Sign in with
Sign up | Sign in
Your question

will dual core be a step backwards short-term wise

Last response: in CPUs
Share
October 5, 2004 12:02:15 AM

Im not a computer expert and don't claim to be as computer literate as most of THs forum posters. My question is that with all the hoopla about dual cores did anyone care to think about the implications of moving from a single core enviroment to multicore desktop enviroment? Doesn't software have to be program to take advantage of a multithreaded enviroment?

I assume that the first multi-core chips from Intel and AMD will consist of cpu cores that are no where near the performance of the highend single cores. In other words, you won't see AMD slap two A64 3800s together from the get go. More like two underclocked A64 2800/3000 in a dual core configuration. When taking account heat output, power usage as well as cost, this is a very likely scenario.

The problem is that most of the software available at the time that dual cores hit the market, will not be able to take advantage of both cores. Lets face it. Game performance will be a major factor in the adoption rates of dual cores. Gamers are what push the desktop cpu performance increases. No one upgrades their cpu to improve their scores on their WordPerfect benchmarks. The A64 wouldn't be %&%* if it could encode a video like a mainframe but played Doom3 or FarCry with a performance of a calculator. It will take time for multi-threaded capable games to hit the market. I doubt a lot of developers will be willing to delay their projects to multithread them, especially if they are within 6 to 8 months of release. They won't waste the money until there is a significant install base of dual cores.

Some say it will be easier to multitask, but how many cpu-hungry programs does one usually run at one time. The ability to encode a video while playing a HL2 on a rig that performs like a xp 2800 (unless valve releases a patch to take advantage of dual cpus) in a gaming enviroment, doesn't seem worth the cost. I highly doubt these things won't cost a pretty penny when they first hit ther market.

I see a slow adoption rates of desktop dualies in the retail market with most gamers spurning them for their faster (GHz wise) single core cousins, until a majority of new PC games utilize a multithreaded enviroment.

These thoughts are opinions and not facts. There are probably aspects of this topic I have not thought about.
October 5, 2004 1:57:00 AM

I agree. There is one more point. The two biggest game engines are just being released. I dont see ID or Valve rushing to make a new engine for dual core.
October 5, 2004 12:39:00 PM

Would a totally new engine need to be built? or could they just re-code parts of the engine and release some kind of "Patch"?

But I do agree that dual-cores won't see wide spread popularity until they have been on the market for 18-24 months, enough time for applictaions and games to be coded to use dual core processors.

XP2800+, Abit NF7, 1GB Dual-Channel DDR333, ATI R9800PRO 128MB, TT PurePower 420W, LG DVD+-R/RW
:redface: <font color=red>My wife says I suffer from premature ejaculation...I don't remember suffering<font color=red> :wink:
Related resources
Can't find your answer ? Ask !
October 5, 2004 12:53:59 PM

They <i>might</i> be able to patch the game engines. Depends on how they're written really. More likely they won't bother, unless CPUs are being a bottleneck, which isn't really the case with either Doom3 or HL2 on any current mid to high-range CPUs anyway, so even if the first Dual Cores are "only" equivalent to a A64 2800+ there won't be any real need for game engines to fully support them.

The next generation of engines perhaps will use them, but that'll probably be at least 4 years away.

Writing Multithreaded apps is extremely painful, and they'll avoid it unless it's actually needed.

---
Epox 8RDA+ V1.1 w/ Custom NB HS
XP1700+ @200x10 (~2Ghz), 1.4 Vcore
2x256Mb Corsair PC3200LL/1x512Mb Corsair XMS PC4000 2.5-3-3-7
Sapphire 9800Pro (VGA Silencer Rev3) @418/742
October 5, 2004 1:00:53 PM

I have been using dual processor machines since the PII, which did not gain any real benifit till the PIII.
Now, I have a hyper-threaded P4.
Your XP os is multi-threaded, and a lot of apps are, which you would benifit from.
Now Games, most are not multi-threaded, I understand, but still benifit from multi-threading slightly by accessing the os.
Would AMD's dual core not be the same concept, and use existing multi threading already being used?



========================
Try everything...
Do not be afraid of failure, for this is how we learn and grow...
Live life to the fullest...
Do not regret what you have not yet done!!!
October 5, 2004 1:00:54 PM

Carmack for the longest time has coded his engines for dual processors. Problem has been getting real performance out of it.

Xeon

<font color=red>Post created with being a dickhead in mind.</font color=red>
<font color=white>For all emotional and slanderous statements contact THG for all law suits.</font color=white>
October 5, 2004 1:11:07 PM

Quote:
Would AMD's dual core not be the same concept, and use existing multi threading already being used?

There would probably be a small benifit like you say, but it'll probably be more because the game can have a whole dedicated core to itself, while the other one runs your background apps (antivirus etc).

There could concievably be an advantage from doing something like having one core do the game AI and/or Physics, while the other handles the other stuff, but until it becomes complex enough to really overwhelm a single processor I can't see them bothering. Plus if it came to that they'd be alienating all the existing Single-core customers (who will be in the majority for years to come of course).

So they won't bother because:
1)Too much effort
2)No real need at present
3)very small % of people would take advantage for a long time.

---
Epox 8RDA+ V1.1 w/ Custom NB HS
XP1700+ @200x10 (~2Ghz), 1.4 Vcore
2x256Mb Corsair PC3200LL/1x512Mb Corsair XMS PC4000 2.5-3-3-7
Sapphire 9800Pro (VGA Silencer Rev3) @418/742
October 5, 2004 2:10:18 PM

This is a bit of a guess but I would imagine programmes for dual cores have become more common so that they can take advantage of intels hyperthreading which has been out for a while now.

