AMD Triple-Core Phenom 8600 Benchmarks

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.


All CPU's have errata. The TLB errata is supposedly rarely encountered, but since B3 was on the way, it hurt sales of B2 stepping. Intel has had just as many "faulty" cores, such faults don't change performance since they are rarely encountered in either AMD or Intel CPU's.

Disabling one core rather than throwing away the whole CPU is akin to what both AMD and Intel have done with cache for many years. Where did you think the AMD Semprons and Intel Celerons came from? The reason that Intel has not released triple cores is that they don't have native quad core. True that native quad core isn't as important if not implemented well. Elegance cannot trump peformance. Equivalent Phenoms lose slightly to Q6600 at stock, and does not overclock as well.

That said, the Phenom 9850 looks to be a great budget quad with close to Intel performance for less. It has a better upgrade path than an Intel build today because Deneb is "only" 9 months away at most. With a new Intel build today, you'll need a new motherboard for Nehalem. It all depends on how often you want to upgrade.

With AMD there are two upgrade choices:

If you have one of the bios supporting AM2 690G boards (i.e. ASUS, my wife has one), you can just slap in a Phenom for now, then wait to see how Deneb pans out to decide whether to go AMD's next socket with Deneb, or Intel's next socket with Nehalem. If one is certain they want Deneb in December, then a 780G, 770 or 790 board for a Phenom 9850 is a good deal, because Deneb will support AM2+ as well as AM3 arriving by February 2009. The big difference between AM2+ and AM3 is supposed to be DDR3.

Phenom's not all that bad. People on the Intel side hate to have Prescott brought up (because it's "old") but when they accuse AMD fans of cherrypicking benchies where Phenom either beats a Q6600 or gets to within reasonable framerates or encoding time of a Q6600, they attack what they relied upon during the days of Prescott, when Intel fans (myself included at the time -- I had two Northwoods) pointed out that Netburst was better in encoding then Athlon 64's, and some Pentium D's could hold their own vs. X2's in a very few benchmarks.

And so it goes....

My big question with Phenom B3 is that it didn't come out when I last upgraded, so I was stuck with building 3 Athlon X2's (3800+, 4600+ and 4200+). One board is a 690G that does have a bios for Phenom, one is a 780G and the 690V won't support Phenom anyways. My whole question is will I "need" a quad core in games prior to December? I might "need" a quad core in encoding, but I don't run a business, an extra minute or so is no big deal. I just don't think that B3 will bring enough to warrant the upgrade I've been saying I'd get.

If the promises of Deneb become reality, then Deneb compared to B3 Phenom will be like an Allendale to a Pentium D. It won't be a completely new architecture, but a 45nm process with higher clocks will make it seem like a new architecture. I didn't say Conroe because I don't think the difference will be that great, so I chose the lesser cache C2D's for the example.

I just wish AMD had managed a way to put two Brisbanes together at 65nm and ditched the native quad core. This time, Intel was right to not go native until the process fully supported it.
 
well at least this way here Intel knows what to expect from a True Quad and this is one point where AMD may have an advantage of having a true Quad for a year so they know what to give it to make it go to new levels. It's like when AMD made the Opteron Dual-Core back in 2003 the first multi-core chip and Intel thought they could do the same but decied to just fuse together two Single-Cores to make a Dual-Core.
 
yipsl, The reason bringing up Prescott is, well stupid, is b/c what does it have to do with AMD's tri core? And BTW Ceadermill and Presler started to close the gap and lowered the power consumption to where they were 65w chips.

Either way the main point is that there happen to be too many variations in the benchmarks per site. Some show the tri core getting beaten by the dual core in most areas. Some show the tri core keeping up or being at an equal level.

To tell you the truth its really hard to tell where the tri core stands. Also since its not a 8X50 I would assume that the tri cores are all B2 steppings which could mean OC'ing will be the same as the B2 Phenom X4's.

I still am going to wounder what will happen once AMD's yeilds improve. The less bad chips the less tri cores. This might just be a short lived market.
 
