• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Phenom II X8

O

octoberasian

2[H]4U
Joined
Oct 13, 2007
Messages
4,081
Thought maybe some people will be interested in this, so making a new thread about it instead of derailing Mr Spock's thread any further.

http://www.fudzilla.com/processors/item/25335-amd-preps-bulldozer-based-phenoms
"Well, according to some fresh leaks, AMD plans to introduce a wide range of Bulldozer based Phenoms, including quad-, six- and eight-core chips. All will feature unlocked multipliers and depending on the clocks and number of cores, they will be available in 95W and 125W flavours. Clocks range from 2.5GHz to 3.2GHz and like their FX siblings, flagship models will feature 8MB of cache."​

http://www.inpai.com.cn/doc/hard/163854.htm

- ZD242046W8K43, AMD Phenom II X8 2420,2.40 GHz, 95W;
- ZD302051W8K44, AMD Phenom II X8 3020,3.00 GHz, 125W;
- ZD252046W6443, AMD Phenom II X6 2520,2.50 GHz, 95W;
- ZD282046W6K43, AMD Phenom II X6 2820,2.80 GHz, 95W;​

capturekns.png


dbe607a7-bde2-48fa-be0a-766f61799dfb.jpg


The naming schemes is confusing or misleading. These are intended to be mid-range to entry-level processors.

Also, if going by that chart, there is another 4100 model with 8 cores. There already is an FX 4100 with 4 cores (2 modules/4 threads). :eek:
 
2.4ghz on a clock-dependent chip that needed double that to convincingly beat their older offerings AND it's at 32nm and still at 95W. 4 core chip at 3.6ghz that should be significantly slower than the quad-core Deneb Phenom II's at the same clock speed at 45nm. That's unbelievable... When you can't beat your predecessor with your best chip then there's little hope for these failed samples will sell to anyone blessed with the gift of thought. Curious, is it 8MB total cache or 8MB L2 + 8MB L3?
 
theres a couple comments on the fudzilla one that makes a bit more sense that the actual phenom II name is just a miss label by the motherboard bios. but who knows, we'll just have to wait and see. as far as them being unlocked goes thats obviously wrong since all AMD engie samples are unlocked so theres no way to say they will or won't be unlocked.



pelo said:
2.4ghz on a clock-dependent chip that needed double that to convincingly beat their older offerings AND it's at 32nm and still at 95W. 4 core chip at 3.6ghz that should be significantly slower than the quad-core Deneb Phenom II's at the same clock speed at 45nm. That's unbelievable... When you can't beat your predecessor with your best chip then there's little hope for these failed samples will sell to anyone blessed with the gift of thought. Curious, is it 8MB total cache or 8MB L2 + 8MB L3?
Click to expand...


2mb L2, 8mb L3 cache.
 
I hope that's the case cause like I said in the other thread, it would be downright idiotic to come out with a Phenom II X8 that is slower than a X6.
 
This is actually not very good news in terms of waiting for Bulldozer price drops. I don't think AMD would have sold or planned these chips had their other parts passed QC and clock-to-TDP tests. The fact that we have crappy Phenom IIs with piss-poor clock speed means that they might be experiencing yield issues at GloFo or that it's very difficult for GloFo to make good BDs at volume on a gate-first approach -- which isn't surprising, really, as a high clock speed design would have been difficult for them in the first place.

Given the large amounts of cache and the reasons above and that the real transistor count above their "estimated" 1.2B, I don't see these chips being sold at a price low enough to be worthy of recommendation. I'd imagine the 8-core chips to cost more than the 4 and 6 core BD FXs, so the BD-based 32nm Phenom IIs cost a bit below their equal core FX counterparts. Of course, I could be wrong and these could potentially be dirt cheap.

AMD, if you're listening, please stop. Just quit it. Don't do anything on the CPU side until you've redeemed yourself with something worthy of the money you're asking for it. Currently you're only embarrassing yourself further
 
There are a number of problems with this strategy from AMD. Most obvious one is that FX needs high clock speeds to be half decent (note half decent not amazing!)
Unless they are amazingly cheap I really see no place for them I'd rather just buy bargain Athlon II's and Phenom II's for builds.

I mean the normal 4 core FX gets thumped with quad core older AMD processors (that cost less) this cut down 4 core isn't going to impress. Ditto on the 6 module FX go buy a Ph II 6 core processor ;-)
And the 8 module ones are too slow clock speed to really get much done at all, unless you are doing major multi threading they could prove pretty damn slow for lower threaded workloads

Had AMD decided to juice up the old Athlon/Phenom II designs and bring them to 32nm tech with some tweaks we'd have a reason to be excited, right now I can only see more horrible reviews for AMD and more lost customers! But Iike I said price well if they are dirt cheap maybe they might appeal to some, if not they are DOA no doubt about it
 
maboblivion said:
If this is right, seems the bulldozer train wreck is only beginning.
Click to expand...

They don't have much of a choice but to embrace their newer architecture, unfortunately it sucks so anything Zambezi/Bulldozer-derived will suck as well. Far too many transistors coupled with poor performance, slow cache and reliance on high clock speed means high TDP and power consumption. CMT-tax means that the fewer core and lower clocked parts suck too but they'll suck under a cooler thermal envelope.

If the Trinity mobile chips show a massive improvement in IPC over Llano then there's still hope in their Piledriver cores, but

Virgo_platform_details.jpg


AMD's own slides have shown that there's a 30% increase in digital media workloads, which is the same shit they were selling us with BD. Unless those high clock speeds are coupled with far faster and smaller cache for the desktop parts (the APUs are already stripped of L3), then this Bulldozer fail-train will run express stops until Intel completely dominates both the server and performance desktop arenas.


maboblivion said:
If this is right, seems the bulldozer train wreck is only beginning.

http://www.tomshardware.com/reviews/bulldozer-efficiency-overclock-undervolt,3083-19.html
Click to expand...

That Tom's article is one of the reasons that site has gone down the drain (along with the horrendous ads, shitty journalism and "News" that's essentially a plug for a company with a new product). They decide to test BD's efficiency and take an in-depth approach but make no conclusions, no assumptions based on benchmarks and fail to compare it to their previous architecture to test whether AMD's claims were correct.
 
Last edited:
Mostly posted it so ppl could see the efficiency it has in best terms with less voltage.

That's why i come here on hardocp , TWITMTBR (The way it's meant to be reviewed ... too long ? )

It's safe to say that amd's claim were incorrect :p
 
saw title - 'Phenom II X8', blood started to rush away from extremities as i thought... 32nm shrink for K10!.
saw that the PII X8's are based on BD cores, blood returned to brain, reality returned.

sigh, what is AMD thinking (if accurate). is there a crack smoking room at HQ where these decisions are made?
 
Maybe they can just glue a bunch of their video cards together and pretend it's a processor. That would probably be better than bulldozer.
 
InternationalHat said:
Maybe they can just glue a bunch of their video cards together and pretend it's a processor. That would probably be better than bulldozer.
Click to expand...

think that's what AMD is ultimately heading towards with their fusion line of CPU's. instead of just integrated GPU on die with the CPU (tho still separate), they'll be completely merged to share resources between each other. for example the CPU will off-load all floating point stuff to the integrated GPU.

what i don't get, AMD has already destroyed one awesome brand (FX) and now they'll be diluting what was a semi decent one (Phenom II). shit the cupboard will be completely bare after this, brand wise. they'll have to rely on their winning personality to bullshit people into buying their stuff from now on. either that or actually release a modern competitive processor.
hmm, there's a thought...
 
Brand new Opteron Thunderbird Barton Athlon64 6000+ processors are going to be/were a hit!
 
As if it wasn't enough the confusion to the customers with Propus cores sold as "Phenom" now they will add Bulldozers into the "Phenom" brandname? Who's the genius that thought of it?
 
Faethon said:
As if it wasn't enough the confusion to the customers with Propus cores sold as "Phenom" now they will add Bulldozers into the "Phenom" brandname? Who's the genius that thought of it?
Click to expand...

gee you know i kinda feel sorry for AMD right now. they are such an easy target. but ham-fisted shit like this isn't doing them any favours.

i thought they sacked all the marketing personal cause this sounds like a move a marketing dept would make. perhaps there is only 1 marketing guy/girl left at AMD and he/she's cursing Rory b/c work once done be a thousand must now be done by one.

marketing types generally being 'glass is half full' people, rather than give up, decides to out-source marketing 'think-tanking' to his kids kindergarten class. sure the quality of the work done is a little sub-standard (being the proposals/reports are submitted on butcher's paper & written in crayon and finger paint) but they work on the cheap. chuck in a few bags of skittles and a couple buckets of red cordial (hey cheaper than ritalin or cocaine) and you're good to go.
 
thread title had me excited...then i read more and i stopped being excited.

