SiGe, sooner or later it will come
it's a well-known fact
Printable View
SiGe, sooner or later it will come
it's a well-known fact
Yeah, my theory is that it will be implimented in the next 65nm rev. That since AMD's inital batch of processors were mainly alotted for OEM's, they decided to pump out the initial batch without SiGe, while improving their 65nm tech. The next batch released for retail would have improved clocking. I guess it all depends on how much time and work is required to mix in germanium to the mix without having to go back to the drawing board.Quote:
Originally Posted by MAS
Anyone kind enough to link me to the theoretical performance for the K8L?
where are you from? :DQuote:
Originally Posted by LiQU!D
see http://virtualexperience.amd.com/ind...re&co=quadcore , it will help u
I went there and couldn't find anything. :S
install Java, k-lite codec pack, OC you inet connection and u'll hear uncle Randy Allen speaking
lol what a load of marketing bullQuote:
Originally Posted by MAS
so why the hell do you assume you need an expensive kit for intel and a cheap ddr667 kit for amd? intel has dividers too...Quote:
Originally Posted by SkunK
I would prefer MAS just relayed that information to me. ;)Quote:
Originally Posted by brentpresley
The lowest divider on the chipsets that would bring the E6300 above 3Ghz is 1:1. The guy says 3.2Ghz is easy with an E6300. The E6300 has a multi of 7. 3200mhz / 7 = 458mhz FSB. 458mhz FSB means 916mhz on the memory at 1:1. There's no way a cheap 667mhz DDR2 kit will do 916mhz for long enough to prove it. Now are you gonna say that with extremely lose timings it's gonna make it happen? Maybe it would. But then the alleged 20%-40% performance gap over the AM2 would be reduced considerably.Quote:
Originally Posted by cky2k6
Not to mention that the 100$ mobo isn't gonna help reach 3.2Ghz.
The AM2 X2 3800 has a base frequency of 200mhz (with a 10x multi). If you raise the frequency to 300mhz and use a 1:1 divider on the memory, it's gonna run at 600mhz and, if you use DDR2 667, is gonna be UNDERclocked (thus buying cheap mem).
But maybe you know more than I do. After all, you have a sig saying that you have a 170 Opteron at 2.8Ghz and mem at 260mhz. For me, running a s939 cpu at 2.8Ghz means running the memory at either 280mhz or, the next best thing, 255mhz. Did you invent a new divider?
You better reread what I said. I read the Locked thread . So I remembered you saying NO good chipsets exist for Intel. So to make you feel good about yourself. I said we can reduce the 3.2ghz . To O/C to the same as an AMD O/C which is stable at or below 2.8ghz. So ineffect I was saying clock for clock C2D is 20-40% faster than X2.Quote:
Originally Posted by SkunK
If we used a 3.2ghz C2d would be 40-60% faster. Sorry you miss read that. As I wasn't real clear on it. Because I wanted to use the cheap DDR2 so as to keep the prices equal other than M/B. C2D does not need expensive ram to hit 2.6-2.8 ghz. So for $100 more for the C2D setup is cheap price/performance wise. At new egg you can pick up DDR2 OCZ PC2-6400 2x1024 kits for $207.99 using 1:1 @400fsb with 965 chipset is very doable.= 2.8ghz.
100$ more as in a 100$ vs 200$, 200$ vs 300$, 300$ vs 400$ video card? If someone buys a computer to use Word and do some browsing, the 100$ more will most likely be unjustified. If someone buys a computer to play games (and is still price conscious), he will probably want to have a better video card for that 100$ more as it would be a more interesting investment.Quote:
Originally Posted by Turtle 1
I admit, though, that I almost bought a C2D system during the Hollidays. But the only mobo that would interest me is the AW9D-Max and it seems plagued with BIOS problems not counting that it doesn't reach that high FSB which wouldn't help a E6300. I would have loved to compare it to my Opty in x86_64 Linux, though.
So than the new low end chips intel will be releasing that run at DDR2/ 800FSB with high multiplier should be very interesting to people such as yourself.
With a release price of about $150 and a second Qt, price of $115 should be very hot selling chip in the budget market. I would never build a new system I was going to O/C and use anything but a top end gpu. If I wanted a wed browser any $300 PC would do the job well.
Maybe. But if you're talking about the E4300, first, it's not here and second, it still has to prove that it will be a good overclocker to interest me. So far, I've read threads that say the E4300 won't clock so well and is crippled in features.Quote:
Originally Posted by Turtle 1
Actually you forgot too much...Quote:
Originally Posted by Shintai
What AMD says will have improved/added:
0. Native quad core
1. Hypertransport up to 5.2GT/s
2. Better coherency
3. Private L2, shared L3 cache that scales up.
4. Separate power planes and pstates for north bridge and CPU
5. 128b FPUs - see 14,15
6. 48b virtual/physical addressing and 1GB pages
7. Support for DDR2, eventually DDR3
8. Support for FBD1 and 2 eventually
9. I/O virtualization and nested page tables
10. Memory mirroring, data poisoning, HT retry protocol support
11. 32B instead of 16B ifetch
12. Indirect branch predictors
13. OOO load execution - similar to memory disambiguation
14. 2x 128b SSE units
15. 2x 128b SSE LDs/cycle
16. Several new instructions
first off, maybe intel boards dont have a 400 divider, but nvidia does... theres more than one chipset manufacturer for intel. second, you have an opty 170, thats not a cheap cpu by any means, so why are you comparing to intel's lowest end conroe? the opty 170, up until a few months ago cost more than 350, so its only fair to compare it to a 6600. and even amd says to use ddr2 800 ram if you want good performance from the cpu. as for my ram, its not at 260, thats it highest overclock...
