| Thread Tools |
16th June 2005, 12:42 | #11 |
Posts: n/a
| Hi guys, reviewer here. Sorry that you were not altogether happy with my review! So I'll explain my reasoning for a couple of things. If you look at the results for Doom 3, you will see that even overclocking the 3700 from 2200 to 3263 has an almost negligable impact on the benchmark - less than 1 fps. The difference between the 3500s and the San Diego were so small, I couldn't believe it, and re-ran the benchmark several times. This would also imply that the 1MB cache is having no benefit at all with Doom3, as if the minimum fps had risen, so would the average fps! As you can see the results in Far Cry, and especially HL2 were a different story, and the performance increase offered by the 1MB L2 cache was very worthwhile. On to using the stock HSF! There were two reasons for this. 1) Performance out of the box. 2) More importantly, to show how well this particular example overclocked in the worst possible case. This was then balanced by putting the CPU under Phase Change! It's pretty reasonable to expect the reader to draw the conclusion from these extremes of cooling: Your overclock may be somewhere in the range of 2700-3300 depending what cooling you are using. Its probably also useful to note that this particular "early" 0515SPW example seems to be one of the best, and that later examples are not doing as well. Hope this clarifies things - any questions just ask |
16th June 2005, 13:06 | #12 |
Posts: n/a
| That's nice to clear up thanks. Since I play and mod mostly Doom3 (and the relevant engine), I think I'll keep my 3500 until I can afford a 4000 San Diego core. |
16th June 2005, 13:22 | #13 | |
Posts: n/a
| haven't read it completely (no time), but euh, comparison at 2 different voltages ? Quote:
| |
16th June 2005, 14:47 | #14 | |
Madshrimp Join Date: May 2002 Location: 7090/Belgium
Posts: 79,021
| Quote:
did you run Doom3 at 640x480 low detail? can't see it in the review at first glance; if you do a CPU review, you need to at least include the runs at lower resolution, putting the load the CPU rather then the GPU take a look at one of our reviewer's report of the previous core release here to see what I mean: http://www.madshrimps.be/?action=get...&articID=23 0
__________________ | |
17th June 2005, 10:36 | #15 |
Posts: n/a
| Cheers for the welcome jmke! More explanations needed huh? I'll have to do a better job next time! 1) The reason for the voltage difference (which I should've explained) was that the 3500 Newcastle would handle 3.7v, the Winchester would not boot at 3.7v, so it's best was at 3.6 . I've made a small edit just to clear that point up. Cheers! :excellent 2) Yes I've soon those results The thing is here, I wanted to try and keep the test relevant, which means running a resolution that most people use. Looking at the results Faiakes was able to deduce that this wouldn't be a worthwhile upgrade for him if he had been a HL2 player that could have been different. Doing a test at 640X480 is artificial - I hope I did enough artificial benchmarking.. and looking at your result, even at that resolution the benefit of the higher clock speed and extra L2 cache was still less than 3%... Something I might try in the future (my DVD player isn't working ATM ) is running a Far Cry benchmark at stock and overclocked, and then changing the rendering mode from D3D to Open GL, and see what effect this has on any gain from overclcoking - unless anyone has already seen any results of course! Thanks for the suggestions/comments, they're always welcome |
17th June 2005, 10:42 | #16 |
Posts: n/a
| Far Cry, mmmmmmm Actually I did play HL2 but is crap at modding and the whole Valve system is crap, so now it gathering dust on my shelf. Anyone interested in buying it for 20 Euro? |
17th June 2005, 10:52 | #17 |
Madshrimp Join Date: May 2002 Location: 7090/Belgium
Posts: 79,021
| I have bought the Silver HL2 package from the web, very nice investment, replayed HL1 in high detail and really enjoying CS:Source, playing against bots >>>> online server with random people.
__________________ |
Thread Tools | |
| |