I7 920 vs Phenom II 965 with an ATI 5870.(Finally!)

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar



This is exactly what happened to me, except I wised up when I hit the 6400+. I decided that the first phenoms were a waste, and I waited for the Phenom II's. I went from the 3800+ to the 4800+ to the 6400+ within a 2 year span, all to get Unreal Tournament 3 to give me 30fps min at 1920x1080. Before anyone says it was the video cards that I had that were the problem, I had the 1 3870 ever since the 4800+. I got the second 3870 after the 6400+. It wasn't until I got the x3 710 that the FPS were able to hold a minimum of 60fps at 1920x1080. Here was the result of going from the 6400+ to the x3 710:

System Information
Operating system: Microsoft Windows XP
System memory: 2.0 GB
CPU: AMD Athlon(tm) 64 X2 Dual Core Processor 6400+
CPU speed: 3433 MHz
Sound system: SB Audigy 4 [E800]
VGA Information
Graphics card: ATI Radeon HD 3870 edit 3870 x2
Memory: 512.0 MB
Driver version: 6.14.0010.6912 (English)
Benchmark Information
Benchmark type: Flyby
Demo: vCTF-Corruption
Motion Blur: Disabled
Hardware Physics: Disabled
Anisotropic filtering: 16×

Resolution: 1280 × 720 (Custom)
Score = 101 FPS



And after:


The benchmark started at 13/05/2009 6:52:43 PM

System Information
Operating system: Windows 7 Ultimate
System memory: 2047 MB
CPU: AMD Phenom(tm) II X3 710 Processor
CPU speed: 3333 MHz
Graphics card: - 0 MB edit 3870 x2
Benchmark Information
Benchmark type: Flyby
Demo: vCTF-Corruption
Motion Blur: Disabled
Hardware Physics: Disabled
Anisotropic filtering: 16×



--------------------------------------------------------------------------------

Resolution: 1280 × 720 (Custom)
Score = 216 FPS

Now before I get flamed to hard for this bench...Yes I know there was XP for the first and win7 for the second, but notice how much of a CPU bottleneck that was obviously present with the 6400+. Then after the upgrade the FPS went through the roof, the problem was that I had to go through 3 cpu's to get that performance. I don't want this to happen again, if possible for the next 2-3 years. BTW at the time the 4800+ cost me around 180$, then the 6400+ was 230$ and then I got the x3 710 for 150$. That was over 550$ in cpu upgrades over a 2 year period, while the whole time my brother was going on his intel core2 chip. Stupid me..I know. 🙁 3 years later my brother upgraded from his e-6600 to his current CPU, the e-8500. This is the position I would like to be in now if possible.

The 3800+ wasn't enough right off the bat with the x1950 I had at the time, so I got the 4800+ and the first 3870, but the single 3870 also wasn't enough with the 4800+, but it would of been with the intel alternative(e-6600). Me and my brother tried it, and the fps was around 15-20fps higher. Then I got the 6400+ and I will say that that CPU did make it so I could get a minumum 30fps at 1680x1050, but only after getting a second 3870 could I game at 1920x1080 at a minumum 30fps. This was only for UT3 though, Crysis, dirt, fallout 3 and a few others weren't so forgiving. Then I got the x3 710. After that everything was perfect, and still is, but for how long?

Basically I dont want to have to go through this all over again, that is why I like to be informed as much as possible nowadays, so I don't make the same upgrade mistakes I did in the past. If I had my time back, when I went to the 6400+ and this motherboard I have right now, I would of switched to the q-6600.

BTW: The bench tool I used changed after each patch that came out for UT3. That is why it looks different, and displays slightly different information in the results.
 
Well a newer cpu is gonna be better than a 2-3 year old cpu...but that isn't what we're talking about here we're talking about two cpu's that are in the same age bracket, same performance bracket. The 4800+ and e6600 were never in the same performance bracket although the 6400+ was pretty close to it.

Compare the e6600 to the x3 710 and the x3 710 will beat the e6600 easily in everything.

I did a bit of digging and it seems UT3 strongly favoured intel systems back then.

http://www.anandtech.com/video/showdoc.aspx?i=3127&p=6

That's quite an interesting article, and shows yet again that the intels do better at lower resolution and not quite so well at higher. The difference there though is, the intels are still ahead at higher resolution - something we are not seeing today with the i7 vs the phenom II, when the gap is usually removed entirely or more unusually, going in favour of the phenom II at the highest resolutions.
 
It is because at high resolution, the majority of the work is offloaded to the GPU. Its called GPU bottleneck. At lower resolution, where GPU bottleneck is removed, one can make a comparison between those two's performance.

And as data showed, Core i7 is just a much superior CPU than Phenom II, period. Its just that since games don't put much stress on the CPU, gamers usually upgrade their GPU before they upgrade their CPU.
 
If it is a gpu bottleneck why is the phenom II ahead at max resolution?

Look at the graphs instead of making up nonsesne. The UT3 ones there clearly show what should be expected to happen in a gpu bottleneck. The core 2's are still ahead, but at a smaller percentage.

The i7 consistently has it's low-res advantage totally wiped out at 1920x and above. That isn't a gpu bottleneck, that is the result of the i7 lacking l2 cache.
 


Tom's results don't look that different to me. Bloomfield has better performance with 4 GPUs than lynfield. But what my links also show is that phenom II is slower when there are no GPU limitations.


Well the title in that section is: "A Quick Look at GPU Limited Gaming" And your quote says it clearly: "Any CPU near the high end, when faced with the same GPU bottleneck, will perform the same in game."

But what happens when you remove the GPU bottleneck? i7 pulls ahead of phenom II.

http://www.tomshardware.co.uk/core-i5-lynnfield,review-31672.html
http://www.tomshardware.co.uk/core-i5-gaming,review-31673.html



Perhaps Q6600 isn't quite fast enough to reach the GPU limit? I don't know, but like I said: Once you remove the bottleneck i7 pulls ahead.
 


The title on that page reads: "GPU Limited Gaming Oddities"

But as I have repeatedly said: If you remove the GPU bottleneck i7 will pull ahead of phenom II.
 


ROFL.gif
ROFL.gif
ROFL.gif
ROFL.gif


With all due respect, I fail to see how L2 is factored into all this.

sandra-mem-bw.gif


As shown here, i7's memory subsystem performance is much better than Phenom II's. Even with smaller L2 cache, i7 managed to achieve higher bandwidth (thus faster) than Phenom II.

As for the reason why i7 920 lost out at high resolution, it may have more to do with the system architecture than the CPU architecture. As all tests show, i7 simply blow Phenom II out of the water in terms of performance.
 


No but the guy is validating my point easily. It took him 3 CPU upgrades to get the performance he wanted. If he had gone with a Core 2 system at the time he would have just had to upgrade his x1950 to a 3870 and maybe a new CPU, possibly a quad core. Thats a large savings. I so far have gotten one GPU and my system is two years old this month yet still can play every current game maxed without even flinching.

For a lot of people, they don't build a new system every year. They build one main system and upgrade the major components. In gaming thats the GPU. Thats why it is important to look at the CPUs true performance and even advantages it has right now for that. Sure triple channel memory is useless now but what if the next gen of games start to use it? Advantage. Core i7 has half the cache of a Phenom II but still keeps up since its faster and well better. Advantage. Core i7 has higher IPC, no matter what way you put it. Core i7 even at a set speed keeps of 2.8GHz easily keeps up with and can beat a Phenom II @ 3.4GHz. Advantage.

All of these advantages are signs that a Core i7 will be less of a bottleneck in the future. Hell a lot of sites already saw a bottleneck with Phenom II with a tri SLI/quad CF setup yet didn't run into that with Core i7. Not saying Phenom II is bad but its not amazing. Its improved the IPC over Phenom but not enough to make a major difference.

But out of all of it, it really seems like you just like to argue. I mean a CPU is the main factor in low res and at the low spectrum of the FPS. Core i7 shows its able to outdo anything. At a high res like 1920x1080+ the GPU becomes the main factor. The CPU doesn't really do much unless the game is optimized for it and the handful that are, most of those are poorly optimized for it. L4D is one of the only ones but since the Source engine is so old its like using Duke Nukem 3D to compare when no matter what you will get 15K FPS.

Bleh its all useless. To each thier own.
 