Show me a divider lower than 1:1.Quote:
Originally Posted by cky2k6
Fair for whom? Find me a 180$ E6600 right now and I'm gonna buy not one but two of them.Quote:
Originally Posted by cky2k6
Thanks for being honest after the fact. The usual MO of Intel overclockers. :slap:Quote:
Originally Posted by cky2k6
If that means 256 bits cache line size then that's not a good thing either.Quote:
Originally Posted by Fred_Pohl
brilliant comeback against someone whose last intel was a 1.3 p4... and again, you bought your 170 most likely for 350, youre complaining about how expensive everything is on the intel side, but yet you yourself bought a cpu more expensive than an e6600 was at launch... wait to be a hypocrit...Quote:
Originally Posted by SkunK
Well, you could have bought AMD stuff as far back as the first paycheck you ever had in your life. It doesn't change the fact that your MO (latin: modus operandi, behavioural pattern, way to conduct yourself, etc.) is the same as the Intel freaks.Quote:
Originally Posted by cky2k6
Also, against you, I don't need to make any comeback since I'm always a step ahead of you (I'll let the readers make their own conclusions about that).
And, finally, read this . Better yet. Try to read anything and, most importantly, grasp it (or at least try) before you make any more comments.
i guess since you like personal insults, your reading comprehension fails you. note how i said opteron 170s cost 350 before a few months ago. i assume you bought your system more further back then when the price drop happened, so again, you are a hypocrit for :banana::banana::banana::banana::banana:ing about intel's prices, when amd used to charge just as much when it was kicking intel's ass... and again, amd wants you to use ddr2 800 for am2 as well, thats why it released it in may and not earlier... and anyways, if intel is such an expensive choice for you, its your peformance to lose, this a star es thread, not why intel is a crappy choice.
You called me an hypocrit first. So I guess there I had to make a comeback.Quote:
Originally Posted by cky2k6
Read all my posts from the last few days. I always stated that I bought the 170 about a month ago. Never have I hidden that fact.Quote:
Originally Posted by cky2k6
I don't think AMD really wants something from me right now.Quote:
Originally Posted by cky2k6
Yeah, I'm very sad I stopped masturbating in front of single Super PI runs when I saw C2D results.Quote:
Originally Posted by cky2k6
Well yes it is. But who instigated the off topic bull:banana::banana::banana::banana:ting?Quote:
Originally Posted by cky2k6
Excellant Post. Could you go threw the improvements 1-16 and add what each point may do for AMD as far as performance is concerned.Quote:
Originally Posted by SEA
What language? Is modus operandi a restricted expression on these forums?Quote:
Originally Posted by brentpresley
Nothing K8L won't be able to fix, though. :with:Quote:
Originally Posted by brentpresley
Uhmm...why make the big list of something that aint so much? Its like trying to spin it to more that it is.Quote:
Originally Posted by SEA
5+14+15+16: Single cycle SSE and SSSE3. No need to make it 3 points unless Core 2 would just get the same 3 and another one with more SSE ports. And 128b FPUs? What did you smoke. Its eitehr 128bit SSE units or 64/80bit x87.
7+8: We already got DDR2..why is this new? And currently AMD dont seem happy about FB-DIMMs and DDR3 is still quite far away.
2+3: You mean the 66% higher latency L2 cache thats been private always? The L3 cache is what fixes the coherency. I doubt it can add much if any performance enchancement besides that. Just look on 512 vs 1024KB K8s.
1: Thats great..for 4x4 and 2xxx/8xxx Opertons. HT aint a problem on the desktop. Try 800 vs 1000mhz today and bench it.
0: Thats a PR joke. Intels 45nm quadcore would be the real native one with shared L2 and direct L1 links. To call aaything for native is a joke. In that case P4D and K8 aint native dualcores either, only C2D and CD? And even tnen CD dont have direct L1 links and would most likely fail then. I count true quadcores as for how they are licensed.
4: Aint this a socket AM3 feature only?
6: Yes the new 1TB memory limit is quite....useful for us. Does it enhance performance? no.
9: And this is new how?
10: RAS feature, not performance enchancing.
11+12+13: Buffers and memory disambiguation.
So at the end of the day, did I forgot anything? Nopes. If I did I would have been forgetting it on both. I only took the direct performance enchancing things. Hell, you even forgot the wider L1 cache and internal data paths.
We are still back to:
Memory disambiguation
128bit vs 64bit paths
Single cycle SSE
Bigger buffers here and there
L3 cache