New Build Phenom II or i7

[sammorris]

More specifically, that's all I get at ANY resolution, 800x600 as that's the CPU cap, with a 3.9Ghz Core 2 Quad. Of course, drop down from Enthusiast and you get a much higher cap (it's 50-60 odd with Gamer mode). However, Quad CF scaling in Warhead is poor, I'm guessing about 1.7-2.2x. (In Left 4 Dead I get 3.5x). The scaling is so bad, CF regularly drops out and returns the other GPUs to idle, so you have to mess around opening the game menu and closing it to force the others to do some work.

DanceswithUnix: That argument was in response to someone claiming the LAG775 socket was 'dead' - I would never use popularity to judge how good something is, we all know the public's taste in things is pretty dire!

[sammorris]

Rough power usage figures extrapolated from graph (Idle/Load)

Core 2 Quad Q9300: 50W idle, 75W load
Phenom II X4 810: 50W idle, 90W load
Core 2 Quad Q9550,Q9650: 55W idle, 95W load
Phenom II X4 920: 50W idle, 105W load
Phenom II X4 940: 50W idle, 110W load
Phenom II X4 955: 55W idle, 115W load
Core i7 920: 80W idle, 150W load (high result probably partly due to X58 chipset consumption)

Rough performance figures extrapolated from graph:
Phenom II X4 810: 101.7
Q9300: 104.5
X4 920: 108.4
X4 940: 114.7
X4 955: 122.7
Q9550: 123.7
Q9650: 130.3
i7 920: 133.5

(est) Phenom II X4 @ 3.7Ghz: 141.9
(est) Phenom II X4 @ 3.8Ghz: 145.7
(est) C2Q 6M @ 3.6Ghz: 150.5
(est) C2Q 6M @ 3.8Ghz: 158.8
(est) C2Q 12M @ 3.7Ghz: 160.7
(est) C2Q 12M @ 3.9Ghz: 169.4
(est) C2Q 12M @ 4.1Ghz: 178.1



Mean Power cost (from 6 different usage cycles: 6i, 12i, 6i+6l, 6i+12l, 24i and 24l) (Baseline cost assume no graphics load, figures based on 14ppu electricity)
Q9300: Basline (Total system cost £108.33 per year)
Phenom II X4 810: +£10.73 /year
Core 2 Quad Q9550/Q9650: + £14.39 /year
Phenom II X4 920: + £16.61 /year
Phenom II X4 940: + £20.10 /year
Phenom II X4 955: + £23.34 /year
Core i7 920: + £56.81 /year

[DanceswithUnix]

DanceswithUnix: That argument was in response to someone claiming the LAG775 socket was 'dead'

Yep, I guessed.

TBH I think the whole "dead platform" thing is a bit of a red herring. I have an LGA775 machine here on my desk, with a P4D in it. If I pop down to my local Novatech store and pick up a CPU to replace this one I have a choice of, drum roll..., nothing. Nada. Not a bean. That is just the way of LGA775, you might get an old CPU in a new motherboard but you will never upgrade an old motherboard with a new CPU.

So, LGA775 is as dead now as it has always been, but I think most of us don't really care that much.

[sammorris]

LGA775, realistically, is many sockets in one. It was about with the latter days of Socket 754, the entire run of Socket 939, and now with AM2, AM2+ and AM3, 775 is still alive and kicking. In all parts of the market, it's AMD vs Intel, and it's AM2/2+/3 vs 775. For the best performance of all for those willing to pay the extra (in up front costs and electricity bills), the i7 is on its own. While it's selling well for an enthusiast product, the mainstream socket for intel is 775, and I don't even see i5 doing much harm when it first comes out.

[DanceswithUnix]

LGA775, realistically, is many sockets in one.

This is my first point. You can't say LGA775 has been around for all that time when, in reality, each incarnation of 775 socket lasts maybe 6 months before there is a pin definition change here, voltage change there, bus speed change maybe and your motherboard becomes obsolete.

