![]() ![]() They never recovered from that smackdown. So for about 6 months they could smack around the Core 2 Quad and charge $250 for the x4 965, and then they got destroyed by the $200 i5 750. So, the Phenom II is as close as AMD ever got post-Conroe to being competitive with Intel, and that's back when they had mountains of cash to throw around. And when they were tested, the Phenom II nearly matched the Core 2 Quad 45nm, but took 30w more power to make it happen:īut then 6 months later Intel released Lynnfield, which had 95% of the performance and features of the i7 920 without the extra 30w of overkill platform I/O power: Both companies had a promising architecture shrunk from 65nm and tweaked. The last time AMD had the same process node as Intel and similar performance was the 45nm Phenom II. They'd probably be starting out at around Sandy Bridge in terms of performance per clock.īut can they do it, and be power-competitive? ![]() And it can get very high utilization of those units thanks to the high bandwidth, low-latency cache architecture, plus the hyperthreading to feed the unused units when the first thread can't.ĪMD can certainly build a core with similar integer IPC to Intel, but I doubt it. Only two out of three of those things actually happened, so we were blessed with the trainwreck that was Bulldozer at 3.5-4 GHz.īy comparison, Haswell can execute up to 4 integer instructions per clock. So the architecture was designed to get LESS done per clock, add slightly more cores, and clock the thing at 5+ GHz at a reasonable power level (to get higher performance than the Phenom II using less resources). Shared decoders reduces fully-loaded scaling by a further 10% versus Phenom II.Īnd high-latency big caches are designed for running at high clock speeds, but cost you some throughput versus a low-latency cache. ~50% less FPU throughput per-core (not as bad as the above) The key differences form it's predecessor the Phenom II were:ģ3% less integer execution units per core (this hurts the most). Right, when AMD announced the Bulldozer architecture people were scratching their heads in confusion. They need a radically redesigned and optimized architecture, shrunk to a smaller process (Intel is on 14nm now for crying out loud! The FX series is 32nm, that SEVERAL GENERATIONS behind.) The rumors that AMD is working on a new architecture with one floating point unit per core is probably good, but that alone won't make up very much in the performance deficit to intel. So, unless you are running code heavy in 256bit floating point math, this will have no impact what so ever. Sure, the two cores in a module share an FPU, but it is a 256bit FPU, and it has the ability to split itself into two dedicated 128bit FPU's, one dedicated to each integer core. However, I think this particular aspect of it has been overblown. Bulldozer and on has been a total failure. Just look at any single threaded benchmark, and you'll find AMD's FX architecture performing as little as HALF the performance, clock for clock, compared to Sandy Bridge and newer Intel chips. Even if you do a single threaded comparison (where this should have no impact at all, because you are only loading one of them, so there is no sharing), the AMD architecture is WAAAAY behind. It contributes to the problem, but it is a minor part to it. Watercooling blocks on cpu gpu and northbridge stable OC'd at 32C.Īll overclocking is completely disabled right now, everything is running at stock speeds. Have also adjusted the power plan to default thinking this may be a compatibility problem with the OS. How do i get my cpu to register as an 8 core cpu opposed to a quad core? i disabled all my overclocking that i had set on windows 7. Only reason i upgraded to 10 was because of dx12 support and rumors that it will increase performance spreading operations to all cores of my cpu and better utilizing a multi-gpu system. Ive had to deactivate my crossfire on several games in order for it to run, which it does in the slowest way possible. It also wont load allot of things on my computer. From boot times to opening any system process sound/gpu options/disc cleanup/HDD properties/task manager/etc. Everything is taking a month to load, much longer than windows 7 did in every way. Also its showing in both the task manager and in cpuz, as 4 cores opposed to 8 and its showing a lower l1 and l2 cashe than what the cpu has. Ive been having nothing but problems with win 10, its showing my cpu is running between 2.2ghz-2.81 and keeps changing actively. Just updated to windows 10 圆4 pro from win 7 圆4 ultimate-n. ![]()
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |