• 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.

Grumpy old man

Associate
Joined
21 May 2007
Posts
1,464
Just looked back at my last ten or so posts, and they mainly seem to be.........


ATI are pants because of a b and c
and
Nvidia are pants because of x y and z


LOL


I think I might just be a bit frustrated at the slow progress in PC's these last 2-3 years......just a wee bit anyway.

When was Moore's law repealed?



I mean between gfx cards that only offer 150-170% the perf of 2 year old models, and CPU manufacturers going from 2 to 4 to 8 to 16 (soon) cores while 99% of software still uses only one core......it's just not as much fun as it has been in previous years.


Old and grumpy of N.Ireland




PS: need some positivity pills....NURSE!
 
I think your frustration lies in you just want something that is a noticeable improvement over what you are currently using and not feeling short changed by your decision.

The problem is this time round there is not really a decisive answer per se as compared to previous generations so it makes decision making just that little bit harder.

It always easy to look at benchmark put up by various reviewers but different peeps have different needs. From reading your posts it seems you like the most AA you can apply without affecting too much of the cards performance.




Both the 4870 and 280 users make compelling arguments for the cards they use. I am one of those peeps I like to test for myself and not rely on benchmarks or people who just used the one card and not the other and then start quoting benches for the reasons they chose their card. There is nothing wrong with what they are doing but I can't help that if I followed their advice and was dissappointed in the choice then that could lead to an conversation I rather not have.

My advice to you ( and it is simply that) is to wait for the 1gb 4870 and that will offer the most potential for the amount of AA you like. It's not guaranteed and I am sure a 512mb 4870 would suffice but you never know. If you find you still can't go on with it then return it to ocuk and get a gtx 280 or something else which may come out in the mean time if it is better. Only through your own testing will you truly know which card is for you.

Funnily enough I have gone back to the 4870 using the 8.7 official drivers I think it is. It's the one that was recently released. I got to admit the 4870 seems like a different card using these drivers. It seems to be better performance wise and I am very very impressed. It could be placebo effect but when I was playing Grid earlier today the low minimum framerates that bothered me before using the previous drivers seems to have been increased therefore making Gridding much more enjoyable with this card than previously.

It still stutters in menus and it still has that split second freeze for me but certainly the ati driver department is earning their moneys worth.

Old and grumpy and skint of S.Wales:)
 
I'd say you pretty much have that one surrounded :D



But I AM miffed that there are no actual CPU upgrades (beyond speed-per-clock tweaks like those Nahalem brings) coming up......at least not for MOST uses....obviously 8,16,32 cores are great for hefty rendering or movie ripping, but for most apps and games you're just gonna be buying 1 CPU and 7,15,31 spare ones ;)

As has been mentioned, proper multithreaded programming is far from easy, which means either Intel and AMD have to either figure out some method to make all the cores transparently act like one uberprocessor......or start seriously thinking about new tech to get past the current speed limitations. I sometimes wonder if the multicores are where they're gonna stay, marketing increasingly useless chips[1]....or whether they are just to distract us and satisfy the upgrade obsessives while they work on something genuinely new.


Anyway, defo going to wait for 1GB 4870 and try one, as you say, if it stinks, it can go home again and I can try a 280. Also tempted-ish by a 4870x2 if it's 2GB.....but don't trust this xfire stuff, and only have 780W to play with (currently peaks just over 400 with the GTX, so it should JUST be OK).

Still grumpy though :D



[1]Look, I know there are benefits when running loads of processes.....but apart from Bootup, when do most of us do that? (mate I just built the box for uses his 4 cores to run 3 copies of WOW, and still have some capacity for using other stuff). @2GHz, a C2D is not 2x a P4 and a C2Q is not 2x a C2D...that's what I'm driving at.
 
PS: need some positivity pills....NURSE!