My second point is that not many people actually seem to care about that. Personally I like to do a CPU refresh as well as a GPU refresh in a motherboards life, but I seem in a minority there.

[sammorris]

Well, the socket is adapted to take newer designs, but never in such a way that a modern board doesn't take an old CPU. Granted, this isn't much use for most people, but it is of some help. The main consideration here is chipset generation and we've had several of those.
I have actually had three different CPUs in the same board (and will fit a fourth into it later, but that's due to it going in a different system). I had an E4300 in a P35C-DS3R originally, upgraded to an X38-DS4 for dual graphics, and kept the E4300 initially. It was then superceded by a Q6600, and latterly by the Q9550. It was replaced with the current Maximus II after discovering it could not handle Quad crossfire, and ironically, neither could its initial replacemet, the X48-DS5, which is why it's been sent back to ebuyer, by the looks of things for a refund as they now only stock the DS4 (Again ironically, the original board I originally wanted instead of the more expensive DS5, but was out of stock everywhere). Sad thing is though, the DS4 is less than 60% of what I paid for the DS5 back then, ouch. The X38-DS4 I still have, and rather than sell it on ebay (it's beyond 12 months age now) for no more than I could buy a cheap board for my third rig for, I'm keeping it, and an E5200 is going in it, for my silent 24/7 system (using a noisy, illuminated Quad CF system to do work and reply to emails is nuts, and I want something quiet I can use as an HTPC, My fileserver doesn't cut it as it has too many noisy HDDs)

[dangel]

More specifically, that's all I get at ANY resolution, 800x600 as that's the CPU cap, with a 3.9Ghz Core 2 Quad. Of course, drop down from Enthusiast and you get a much higher cap (it's 50-60 odd with Gamer mode). However, Quad CF scaling in Warhead is poor, I'm guessing about 1.7-2.2x. (In Left 4 Dead I get 3.5x). The scaling is so bad, CF regularly drops out and returns the other GPUs to idle, so you have to mess around opening the game menu and closing it to force the others to do some work.

...and that's some of the reasons why I don't bother with multi gpus. I can only pray for a proper hardware solution one day whereby the cards delegate rather than the drivers. Add on top of that heat, power, cost, noise etc.. OTOH I still have a hankering from the point of view of an addict.

[sammorris]

To be fair, that's one of the worst examples of all. Granted, had I not managed to get my pair of X2s for £536 delivered versus the £650+ retail price, I wouldn't have bothered, but frankly, the performance in some games really does justify the cost. Left 4 Dead at 2560x1600 with 4x or even 8x AA with no lag at all except a little physics lag if someone blows up several gas canisters at once (CPU side), Supreme Commander:Forged Alliance is the same, yet was barely playable at any setting on an HD3870, the only big disappointments are Crysis / Crysis Warhead and Far Cry 2. The power is a bit absurd though, I'll grant you that, and amen to a hardware-side dual graphics solution. Still, the X2 products versus proper CF are a step in the right direction. Unfortunately for Quad CF you need to use both

[shaithis]

but frankly, the performance in some games really does justify the cost. Left 4 Dead at 2560x1600 with 4x or even 8x AA with no lag at all except a little physics lag if someone blows up several gas canisters at once (CPU side), Supreme Commander:Forged Alliance is the same, yet was barely playable at any setting on an HD3870, the only big disappointments are Crysis / Crysis Warhead and Far Cry 2. The power is a bit absurd though, I'll grant you that, and amen to a hardware-side dual graphics solution. Still, the X2 products versus proper CF are a step in the right direction. Unfortunately for Quad CF you need to use both

I get no lag what-so-ever in every game i play with a single x2 with all options maxed @ 2560x1600 (with the exception of the Crysis series of course, which still runs very well @ 1920x1200)....in fact never does Left4Dead glitch in the slightest....so there is something wrong with your setup or your CPU isn't powerful enough.

Are the x2 definitely running at 16x PCI-E?

[sammorris]

No, I use a Maximus II Formula which is P45, hence 8x (though it is PCIe 2.0, so still 16x old PCIe bandwidth, or close to it)
When I say there's lag when the barrels go up, it ONLY happens in the new Survivor map Lighthouse with all the barrels clumped together, the frame rate never dips below 55 otherwise.
Both the full 16x boards I've had can't run two graphics cards without the PCI express controller crashing, so I bought a mate's P45 board instead, which suffers no such issue.

[scaryjim]

Scarjim: Exactly as I predicted, the 955 does better, but falls behind the Q9550.

That's an interesting interpretation of the results! The Q9550 beats the X4 955 in single-threaded Pi-fast, but in the multithreaded tests (which are surely more important if you're buying a quad core processor) the X4 955 beats the Q9550 in all three tests, and the Q9650 in two! How exactly does that show it "falling behind" the Q9550?

[sammorris]

The Q9550 is ahead in all three of the game tests, 2/4 of the Megatasking tests and wins 2 of the three initial tests (Bandwidth and PiFast)
That makes it 7 wins to the Intel, 3 wins to the AMD.

[scaryjim]

The Q9550 is ahead in all three of the game tests, 2/4 of the Megatasking tests and wins 2 of the three initial tests (Bandwidth and PiFast)
That makes it 7 wins to the Intel, 3 wins to the AMD.

Sorry, you're wrong. 1) there are only 3 megatasking tests (the first graph is the ordinary DivX with no other tasks). 2) The Q9550 only wins on the gaming benchmark during megatasking, giving 2/3 to AMD. 3) The difference in all the gaming tests in imperceptible due to the extremely high base frame rates (> 80fps throughout). 4) You seem to have forgotten those all-important multithreaded test: even giving Intel the gaming tests I make it 6 - 6 overall.

In fact, the difference in all the benchmarks is pretty imperceptible. My personal opinion is that you simply wouldn't notice the difference in a blind test between the Q9550, Q9650 and X4 955 BE. And, indeed, I didn't come here claiming the 955 BE beat the ass off of everything else. I came here making the point that AMD's Phenom IIs are competitive against Core 2 Quads, and the Hexus review suggests just that - the 955BE gives as much as it takes against the 9550; gives a little more than it takes against the 9650. Given it's priced around the same as a 9550, I'd say that makes it competitive.

[sammorris]

There are definitely occasions where I would notice the Phenom II 955 being slower than the Q9550 (but only in the same sort of way as I'd notice the Q9550 being slower than the Q9650). That said, it doesn't make the CPU worth the same amount of money. The Q9650 is £50 more than the Q9550, and I don't think you'd immediately notice the difference with that either, but you'd still pay £50 more for it. Hell, I would. The only reason I didn't is that the gap was £100+ at the time I bought my CPU.

Also, who cares what the test was in the list, that doesn't make it any more or less relevant. There's nitpicking and the there's just being perverse.

Page 6: Intel, AMD, Intel
Page 7: Intel, AMD, AMD
Page 8: Intel, Intel, Intel (Though the last test is appreciably irrelevant)
Page 9: AMD, AMD, AMD, Intel

Intel: 7 wins
AMD: 6 wins

Overall, it's about even. Percentage-wise though, things are different. The Intel is rarely beaten by more than 2-3%, which really is irrelevant. Only when the balance shifts the other way round do the 8, 10, 12% figures show up. On merits of performance alone, I'd buy the Q9550 if it was priced anything up to £30 more than the AMD. On merits of performance, power and heat, I'd go as far as £50. Fortunately, I don't have to, they're the same price.

What the Phenom II X4 955 does allow is to get very similar to Q9550 like performance out of an existing AMD system at a competitive price, which is great.

[aidanjt]

Also, who cares what the test was in the list, that doesn't make it any more or less relevant.

Of course it's relevant. You wouldn't buy a CPU which excels in every benchmark except media when you want to use your new machine for editing video and transcoding.

[sammorris]

Of course, but if it's a case of video encoding by itself, or video encoding and multitasking, the people it caters for are the same.