I can't add anything else :p
 
Probably a mislabel, but could easily be an evil marketing strategy to further drive Phenom II out of the minds of future customers. Currently people are scratching their heads trying to figure out which CPU to buy, the Phenom II or FX. AMD doesn't like or want this so the easiest solution is to release a cripped FX as a new Phenom II so the choice is obvious: buy either slow, or super slow.
 
Admittedly my knowledge of CPU fabrication is poor, but given the lackluster performance of Bulldozer, wouldn't it have been more cost-effective to just build more Deneb or Thuban cores on a 32nm process? AMD's Phenom II line still offers a great value in the lower-priced segments, with a chance for better performance through overclocking, but they are not really enthusiast chips. Wouldn't going to a 32nm process improve power consumption and heat output a little, therefore making a chip in the same price range that uses less power, runs cooler, and might have more overclocking headroom, an even better value?
 
FlawleZ said:
Probably a mislabel, but could easily be an evil marketing strategy to further drive Phenom II out of the minds of future customers. Currently people are scratching their heads trying to figure out which CPU to buy, the Phenom II or FX. AMD doesn't like or want this so the easiest solution is to release a cripped FX as a new Phenom II so the choice is obvious: buy either slow, or super slow.
Click to expand...

MrCaffeineX said:
Admittedly my knowledge of CPU fabrication is poor, but given the lackluster performance of Bulldozer, wouldn't it have been more cost-effective to just build more Deneb or Thuban cores on a 32nm process? AMD's Phenom II line still offers a great value in the lower-priced segments, with a chance for better performance through overclocking, but they are not really enthusiast chips. Wouldn't going to a 32nm process improve power consumption and heat output a little, therefore making a chip in the same price range that uses less power, runs cooler, and might have more overclocking headroom, an even better value?
Click to expand...

I know it's been mentioned a few times already here on [H] that AMD is planning on stopping production on Deneb/Thuban models. In my opinion, it would be a bad idea. The Phenom II CPUs already outperform or match a "high end" Bulldozer FX CPU, and AMD don't see that or admit to its failings. The only areas I've seen Bulldozer FX excel in over the older architecture is video encoding, AES performance, and memory bandwidth (compared to Phenom II, not Sandy Bridge).

I agree with you and a lot of other posters who thought of the same suggestion: die-shrunk the Thuban/Deneb models to 32 nm. It would have been preferable to do that instead seeing on how the Bulldozer architecture performs against the older one. As we've seen with the 28 nm GCN-based 7970, the move to a smaller fabrication process node helped reduce temperature and power consumption.

AMD really needs to get its ass in gear or just stick to its APU and mobile brands because come March, Intel is releasing a whopping 11 CPU models based on Ivy Bridge if going by the last article regarding them. And, Intel isn't stopping there either. They have Haswell in a year or so after Ivy Bridge. May as well throw in the towel and secede the desktop CPU market to Intel if you aren't willing to get better engineers and CPU designers, or a better CPU architecture to compete in this market.

With Piledriver, AMD has to achieve several things in my opinion before bringing it to market:
  • Vastly improve pipeline length and the amount of registers and FPUs in the CPU. This here holds the CPU back. 2 ALU & 2 AGU integers (per core)/4 FPUs (shared per module between 2 cores) in BD versus 3 AGU & 3 ALU integers (per core)/3 FPUs (dedicated per core).
  • Decrease power consumption and heat output.
  • Increase IPC per core.
  • Increase overall performance significantly instead of the suggested 10-15% improvement over Bulldozer.
  • Increase cache speed and decrease cache latency. Intel's Sandy Bridge cores have L3 cache clocked at the same speed as the core, and a vastly superior interior I/O bandwidth and inner-chip connection.
But, doing this will require AMD to rethink and redo Bulldozer's architecture, and possibly rework the whole thing from the ground up.

For all intents and purposes, I know that will never happen. Even branding lower spec-ed BD CPUs as Phenom is adding insult to injury to the brand name.
 