Core i7 will be less of a bottleneck in future... Okay, has Pentium IV HT been future proof for users who bought it? And if you test it in Crysis will it be much better than a standard Pentium IV? No... Really, you just need to get that all products have their time, and when that time comes both PII and I7 will be utter garbage. On low resolutions where GPU isn't used as much, I7 wins. Tada. Great. And who plays at 800x600? Nobody, so that is irrelevant. Since I7 and PII play games on high resolutions equally, it is logical that newer games will get even higher details and CPU will become totally unimportant! Get it!
 
Whatever the reason is for it I don't actually know. Anandtech have done more research on this than most I'd wager, and even they aren't 100% sure.

What he do have are the facts, and the facts are the i7 does not maintain it's low resolution 'wins', if you can call 640x480 resolution gaming meaninful in any way. At higher, or real gaming resolution, the Phenom II either matches or wins most times.

Intel fans might like to imagine it's a gpu bottleneck, but the graphs simply point to something else. Not that I'd expect any intel fans to admit that the i7 isn't as good as they believe regardless of what the facts plainly show.

If the i7 was anywhere near as good as most of you believe, it would still be ahead at high resolution. It isn't. The graphs clearly show there is no gpu 'bottleneck', if that was the case then all the cpu's tested would be scoring the same but they aren't.
 
^ of the links you have posted the first one, done with a 2.8 GHz AMD cpu that was out of production when the benchmark was made was one of the worst benchmarks i've ever seen Toms do. It was an attempt to compare clock speeds even though the out of production AMD cpu didn't even have a DDR3 memory controller. The second benchmark just proves the point about gaming with the PhenomII vs the i5/i7. I wish Toms had bothered using a AMD SLI capable motherboard for the testing, but they didn't.
 
Yep I'm not sure what you think those links prove kettu. What we have in those benchmarks is crossfire looking a bit less mature than sli, but that's nothing new. ATI was a long, long way behind Nvidia on multi-gpu setups but they are closing the gap - but the issues are still there that is why some games throw up strange results and why sli'd 285gtx's can sometimes (rarely) beat crossfire 5850's. It's just the crossfire drivers, it's nothing to do with the cpu's.

 
Too bad Phenom II is about the same speed (+/- ~5%) clock for clock as a 3 year old CPU. :) (Q6600) Even in gaming.

This should tell you something.

For gaming, Phenom II is a great choice. But so is a dual core even. I wouldn't invest my money into something that is 20%+ weaker (in things other than gaming, and in gaming possibly in the future) as the competitor, and only as strong as something as old as the Q6600, and weaker per clock than C2D 45nm which has been out for a while now as well.
 
Too bad Phenom II is about the same speed (+/- ~5%) clock for clock as a 3 year old CPU. :) (Q6600) Even in gaming.

This should tell you something.
Click to expand...

It does, it tells me you haven't got a clue. I have both cpu's and the Phenom II walks all over the Q6600 with both at 3.4ghz.
 


DDR3 doesn't improve realworld performance that much. That doesn't explain the performance differences seen here. According to anandtech 3-5% performance increase. ixbt labs doesn't notice any improvement between similar speed modules. Tweaktown doesn't see a huge difference. I'm specifically refering to gaming performance.

http://www.anandtech.com/memory/showdoc.aspx?i=2989&p=7
http://ixbtlabs.com/articles3/cpu/ddr2-800-vs-ddr3-1333.html
http://www.tweaktown.com/articles/1782/amd_phenom_ii_ddr2_vs_ddr3_performance/index11.html

Before you say "different memory speeds":
http://www.tomshardware.co.uk/phenom-ii-ddr3,review-31596-6.html

The biggest difference between slowest 1066mhz and fastes 1333mhz is less than 7%. Looking at at the other gaming tests you see less than 2% and 0% difference. But let's be generous and add 10% to phenom II scores on the first link:

HD4870x2
Stalker, high details no AA/AF
1680x1050 pII 53.4*1.1 = 58.7 vs i7 66.8
1920x1200 pII 52.7*1.1 = 58.0 vs i7 64.9

Stalker, high details 4xAA / max AF
1680x1050 pII 39.0 vs i7 40.5
1920x1200 pII 35.4 vs i7 35.9
Now at these settings pII and i7 are pretty much equal. Giving 10% to pII would put it ahead. But would that be fair? Is pII limited by its memory subsystem or graphics subsystem? Consider the following numbers with 2x HD4870x2:

2x HD4870x2
Stalker, high details 4xAA / max AF
1680x1050 pII 45.9*1.1 = 50.5 vs i7 55.1
1920x1200 pII 42.8*1.1 = 47.1 vs i7 52.2

Both pII and i7 improve their scores over single hd4870x2. But again once the GPU bottleneck is removed i7 pulls ahead. Even when adding a very generous 10% increase to pII score.