Is it possible to disable one core on a 9000 series in BIOS or from AOD? If so then maybe blackpanther26 would run some CPU benchmarks for us. Then we'ed have an idea of what could be expected from a user built, overclocked system.
 
I hope he puts an extra mobo screw in the right hand side of the mobo and makes sure the HSF is on tight before disabling a core.

Check the EMF also doesn't mess up the telly.

Those harmonics could cause epilepsy ... that's what happened to all of those 5 cylinder Audi drivers ... they went mad.

I live under a high voltage power line ... I suspect TC does too !!!

heh heh.

I still bet the system people will flog heaps of these at the bottom end of the market.

I'd imagine low speed triples would make excellent servers and HTPC's on the new IGP mobos too.



 
OUCH, THAT HURTS !
ouch.jpg


Could we expect to see any price competition from AMD at least? I hope so 😉
 


The gap is the reason to bring up Prescott. The way Intel fans are talking, Phenom is a POS and anyone who gets one is not an enthusiast and really knows squat about PC's. The fact that 9850 closes the gap like Pressler did is something that you guys seem to want to ignore. That shows me that some people have an irrational bias against whichever company is down at the moment (I'd put TC in that category) and others have an irrational bias for Intel or AMD (Thunderman is an example of an irrational bias towards AMD).

That's the only reason to bring up Prescott, Pressler and Cedar Mill. The type of arguments used by Intel fans at the time when Intel was down technologically are the type of arguments those same Intel fans reject when used by AMD fans today. See my point?

I'm not making a technological comparison between triple cores and Prescott. I made a technological comparison between triple cores and Semprons and Celerons in that one part of the CPU (or CPU cache) was disabled to meet a market, regardless of whether a part of that CPU (or CPU cache) was defective to begin with and had to be disabled and remarketed or the whole thing scrapped.




I'm avoiding B2 triple cores, since I don't buy OEM PC's. A B3 triple core might work for me if I don't want to go B3 now. It would have to be a significant improvement over my 4600+ to warrant not just waiting for Deneb. My jones for the best AMD CPU now is warring with my conservative "keep a CPU for two years" upgrade style.

Will any games besides Supreme Commander need 4 cores before December? I just don't know. I do know that The Witcher, Oblivion and LOTR online do not need it right now.



I think it will last as long as 65nm Phenom's last. I do not think that Phenom yields will change prior to Deneb because there's no hope for 65nm Phenom other than as a B3 improvement to B2. We won't see higher clocks till Deneb. Even then, if all the Nehalem's have hyperthreading, then they will probably trounce SOI Deneb at the midrange and enthusiast and AMD won't make a complete comback until the second version of 45nm Deneb, at best, or the next architecture at worst.

Still, AMD has a place in the OEM and budget enthusiast market, not to mention notebooks because of their better chipsets. I think they'll survive, even if they "lose" the enthusiast market because their CPU's bench a few fps less than the equivalent Intel.
 
AMD gave up the high-end a long time ago. I'm not sure why you guys get so worked up over this. It's not like any of you work at Intel lol. Or maybe you do? The mail room?
 
ok, i haven't read this whole thread, but from what i am hearing, you guys think that the tri core has four cores and one is faulty and disabled. that is totally false. ask anybody from amd and they will kindly show you a schematic. secondly, DON'T FLAME THEM FOR THE IDEA OF SELLING DISABLED CORE CPUS!!! research the CORE SOLO!
im done ranting, let the flaming begin.
 

Thunderman? BaronMatrix? Whoever you are I like your style 😀
 
He has a point though, selling disabled cores at a cheaper cost has a history. The Durons were Athlons with some L2 cache disabled, as were Celerons at the time. Graphics card companies do it also, the X800 and 6800 series were full of cards that had disabled hardware units, some unlockable, some not. Even my 7900GS is a slightly disabled 7900GT.
 


The main"stupid" thing you said was that the tri cores were not quad cores with a bad core disabled. This is not true since AMD themselves have stated that these are Phenom X4's with a bad core that didn't make the cut.

And you are right that there have been many different CPU variations that were just lower versions of the same chip minus cache and what not. Hell my HD2900Pro 1GB is a HD2900XT 1GB with lower clocks. But GPU-Z still sees it as a R600XT chip.