Last edited:
MrCaffeineX said:
Admittedly my knowledge of CPU fabrication is poor, but given the lackluster performance of Bulldozer, wouldn't it have been more cost-effective to just build more Deneb or Thuban cores on a 32nm process? AMD's Phenom II line still offers a great value in the lower-priced segments, with a chance for better performance through overclocking, but they are not really enthusiast chips. Wouldn't going to a 32nm process improve power consumption and heat output a little, therefore making a chip in the same price range that uses less power, runs cooler, and might have more overclocking headroom, an even better value?
Click to expand...

AMD should hire you "NOW" ;)

Yes bottom line is they could have re-worked the older processors at the new 32nm process and they would be superior to the FX ones we have now.

So damn simple but some noob must have been running AMD because reality completely escaped them.
 
I like the over-arching goals that seem to be driving the bulldozer architecture. I believe that it is too early in the game to lean into leveraging GPU for floating point. It looks as if, for simple floating point, that each module would be 2 full-fledged cores, and that the only time you get a wait is when 128 bit floating point comes through. I am guessing that poor cache latency and L1 cache size is holding back somewhat, along with the real world performance of the FPU not being as seamlessly adaptive as on paper.

Despite the CPU line not being anything near as bad as the Pentium 4 architecture, the tradeoffs do ring true to a step in that bad direction. I must admit that a lower clocking, low-latency cached part, with dual 128 FPUs would make more sense. Without knowing how these processors work from a low-level standpoint, I hesitate to call them "bad" when the performance will actually improve as the OS platforms continue to evolve towards the same operation mode as these CPUs.

Since I intend to keep the bulldozer that I eventually order, I have no issues waiting a year or so for it to be more relevant. I keep my procs for 1 to 3 years, with 2 being typical. I believe that the FX should be a 3 to 4 year part for my use.
 
Well I'm FX ready but to date I can't say anything grabs me (price is a problem IMO they are not priced well at the moment) but that might change. If AMD can sort out performance (ie quite a lot better), price it at more competitive levels, and address the power consumption issues.

It might be worth a look.
I suspect it's one of those "best to wait" strategies for FX ready folks I'm perfectly fine with my current CPU which is quite decent performance wise for my needs. I'll take a look at Bulldozer later on, but only if it's logical to do so. I don't really mind if they call another version Phenom II if it's priced right and improved it could be worth a look.

At the moment it's Phenom I all over again at the time not as good as expected and it took a while to get it to be at least worth looking at.
 
I think they don't have time for all this.
Wrong strategic planning from amd.as soon as they saw BD lacking in performance they should of worked on a cpu with raw power,based on native cores.and perhaps that is what they should be doing now to get back some trust back from their fans.and to have time to fix their shit.
I see the BD more as like a prototype.But i like the idea of BD,but its also based on a server cpu if,
i am not wrong about this.
 
biggston said:
I see the BD more as like a prototype.But i like the idea of BD,but its also based on a server cpu if,
i am not wrong about this.
Click to expand...

I was thinking the same thing about the server CPU architecture. The benchmarks where Bulldozer seemed to do well were in server/workstation applications like encryption (TrueCrypt 7.1 Benchmark), compression (7-zip Benchmark), etc. One possible conclusion that I can draw is that AMD was looking at gaining ground in the enterprise market, but their marketing team blew it by trying to entice enthusiasts with unlocked multipliers and by bringing back the FX-branding.

From that perspective Bulldozer at least makes sense, even though it still under-performs in some areas (Microsoft Excel - Monte Carlo Simulation). If you have an organization that regularly compresses backup archives for example or works with encrypted data the extra "cores" and design changes seem to be useful. The productivity performance doesn't appear to be that bad either as the 2600K is some stiff competition.

IBM seems to have done quite well ditching the consumer market in favor of the enterprise market (which I would guess was a large part of the motivation behind HP considering the same thing a few months back). I even remember reading an article (though I can't find it at the moment) where Acer was talking about focusing on building quality machines to attract business customers instead of continuing to saturate the market with the lower-end products you see at Wal-Mart.

Out of curiosity, has anyone had any experience with a Bulldozer-based server running database applications or Exchange that could provide some information regarding Bulldozer's performance in these enterprise areas?
 
AMD need to get some chips out the gate. It's getting slim pickings out there. Stocks of all AMD CPUs (expect the lower BDs) are drying up.
 