Burn your PC, shoot your mobile and live on a desert island! Much better in the long run... Seriously though, I hear you, I find new tech disappoints more as it progresses, whether that's to do with our expectations or not I don't know. I've been in IT for 21 years, and it's not as much fun as it used to be!! My new rig (in sig, soon to be 4870'd) is great, but hasn't blown me away as much as previous incarnations...

Even Grumpier of Worcestershire
 
When was Moore's law repealed?

Contrary to popular opinion, Moore's law makes absolutely no statement on the real-world performance of any piece of silicon.

It simply states that the number of transistors which can be placed on a piece of silicon, for a given cost, will double every two years. That's it. CPUs have been keeping track with Moore's law (and GPUs have been exceeding it since their genesis), since it was introduced in 1965. How much longer it can be maintained is anyones's guess...
 
Contrary to popular opinion, Moore's law makes absolutely no statement on the real-world performance of any piece of silicon.

It simply states that the number of transistors which can be placed on a piece of silicon, for a given cost, will double every two years. That's it. CPUs have been keeping track with Moore's law (and GPUs have been exceeding it since their genesis), since it was introduced in 1965. How much longer it can be maintained is anyones's guess...


Was it not 18 months?

Surely a C2D and C2Q have the same component density? Hence a Quad is just 2 C2D's under one cover?

Anyway, in 10 years will we have 128 core chips @ 3GHz, with 127 cores doing a whole pile of sod all 99.9% of the time?

I see your point about GPU's but it doesn't seem to hold true (not that I know much about this). a 280 is 2 years younger than an 8800GTX, so it SHOULD have 2x transistor density, it also has a die which is considerably larger, yet it barely manages 150% the ultimate performance.

Time for quantum computing or something methinks;)



My CPU's......

8MHz,233MHz,533MHz,1200MHz,2000MHz,2800MHz(tortured to 3.2G),2400MHz(tortured up to 3.4G), and in another two years..... another 2400MHz....but with 8 or 16 of them....you see what's buggin me :D
 
Was it not 18 months?

Surely a C2D and C2Q have the same component density? Hence a Quad is just 2 C2D's under one cover?

Anyway, in 10 years will we have 128 core chips @ 3GHz, with 127 cores doing a whole pile of sod all 99.9% of the time?

I see your point about GPU's but it doesn't seem to hold true (not that I know much about this). a 280 is 2 years younger than an 8800GTX, so it SHOULD have 2x transistor density, it also has a die which is considerably larger, yet it barely manages 150% the ultimate performance.

Time for quantum computing or something methinks;)



My CPU's......

8MHz,233MHz,533MHz,1200MHz,2000MHz,2800MHz(tortured to 3.2G),2400MHz(tortured up to 3.4G), and in another two years..... another 2400MHz....but with 8 or 16 of them....you see what's buggin me :D


Nope - every two years. Check out the wikipedia page if you like, there are plenty of referenced sources there. Also, Moore's law is not about "transistor density", it's about the number of transistors on a piece of silicon. GT200 has twice the number of transistors as G80.

As for quantum computing - that's going to be a highly specialised tool which will make the breaking of existing encryption algorithms a trivial process, will allow the formation of truely unbreakable encryption (through quantum entanglement), and will allow the rapid simulation of molecular models (among other things).

...what it won't do is run your games any faster. You will never use a quantum computer for 'everyday' calculations, as it would not be able to compete with a traditional transistor based computer.



As for your overall point, well we're rapidly approaching the limits of the technology. We are starting to run up against problems at atomic lengthscales (quantum tunneling etc), and the designs which are implemented have very much matured. Like all new technologies (think of aviation as an example maybe?) we've had the 'boom' phase. Yes, there are significant improvements still to be made, but the real-world performance improvements from these will continue to slow down.

I do think, however, that GPUs and other naturally parallel devices have a lot more 'room' for improvement than CPUs (serial devices). The main problem with massively parallel computing will continue to be power requirements (and as a consequence, heat production and removal).
 
Last edited:
Back
Top Bottom