis44]

I don't think buying the best option available makes someone an Intel fanboy. I've owned 3 different AMD systems during the period when they were better option. And I only buy AMD video cards. But I doubt any of this matters to you.

All of it matters to me, or I wouldn't be on this forum. You keep missing the point. The 'best' option is not only one option. Life isn't black and white. What makes sense in one scenario doesn't apply to other upgrade scenarios. If you are looking to build a brand new system from scratch, and you can spend more than $200 on a CPU, then you can justify building an Intel setup. No problems there. I have never argued an Intel setup doesn't make sense for some people. What you don't seem to get is that going Intel might not make sense if you already have an AMD system, you aren't a professional videographer or person who uses heavy duty photoshop, and if you can drop in a Phenom II without buying any other parts. It's such a simple argument, I can't see why you're missing it.

[Soul Assassin]

I just don't understand what you're trying to argue. If you're not planning on replacing your motherboard than obviously a processor which will fit in your motherboard is the cheapest option. Whether it's an AMD or Intel. Does it even needed to be mentioned? It just sounds kind of irrelevant. I thought we were talking about building a new system. Rereading your original post makes it sound like you're looking for what to recommend for a new system, not an upgrade. Maybe I'm just misreading it.

And you don't even need to spend $200 on a CPU. For $125 you can get an i3-530 which will OC like mad.

If you're building a NEW machine (or even replacing your motherboard as well as your CPU) I can't see a single viable reason to get an AMD solution unless you just don't want to buy Intel, which is what I said already. There's nothing wrong with that reason so I don't see the big deal about it. Even when Japanese cars were notably better than American ones people still bought American cars just because they wanted to support them.

[anubis44]

I just don't understand what you're trying to argue.

Well, I have made it easy enough for a 5 year old to understand, so at this point, I give up.

Yay Intel!! Woohoo!! Intel's better!! ALWAYS, no matter what!!

[Nater]

Ah, so now the truth comes out. Not merely an Intel fanboy, but an NVidia one as well. You must be really pissed that ATI is eating NVidia's lunch right now.

The fact is, ATI and NVidia GPUs both produce single or double precision floating point calculations. In the end, that's all they do, and that's why they're faster than x86 floating point architectures right now - because they're optimized from the ground up to do only that. They use all sorts of slightly different architectural tweaks from one another to arrive at this, but ultimately they produce EXACTLY the same kind of calculations. Saying NVidia's double precision floating point is somehow 'better' floating point than ATI's is ludicrous and you know it. If there's any variation in the quality of the transcoding produced by either one of them (not the speed, the QUALITY), that is PURELY a software issue at this point, since they both do double-precision.

By the sounds of what you seem to know, you should know better than to argue this kind of tripe. Shame on you.

The only fanboi here is you buddy, and you don't know what you're talking about, clearly. Giving me a 'lesson' on how GPUs do FP intensive tasks better than x86 (or any scalar architecture) only goes to show it.

As for the borderline insane fanboy charge, I probably have more AMD and ATi hardware sitting in my closet than you've ever owned. This is just what I can remember off the top of my head:

AMD: Athlon 500MHz, Athlon 3700+, Opteron 165, Opteron 170, Athlon X2 4800+

ATi: Radeon 8500, Radeon 9700, X1900XTX, X1900XT CrossFire and I have a 4890 running in a PC.

I have four nVidia cards and three intel processors. Tell me how many intel processors you've owned before you start throwing the word fanboy around. I buy the hardware that is going to suit me best when I'm buying. As far as CPUs goes, that's been intel for the last three years and I don't see that changing anytime soon, if ever. The next time I need a GPU, it'll probably be nVidia. Why? Because in the middle of this year they'll most likely have a faster card and here is the kicker: their stuff actually has real drivers for Linux.


GPU compute is 80% about software assuming the chips are compliant with the same IEEE specification. So who has the better software implementation? It's not AMD, what do they even use? Is it Brook+? Is it CTM? Is it Stream? Is it OpenCL? Seriously, how many times have they fumbled that ball? Eventually something like OpenCL or Direct Compute is going to come around, it'll probably end up just like Direct 3D vs. Open GL did as unfortunate as that is. If you need an example of how bad it is on the AMD side for GPU compute, just look at the only really widely used GPU compute program in the PC world, Folding@home. How about the professional world? How many workstations run FireGL cards? ~10% How many servers do you see out there with AMD Stream cards in them? Few. Do you see US National Laboratories talking with AMD for GPU pumped supers? Nope.

Then again, saying GPU compute is completely a software problem is also incorrect. How much level one and level two cache does the GPU have? Does it support ECC memory? How efficient is the dispatch? How many threads can you run? Theoretical peak floating point figures don't mean a whole lot. Theoretically, a HD 5870 will pump out a good bit more SP FLOPS than GT300, care to take a bet on which one ends up faster in real life running a bench like SGEMM or actual applications?

All you can do is regurgitate some boilerplate explanation written in every article that mentions computation of GPUs. 'They're faster than x86 cuz they're build for floating point'. Well no **** Sherlock, of course they're theoretically faster for FP intensive applications. Too bad they usually can't do it for a host of technical reasons that I just barely touched on in my last post that you apparently didn't even notice.

Now, back to the original argument about CPUs. The K10 architecture is just plain slower clock for clock than Nehalem. Period. Intel's speed more than makes up for the higher price of entry. AMD can only compete on price (desktops) or by using 50% more execution cores than intel (servers).

[Timman_24]

I think this new Tom article shows what we are talking about:

http://www.tomshardware.com/reviews/...pu,2499-8.html

This is the games portion. They locked the clocks at 2.8ghz and tested the chips. The Intel chips pretty much dominated. Hyperthreading hurt the i7 on a few of the games tests. But if you turn that setting off it totally beats the AMD. This is properly what happened in that article the OP posted.

[fluffmonster]

Benchmarks like that though don't permit the main optimization objective where AMD is at all competitive, which is a minimum-cost configuration that achieves a certain level of performance. If we ask instead what the lowest cost config would be to get "acceptable" framerates, the answer to that can in many cases be an AMD system. If one's definition of acceptable there is 45-50fps, then an intel fps advantage when fps is already in the 70s or more is irrelevant. I'm not aware of any review which looks at a minimal-cost measure which is probably why Intel is so dominant in comparisons. The minimal-cost measure does implicitly show up in budget gaming build recommendations though which is why we see more AMD representation in that group than typical benchmark comparisons would suggest.

This doesn't do anything to support a recommendation of a PhII 965 which is obv not in that budget category, nor does it really do much to give faith in the sustainability of AMD as a firm since a "budget king" crown delivers neither margins nor volume in this industry, but it does mean that claims that AMD is never the best choice are suspect. Maybe the i3 cpus will change this analysis, but given the relatively high prices on those pieces and the price disadvantage on mobos its not a given.