All I am saying is we have to wait until someone gets a Phenom X3 to see what it is capable of. I am not biased against AMD but with so many benchmarks from different sites not showing the same thing its confusing to know why. Did they nerf the Intel setup with low grade crap? Or did they use a AM2+ board? Why is it that there are so many different figures for the same benchmarks?

Either way I don't see the tri core lasting long. After a while duals will be the low end, quads will be the norm and octos will be the high end.

And yipsl, I still don't see how a die shrink will change anything in terms of performance. Yes maybe it will help unlock higher speeds but thats to bee seen. Also a process change might help but I am going to doubt it will do anything more than lower leakage and allow higher clocks. And vs Nehalem, a brand new archtecture, it wont be good if Deneb doesn't improve enough to get AMD some money flowing. If they have to wait until the next architecture to be truly competative that would be very bad.

I also don't see the Phenom 9850BE as fully competative as it still is beaten in quite a few benchmarks by the old Q6600. It even has a 100MHz advantage. I will say it did close the gap a bit so the IPC level is a bit better bit it still isn't going to give AMD a miracle. Maybe Ruiz should start praying to Jesus for one now.
 

Well at first I thought you were trolling since you were so completely and obviously wrong...

Anyway the tri-core Phenom is in fact a quad-core with one core disabled :hello:
 
Well, if i can remember this correctly (bearing in mind I've had *quite* a few to drink tonight) the idea of a tri-core started out as a stand-alone design. However, this stand-alone design had no actual differences than if a quad core simply had 1 of it's 4 cores lasered off, and it would work out cheaper (guessing) to simply keep quad core pattern in the press than taking time (and spending money) to switch in the tri-core pattern (or simply cheaper due to poor-ish yields of quad parts... who really knows?)
 


We can wait and see about Deneb. I believe it will be competitive against Penryn, but fall behind against Nehalem. AMD will probably be in the same situation vis a vis Intel in 2009 that they are in today, except they should have debt paid off from the ATI acquisition and the GPU division should start earning money that can be put into research.

Swift will help vis a vis notebooks, as will great ATI/AMD chipsets, but we'll have to wait for the next architecture for a possible return to the glory days of the Athlon X2 vs. Netburst. Frankly, I don't think Intel will willingly let AMD beat them again, but I hope AMD closes the gap.

I want to amend what I said earlier about the 9850BE and 7xx chipset boards as upgrade friendly for Deneb later this year. It turns out that most 780G boards can't handle 125 watt Phenom B3's because of voltage issues. This isn't a chipset limitation, but a motherboard manufacturer choice because they market the boards for SOHO, HTPC and home PC with light gaming. Many 770 boards are in the same boat, so people need to check their motherboard's CPU list at the manufacturer's site before they buy or upgrade.

AMD said that the 780G's would be a good match for the 95 watt triple cores, so when B3 triple cores come out, that might be the way to best use those boards. I'm not sure performance will be that much greater than an Athlon X2, so if you own one already, waiting for Deneb might be the best choice.

Also, just because many AM2 boards (like ASUS's 690G) have a new bios that support B2 Phenom's don't assume they'll support B3's (even 95 watt) or triple cores. Check the CPU support list first. There's a thread where one person is doggedly insisting that his board supported the 9850BE, but he got a bad CPU, so he sent it back. The truth is that his board did not support a 9850BE. Read the motherboard manufacturer's updated info on their site first to avoid making that mistake.
 
yipsl seems more like baronmatrix every day ...

And that's a compliment.

I miss the baron ... not many uberAMD ppl left out there.

keep it up yipsl !!

I'll try some thunderman emulation as it requires significantly less cognitive effort ... tho the writing style is admirable.


AMD4Life ... <which means I may be stuck with this cpu for a long time>

damm ...

Would praying help ... or should I just keep going with the Intel disinformation strategy??
 

I don't see where you're getting that from. BaronMatrix was either a fool or a troll, but yipsl is just speculating and stating the pros and cons, something that you would never catch the Baron doing :)