CPU cache debate- need to get some support

MumboJumbo · 9 Mar 2008 at 13:56

Bascialy this guy refuses to beleive that cache on cpu's doesn't make a huge difference in gaming.He has claimed that the tests done on this very forum are wrong. He claims on one of them that differnt FBS's were used and therefore this caused the lower cached cpu to do just as well as the higher cached one.

Can you lot please put an end to this?

He also constantly argues with me that any voltage over 1.5v and especialy 1.6v will kill your cpu in weeks via ion migration. He is doing my nut in lol.

gurusan · 9 Mar 2008 at 14:01

It depends which game you're talking about. For example UT3 seems much more cache dependent than others....but for the most part though from what I've seen cache has a very minor impact on gaming performance if any. (yes at same FSB and clockspeeds).

As for voltages over 1.5v, theoretically yes they will die earlier....but in a matter of weeks? come on.

I know people who have run 1.6v through their 65nm C2D chips for almost 2 years now! If what he says were true then C2D chips would be getting zapped left and right....and I can't even think of 1 instance I have heard of where a C2D chip has died on this forum or any other.

Cyanide · 9 Mar 2008 at 14:03

Well it seems to make a difference for me, my scores in benchies and real games are pitiful compared to the same spec system with a larger cached CPU.

gurusan · 9 Mar 2008 at 14:06

Which games? Benchies tend to show a difference but as far as real world gaming goes I've I've only noticed a difference in UT3 (and that's only with the FPS counter on)

Cyanide · 9 Mar 2008 at 14:08

gurusan said:
Which games? Benchies tend to show a difference but as far as real world gaming goes I've I've only noticed a difference in UT3 (and that's only with the FPS counter on)

Crysis, Oblivion, Forged Alliance... to name a few.

I just don't think the CPU can handle having 2 GPU's without a larger cache, Crossfire is very CPU hungry afterall.

Richie P · 9 Mar 2008 at 14:09

Depends on what you both classify a 'huge' difference.

The E1200 Celeron is severely crippled by the lack of cache, but when you put a 2MB C2D versus a 4MB, it depends on the application. From what i remember, it's a matter of a few percent.

1.5-1.6v could be harmful to a CPU, but again it's dependant on temperatures.

Sounds like you're both kind of right.

gurusan · 9 Mar 2008 at 14:10

With 3870s in crossfire I would have thought detecting any CPU speed or cache impacts to be very difficult.

Cyanide · 9 Mar 2008 at 14:12

gurusan said:
With 3870s in crossfire I would have thought detecting any CPU speed or cache impacts to be very difficult.

You'd be suprised, can't get much past 32-35fps in Crysis at the moment which is a lot lower than a similar system with a larger cache.

Forged Alliance grinds to a halt when playing skirmish games and the outside parts of Oblivion run rather choppy at times too.

I know my res. probably isn't helping but a pair of 3870's should be doing a lot better than they are.

gurusan · 9 Mar 2008 at 14:14

Well I cant see there being that much of a difference, I think something may be wrong with your driver setup or settings? In the Crysis bench I get 31fps at 3.3ghz and with my 3850 512mb at stock.

However if cache makes a difference I would expect to see it in Forged Alliance....

Cyanide · 9 Mar 2008 at 14:19

gurusan said:
Well I cant see there being that much of a difference, I think something may be wrong with your driver setup or settings? In the Crysis bench I get 31fps at 3.3ghz and with my 3850 512mb at stock.

After installing the 8.3's it's currently running at 29fps so there's an issue at the moment. On 8.2's the most I could get from it was 38fps when it was clocked to within an inch of it's life. Which means I got 7fps more with an overclocked Crossfire setup than a stock 3850

gurusan · 9 Mar 2008 at 14:21

yeah well I think there is some other factor at play here....

Tooks · 9 Mar 2008 at 14:26

Crysis is the only game I've run recently on both a 1mb, 2mb, 4mb and 6mb cache CPU, with the other hardware the same.

Using the built in CPU benchmark app, my experience is marginal differences on the max fps, but significant differences on min fps to the order of 40% faster on the 2mb/4mb/6mb cache chips.

Probably wouldn't be something you'd notice on games that are running faster than 60fps to be fair, but noticeable on Crysis.

Talrinys · 9 Mar 2008 at 14:49

Like someone said, when cache gets really low, 512KB currently, you'll see quite a big hit. Same with 1mb but moving from 2mb to 4mb to 6mb won't result in huge gains in all games. They'll benefit some, titles have already been mentioned, but i believe the main gains are seen during encoding/media.

tetsuo · 9 Mar 2008 at 15:01

Talrinys said:
Like someone said, when cache gets really low, 512KB currently, you'll see quite a big hit. Same with 1mb but moving from 2mb to 4mb to 6mb won't result in huge gains in all games. They'll benefit some, titles have already been mentioned, but i believe the main gains are seen during encoding/media.

Interesting, i'm trying to decide between a 9300 vs 9450 non OC for gaming at the moment. Guess its just best to wait and see reviews etc..

Talrinys · 9 Mar 2008 at 15:41

tetsuo said:
Interesting, i'm trying to decide between a 9300 vs 9450 non OC for gaming at the moment. Guess its just best to wait and see reviews etc..

Yeah like i said in the other threads we'll just have to wait and see but AFAIK the 9300 has 3mb cache per dualcore, 6mb total right? That sounds like enough, i mean my E6400 has 2mb and that isn't crippled by it last i checked.

Geffen · 9 Mar 2008 at 17:14

On average doubling the cache size seems to add about a 10 to 15% increase in performance in games. So it does offer an improvement but personally I don't think the increase is normally worth the extra cost.

Moving from an E2XXX to E4XXX you get 15% increase in performance for 35% increase in cost and again going from a E4XXX to E6XXX you get about 15% increase in performance for 30% increase in cost. If you're in to games then your generally best to buy a E2XXX processor overclock it and put the money you've saved not buying a E6XXX or E8XXX towards a better graphics card, which will have more impact on the FPS. There are a few exceptions to this such as Supreme Commander when you would be better spending more on the CPU.

gurusan · 9 Mar 2008 at 17:14

Where are you getting all these numbers? :confused:

Amnesia · 9 Mar 2008 at 17:20

to test if it makes a difference you would need to test games which are bottlenecked by cpu or atleast very cpu dependant...crysis isnt one of those games.

most online fps games will be heavily dependant on cpu power.....find any server with lots of players (20+) make a demo with graphics set to very low ...then do a timedemo.

Geffen · 9 Mar 2008 at 17:25

The review on Toms Hardware has quite a good comparison of impact of Cache see
http://www.tomshardware.co.uk/processor-cache-size,review-29707.html but almost all review site compare different version of Intel chip against each other.

Jokester · 9 Mar 2008 at 17:29

So it looks like going from 1MB to 4MB cache models, at most seems to give a 10-15% boost, but in most things makes next to no difference.