If you are interested you could calculate more results yourself. I did a few myself and they conformed with the result stalker gave. So in my opinion the first test is valid and complements the second test very well in demonstrating the performance advantage of i7 when GPU is not a limiting factor.
 


Considering that i7 consistently outperforms pII using crossfire technology with both CPUs it is clear which one is more powerfull.
 


I've posted earlier the evidence that I think proves i7 is a more powerfull CPU. But you should review it yourself and come to your own conclusion. In this post I'll just give you my opinion on your possible upgrades.

While i7 is more powerfull pII has a more attractive price. Unless you get a very good price for your current computer I wouldn't recommend a new i7 system. The speed gain is not that significant in my opinion. A better choice would be to buy a X4 955BE instead. It has a pretty good price/performance ratio especially since you allready have a platform for it. But again in my opinion the speed gain is not significant enough to justify the cost. Unless you can get a good enough price for your current CPU.

I think the best choice would be to forget the CPU upgrade for now and just buy a new GPU. I think a HD5850 would be worth the price.
 


Unfortunately, you're not the first person to come up with that argument. However, I do agree with you. If you just want a CPU that's purely for gaming, i7 is an overkill. In fact, most quads are overkill for pure gaming, as dual cores and tri cores can do the same job for a fraction of the price.

I'm just pointing out that from the performance point of view, Core i7 is a much more capable CPU than Phenom II. As its been shown consistently, Nehalem architecture pretty much outperformed K10.5 in all CPU intensive applications, and achieving almost twice the performance of it in some situations.

Plus, the HyperThreading in i7 is actually very very powerful. At the moment, I can run folding at home on my i7 with 8 threads, and completing work in the same time as an eight core Shanghai (dual CPU) machine.



And which way did it point? L2 cache?
ROFL.gif


In my opinion, it maybe have more to do with the fact that Intel still uses DMI bus between X58 and the PCI-E slots. As data transfer between CPU and GPU heightened due to higher resolutions, the bus may be saturated. Hypertransport bus used on AMD boards, however, have much more bandwidth than the archaic DMI bus. It may be the reason why Intel systems tend to lose out at high resolution (>1920 x 1080), even though the CPU is clearly superior.

EDIT: Looks like I'm incorrect. QPI is used between the CPU and NB, not DMI. However, someone points out that the reason is because QPI is incompatible with PCI-E. Therefore in order to make them compatible, Intel implemented a device in IOH to convert the signal. Therefore the latency in conversion resulted in lower performance.
 


Interesting theory. But there is a discrepancy between results with single GPU cards and dual GPU cards. Phenom II sometimes pulls ahead when a one single GPU card is used. DMI bus staturation looks like a reasonable explanation. But using one dual GPU board (HD4870x2) i7 pulls ahead again. There's something more complicated going on I guess. Unfortunately I have no idea what that might be :).
 


In my opinion for current games all the tested CPUs in my links are mostly the same in practice. You will get playable fps with all of them. Note though that in my second link CPUs are tested at stock speeds. At similar clockspeeds i7 is a faster CPU. Wether one feels like paying the price premium for that is their business. Phenom II has a price/performance advantage in my opinion. Especially when used at stock speeds. Though it decreases somewhat or might even disappear for overclockers.
 
Kettu that benchmark was done with Catalyst 9.6 right? The big improvement for Phenom II crossfire came with 9.8.

I'm sure Jaydee had a link somewhere...I'll try to find it.

http://www.pcgameshardware.com/aid,692942/Catalyst-98-reviewed-HD-4870-X2-up-to-47-percent-faster-failing-in-Anno-1404/Practice/

(Check out the Farcry bench)

http://www.xbitlabs.com/articles/video/display/catalyst-9-8_5.html#sect0

Some more there. Catalyst 9.8 brought the Phenom's in line with the i7's for crossfire performance. It was never that much more powerful, for some reason ATI's drivers just worked better on the Nehalem than it did on the Phenom before 9.8.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom