bfar said:
Yes, but I am stressing because I spent four figures swapping all my existing rigs to Core2Duo and not only are they not any faster, they are actually slower than the old P4's. Look out for a slow trickle of E6600's on MM when it reopens.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
C2D v X2 = Not as good as you think?
- Thread starter FatRakoon
- Start date
Yes, but I am stressing because I spent four figures swapping all my existing rigs to Core2Duo and not only are they not any faster, they are actually slower than the old P4's. Look out for a slow trickle of E6600's on MM when it reopens.
![[HoL]Cobra](/styles/overclockers/avatars/2.png){kind=avatar}
Associate
- Joined
- 23 Oct 2003
- Posts
- 185
- Location
- Coventry, UK
WJA96 said:
Oh deary deary me, you must really have something wrong, like forgetting to plug it in, if your old P4 was quicker than your C2D
Shakes head at some comments on here, worries me sometimes.
PS, If there cheap enough, i'll buy the lot.
Everytime I see these threads I have to laugh.
The simple fact is that C2D should beat an old generation chip Like the X2 4800 with ease. The fact that it took a big corporate giant like Intel to get off its butt and beat AMD so long is testament to how good the AMD chips are. An average speed increase over old technology of 25%, is that really something to shout that loudly about?
Phil
The simple fact is that C2D should beat an old generation chip Like the X2 4800 with ease. The fact that it took a big corporate giant like Intel to get off its butt and beat AMD so long is testament to how good the AMD chips are. An average speed increase over old technology of 25%, is that really something to shout that loudly about?
Phil
Combat squirrel said:
makes sense to me.
Stack up the number of tasks your doing (Super Pi, CSS, 3dmark, etc..) and the AMD copes better due to its use of an onboard memory controller.
it stands to reason a hypertransport bus and onboard memory controller would cope much better with lots of very large requests for data, than an old fashioned FSB and northbridge.
Just looking at the stats, a 2ghz AMD Hypertransport bus has a bandwith of 16GB/s whereas an 1066mhz bus is only 8.5GGB/s
so hypertransport has double the memory bandwith and double the speed of an intel 1066mhz bus. This of course doesnt normally make a difference, as playing 2 things at once is unlikely to saturate all 8.5GB/s bandwith. But stack up Super Pi, Prime, 3Dmark, CS etc... and you will saturate the bus. But the AMD bus, with twice the bandwith doesnt saturate.
seems to make sense to me. But given that nobody would want to run all these things at once, it wont make a difference for everyday use.
Last edited:
Soldato
- Joined
- 7 Aug 2004
- Posts
- 11,175
Yup, also another thing I thought of, while sitting on the bog no less (all the best ideas are created there! haha) Intels quad pumped bus was designed for use with netburst architecture, a cpu that did few calculations very fast, conroe does more calculations in one cycle, thus its design is MUCH better suited to a hypertransport, on die mem controller config. Akin to 50 lane road (ram) going into 2 lanes (intel quad system bus) back to a 50 lane road (processor), it will all get clogged up in the system bus
Again INTEL sort out your system bus specification thats suited to the way your now doing processors ! lol
Again INTEL sort out your system bus specification thats suited to the way your now doing processors ! lol
They are sorting it out 
It's called Common Serial Interconnect or CSI and is slated for Q4'07 last I heard.
The current FSB design isn't really holding Conroe back but yes I agree that it's not good for scalability or going forward. The quad core chips especially are pushing the FSB to its limits. Some of Intel's server boards have dual FSBs, one for each CPU socket, but obviously this is just a temporary workaround.
An on-die memory controller won't help the Core architecture much at all. Precaching tends to erode the benefits of it. It probably could improve performance by a few percent but at the cost of die space that would be better spent on L2 cache or more branch prediction logic. Plus there's the old bad boy called "heat" which Intel has a new found hate for. PC's are increasingly being packed into smaller and smaller spaces (think slimline laptops, media center PCs etc) and so it's best to spread the heat around the motherboard rather than have it all coming from the CPU.
It's called Common Serial Interconnect or CSI and is slated for Q4'07 last I heard.The current FSB design isn't really holding Conroe back but yes I agree that it's not good for scalability or going forward. The quad core chips especially are pushing the FSB to its limits. Some of Intel's server boards have dual FSBs, one for each CPU socket, but obviously this is just a temporary workaround.
An on-die memory controller won't help the Core architecture much at all. Precaching tends to erode the benefits of it. It probably could improve performance by a few percent but at the cost of die space that would be better spent on L2 cache or more branch prediction logic. Plus there's the old bad boy called "heat" which Intel has a new found hate for. PC's are increasingly being packed into smaller and smaller spaces (think slimline laptops, media center PCs etc) and so it's best to spread the heat around the motherboard rather than have it all coming from the CPU.
Last edited:
[HoL]Cobra said:Oh deary deary me, you must really have something wrong, like forgetting to plug it in, if your old P4 was quicker than your C2D
Shakes head at some comments on here, worries me sometimes.
PS, If there cheap enough, i'll buy the lot.
Who are you folding for? What's your daily output from your C2D system? I have no idea who you are or what team you fold for but I can assure you that I take my F@H very seriously indeed and I'm building up Prescott and Presler Core P4 systems now to re-run these WU's so I can see if they have been optimised for the older processors. If they are quicker, then I'll switch back, or if AMD X2's are quicker, I'll switch to them.
As for taking them all, you have no access to MM so there's not much chance of that now, is there?
Is this an argument for C2D FSB overclocking? I mean that 1066MHZ is based on a 266MHz clock, overclock that to 450MHz as many boards seem to do, maybe drop the CPU multiplier a little and you've got an 1800MHz bus or or 14.4GB/s (assuming a 69% increase on that 8.5GB/s figure).MrLOL said:
Associate
- Joined
- 23 Oct 2003
- Posts
- 185
- Location
- Coventry, UK
WJA96 said:
No mention of you doing folding at home on your post, or am I blind?
If FoH is not optimised to run adequetly on a C2D, I would suggest you mail of to the people responsible, and see or ask if their working on optimizing their code for them. To say in your post that a C2D is slower than a P4 is laughable, again, no mention of FoH so think on your reply.
Edit
Just read your other posts, but tbh, dont read the names of people who post, just the post themselves, and my post was from your post that I quoted.
WJA96 said:
If that is all you buy computers for, then good luck to you.
I on the other hand, buy to have the best for what I use it for, which in this case is an E6600. Fast, reliable and fairly cheap, and does the job excelently for myself, as your's does for you.
I do not Fold for no one, and if I did, I would not beat about the bush trying to score higher than everyone else, but just let it work for the team, as I did with SaH years ago.
I did on the other hand try a FoH WU the other day to try to push my computer to it's limits, for a stress test to see if notepad loaded within 3 seconds (it did btw) to compare to the OP's problem. It has since been uninstalled.
As to MM, indeed I do have access, just not on my account, as i'm more of a reader than a poster, but I only have to nip accross the road to have a browse thank you very much..
Last edited:
I've been watching this thread for a while now. It's interesting to see how firmly some people believe the C2D cannot be beaten in anything. I'd pretty much assumed that too until I started seeing these kinds of reports. While it's clearly the best CPU for the average user it doesn't surprise me that it takes a bigger hit than the dual core A64. I think, as several people have pointed out, that the on-die memory controller is the key here. It's important to have an open mind and to consider what's right for your own needs.
It's actually got me thinking about my next upgrade. C2D is absolutely the best way forward at the moment. But it occured to me that I may be able to pick up some older used 2** series Opterons and take a look around for one of the rare-ish s940 boards with PCI-E and SATA. I can already see a place selling new Opty 265's for about £150 (not sure if that includes VAT as it's a little unclear). A couple of those would give a reasonably cheap way to getting a four core setup depending on the price of a used motherboard and RAM.
I actually think the way forward is multiple cores rather than fewer faster cores. As a software developer I can see a lot of ways to improve efficiency in this way. I'll probably start a separate thread on it at some point.
It's actually got me thinking about my next upgrade. C2D is absolutely the best way forward at the moment. But it occured to me that I may be able to pick up some older used 2** series Opterons and take a look around for one of the rare-ish s940 boards with PCI-E and SATA. I can already see a place selling new Opty 265's for about £150 (not sure if that includes VAT as it's a little unclear). A couple of those would give a reasonably cheap way to getting a four core setup depending on the price of a used motherboard and RAM.
I actually think the way forward is multiple cores rather than fewer faster cores. As a software developer I can see a lot of ways to improve efficiency in this way. I'll probably start a separate thread on it at some point.
Associate
- Joined
- 28 May 2004
- Posts
- 143
- Location
- Cardiff
FatRakoon said:
Don't know if this has been addressed yet (long thread...) but the timing is out and will result in poorer performance.
You should be able to run 4-4-4-12 timing on that memory (I'm assuming its Geil PC6400 now). The fact its reporting it as PC5300 in the SPD tab is normal. I believe this may be for compatibility reasons because the memory requires higher than JEDEC standard voltage to operate at its rated speed. It needs 2.1v. In other words it probably wouldn't POST in most motherboards if the SPD was set for it to operate at 800mhz and 4-4-4-12 timings.
Last edited:
Associate
- Joined
- 28 May 2004
- Posts
- 143
- Location
- Cardiff
I don't have the same motherboard, I have an Abit AB9 Pro, but I do have that memory.
In my case, I run the memory at 840mhz (FSB 420mhz) with 4-4-4-12 timing @ 2.1v. The Abit doesn't seem to like very high FSB's unless its volt modded, but the memory could run a lot faster I suspect, with slightly slacker timing. With the board I use, you have to set the memory command rate at 2T to run those sort of FSB's.
I would therefore say if you are having problems changing the timing, make sure the command rate is 2T if you have that option. And make sure the memory voltage is correct i.e. 2.1v (or +0.3v if thats how your motherboard shows it).
In my case, I run the memory at 840mhz (FSB 420mhz) with 4-4-4-12 timing @ 2.1v. The Abit doesn't seem to like very high FSB's unless its volt modded, but the memory could run a lot faster I suspect, with slightly slacker timing. With the board I use, you have to set the memory command rate at 2T to run those sort of FSB's.
I would therefore say if you are having problems changing the timing, make sure the command rate is 2T if you have that option. And make sure the memory voltage is correct i.e. 2.1v (or +0.3v if thats how your motherboard shows it).
Yes, +0.3v ... 2T? I dont know to be honest, Ill check that out.
The timings is extremely slack, but then this is obviously why Im being allowed such a high FSB perhaps?
Kind of defeats the object of the whole idea of clocking it realyl if this is the kind of performance I am getting?
Corsair @ 400 & 4-4-4-12 = 18 Seconds 1MB SuperPI
Geil @ 430 & 5-7-7-20 = 21 Seconds 1MB SuperPI
Rather be on 400FSB if thats what I can expect.
Will try more out when it nexts restarts
The timings is extremely slack, but then this is obviously why Im being allowed such a high FSB perhaps?
Kind of defeats the object of the whole idea of clocking it realyl if this is the kind of performance I am getting?
Corsair @ 400 & 4-4-4-12 = 18 Seconds 1MB SuperPI
Geil @ 430 & 5-7-7-20 = 21 Seconds 1MB SuperPI
Rather be on 400FSB if thats what I can expect.
Will try more out when it nexts restarts
FatRakoon said:
I don't think that's possible. The RAM timings might make a few decimals of a second difference, but not 3 seconds, especially when you've increased the speed of the processor by almost 10%.
Have you checked the CPU isn't throttling because it's overheating? YOur results are extremely unusual.
Associate
- Joined
- 28 May 2004
- Posts
- 143
- Location
- Cardiff
I have some experience in this error due to unintentionally ballsing up my RAM timing 
I once accidentally set my memory timing at 4-12-12-12 (instead of 4-4-4-12!!)
Running my E6300 @ 3 ghz-ish the difference between the two timings in SuperPi 1mb was maybe a second or so.
However on the 32mb calculation the difference was huge, with the 4-12-12-12 timings being well over 2 minutes slower than the correct ones, at the same CPU speed.
The system also felt noticeably more sluggish (just contributed to a hunch I had that something wasn't quite right) but surprisingly, was completely stable even though the settings were way out.
I once accidentally set my memory timing at 4-12-12-12 (instead of 4-4-4-12!!)
Running my E6300 @ 3 ghz-ish the difference between the two timings in SuperPi 1mb was maybe a second or so.
However on the 32mb calculation the difference was huge, with the 4-12-12-12 timings being well over 2 minutes slower than the correct ones, at the same CPU speed.
The system also felt noticeably more sluggish (just contributed to a hunch I had that something wasn't quite right) but surprisingly, was completely stable even though the settings were way out.
@FatRakoon - is your system prioritised for applications or background processes?
OH?
I cant get into the BIOS?
Its thrown the cursor into the middle of the screen ( Slightly up and to the left ) instead of going into the BIOS
It did this before when I had a Maxtor 40GB attached to the PCI Card, but its been fine with just DVDRW Drives?
I took the card out and it still wont go in, so I took all the cards out and cleared the CMOS and its still not going in???
Any clues?
I cant get into the BIOS?
Its thrown the cursor into the middle of the screen ( Slightly up and to the left ) instead of going into the BIOS
It did this before when I had a Maxtor 40GB attached to the PCI Card, but its been fine with just DVDRW Drives?
I took the card out and it still wont go in, so I took all the cards out and cleared the CMOS and its still not going in???
Any clues?
Soldato
- Joined
- 3 Nov 2004
- Posts
- 9,871
- Location
- UK
Don't know the board, but does it change chipset strap latencies over 400Mhz? That'd slow things down, aside from memory timings.