95 Watt 8 Cores from AMD is a good sign that PD will definitely happen. Plus their fab yields are working out they way they want them to or they wouldn't have all these other chips coming out. I agree with the guy above me, they NEED more than 4 chips on the desktop platform ASAP! SO this is good news, things are running very smooth for them now.

The more chips AMD puts out the more money they make, and the more their engineers will continue learning about how their new design can be improved upon. You people that are hating AMD FX have issues. An 8 core 95 watt version sounds very interesting if priced aggressively enough, AMD is on the right path. I'm in.
 
aphasia said:
saw title - 'Phenom II X8', blood started to rush away from extremities as i thought... 32nm shrink for K10!.
saw that the PII X8's are based on BD cores, blood returned to brain, reality returned.

sigh, what is AMD thinking (if accurate). is there a crack smoking room at HQ where these decisions are made?
Click to expand...

My exact reaction :mad:
 
what is AMD thinking (if accurate). is there a crack smoking room at HQ where these decisions are made?

I think intel paid some people at AMD to make sure that BD fails. LOL
They need to hire new staff and fire all of them.
All this is looking really bad.
 
AMD needs a radical change to keep them in the game. Intel could afford the Pentium 4 fail... im not sure if AMD could handle a pile driver fail
 
According to this site, Vishera will be AM3+ compatible which is good news for those of you who own AM3+ motherboards and may contain up to maximum of 10 cores as opposed to the rumored 8 cores and will have a "more advanced generation piledriver chips" (whatever that means). I have a feeling that they will get it right when Vishera comes out in late 2012 or early 2013. As someone said earlier, think of BD as the Phenom I and Vishera as Phenom II.

Check out the link below to see AMD's new roadmap.

http://www.xbitlabs.com/news/cpu/di...ure_Peculiarities_to_Software_Developers.html
 
psy81 said:
According to this site, Vishera will be AM3+ compatible which is good news for those of you who own AM3+ motherboards and may contain up to maximum of 10 cores as opposed to the rumored 8 cores and will have a "more advanced generation piledriver chips" (whatever that means). I have a feeling that they will get it right when Vishera comes out in late 2012 or early 2013. As someone said earlier, think of BD as the Phenom I and Vishera as Phenom II.
Click to expand...

but will they give me a chipset with PCIe 3.0 and native USB 3.0?

http://hardforum.com/showthread.php?p=1038221590#post1038221590
 
I forget where the article was posted, but basically Bulldozer was computer optimized and not hand tuned at all. The end result was bulldozer has a IPC approx 25% lower than what was expected from an optimized design. The really bad news is that even if pile driver features some of those improvements its still not going to be on equal footing with Sandy Bridge. If Intel ends up having manufacturing issues this summer, AMD at best can put themselves back in a similar situation that they were in with the phenom II. All in all it does not look good for AMD.
 
cloudpost said:
AMD needs a radical change to keep them in the game. Intel could afford the Pentium 4 fail... im not sure if AMD could handle a pile driver fail
Click to expand...

Unfortunately I don't think AMD has the funds to do dirty business practices that Intel did during those Pentium 4 fail days.
 
I still believe that fixing the slow high latency cache and the power usage when overclocked are both essential for pile driver to achieve the position it was designed to be. I mean the top cpu performing somewhere above Intel's mainstream platform and competing with the low end of Intel enthusiast platform.
 
Hi Folks,

i've been running 16 of the cores (2 per die) on my 48 core Opteron (K10 based) system at 3.3 Ghz... i set thread affinity to these quick cores, and ran Cinebench 11.5 with 8 threads... just to see how a simulated Phenom X8 might have performed :)

https://www.youtube.com/watch?v=ppvOUAVrHfQ

scores ranged from 7.49 to 7.6 at 3.3 Ghz... not bad.

Just for reference... FX-8350 typically runs 6.90 at 4.0 Ghz... i7-3770K typically runs around 7.6 at 3.5 Ghz.
 
Its ballpark. This system is running at DDR3-1333 so i would imagine that a real X8 would have been slightly faster. CB 11.5 doesn't seem too ram speed dependent... i overclocked my 12 core Opteron 8439SE (running DDR2-666) setup to 3.2 and the per core results were not much different from running 12 threads on the quad at 3.2 GHz...

JustReason said:
You would have to simulate the entire system as well not just CPU.
Click to expand...
 
Last edited:
You must log in or register to reply here.
Back
Top