I respect your ideas but i think you have probably painted a worse case scenario picture, for example cool'n'quiet type technology would somewhat reduce the power problem.

remember that it will be aaggggeeeess until dual cores really take over, only super high end chips will be dual core for ages.

Also dual cores will come in at approximately the same time as 64bit windows so maybe when programers are redesigning for 64bit they will make the dual core change as well?

.
October 5, 2004 2:25:46 PM

Quote:
Also dual cores will come in at approximately the same time as 64bit windows so maybe when programers are redesigning for 64bit they will make the dual core change as well?

Existing stuff doesn't really require much in the way of change for 64-bit. Changing a single-threaded app into a multi-threaded one is a nightmare.

---
Epox 8RDA+ V1.1 w/ Custom NB HS
XP1700+ @200x10 (~2Ghz), 1.4 Vcore
2x256Mb Corsair PC3200LL/1x512Mb Corsair XMS PC4000 2.5-3-3-7
Sapphire 9800Pro (VGA Silencer Rev3) @418/742
October 5, 2004 2:42:25 PM

really? are we talking about complete start from scratch? i always got the impression that people were dreading the 32=>64bit, if dual core is worse then there are going to be some angry programers over the next few years, and a lot of buggy software.

.
October 5, 2004 3:43:47 PM

Quote:
Now Games, most are not multi-threaded, I understand, but still benifit from multi-threading slightly by accessing the os.


actually performance suffers in most games, theres been articles but im too lazy to link them

-------
<A HREF="http://www.albinoblacksheep.com/flash/you.html" target="_new">please dont click here! </A>
Brand name whores are stupid!
October 5, 2004 3:53:41 PM

Dual core isn't worse. you can write multi-threaded apps for single processor systems (assuming the OS supports it, which Windows has for years). It's just debugging a multi-threaded app is a complete headache.

If they don't bother writing stuff as multi-threaded, then It'll just use the one core all the time, and that'll get handled by the OS.

It's just like using SSE2 or MMX or something - the extensions are there, and may be useful for some things, but you don't <i>have</i> to use them.

---
Epox 8RDA+ V1.1 w/ Custom NB HS
XP1700+ @200x10 (~2Ghz), 1.4 Vcore
2x256Mb Corsair PC3200LL/1x512Mb Corsair XMS PC4000 2.5-3-3-7
Sapphire 9800Pro (VGA Silencer Rev3) @418/742
October 6, 2004 3:25:38 PM

Article on the THs homepage

"Once again it is AMD that takes a step forward and promises that dual-core Opterons will deliver at least 30 percent, and in a best-case scenario up to 55 percent more performance than their single-core siblings."

you think that was a single core vs. multicore using a singlethreaded app. Probably not. I bet any singlethread app will run faster on a single core opteron. Since the cores on the dualcore will be three to five speed steps slower than their single core counterparts. This won't be a problem with the Opteron line. Since there are multithreaded apps already available in the server or workstation market. But desktop, thats a different a story, especially games.
October 6, 2004 3:51:53 PM

yeah, i wonder how they will do the fx line. maybe they are producing opteron first to give them time to revise the spec to where they could get 2 ghz or more per core on an fx maybe, or to give developers tiem to make optimizations lol.
October 7, 2004 1:20:04 AM

This is really coming from left field, but Amd, Intel and M$ are probably working on some type of extensions that would allow some division of load, say seperating AI and physics to the different halves. It might even be possible to do through Directx.
October 7, 2004 8:59:38 AM

hmm.. I hadn't thought of it being done that way, but that would make a great deal of sense, and would make life much easier for the developers.

I guess we'll just have to wait and see.

---
Epox 8RDA+ V1.1 w/ Custom NB HS
XP1700+ @200x10 (~2Ghz), 1.4 Vcore
2x256Mb Corsair PC3200LL/1x512Mb Corsair XMS PC4000 2.5-3-3-7
Sapphire 9800Pro (VGA Silencer Rev3) @418/742
October 7, 2004 9:25:51 AM

<i>Since the cores on the dualcore will be three to five speed steps slower than their single core counterparts.</i>

Just wondering if this is based on something AMD has officially announced?

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
October 7, 2004 12:50:27 PM

I thought DirectX already could deal with AI and physics calls, I could be wrong though.

Xeon

<font color=red>Post created with being a dickhead in mind.</font color=red>
<font color=white>For all emotional and slanderous statements contact THG for all law suits.</font color=white>
October 7, 2004 5:14:06 PM

http://www.anandtech.com/news/shownews.aspx?i=23132 or
http://www.eweek.com/article2/0,1759,1666804,00.asp

"Kevin McGrath, an AMD fellow, noted that the company's dual-core architecture built on the groundwork laid by AMD's "Hammer" processor—now known as the AMD Opteron....McGrath also answered key questions about the performance of the dual-core chips compared to their single-core cousins. To minimize power under 95 watts, the dual-core chips will be under-clocked. As part of his presentation, McGrath compared the performance of dual-core chips versus a standard two-way, unicore system.

A two-processor dual-core system clocked 5 clock speeds down from AMD's fastest part will generate performance of approximately 125 to 140 percent performance of the dual-processor unicore system, McGrath said.

For example, a dual-core, dual-processor system clocked at only three grades slower than the fastest AMD chip will perform at between 130 to 160 percent of the performance of the base system, McGrath's slide indicated. McGrath's performance figures used synthetic benchmarks, such as SPECint_rate 2000 and SPECifp_rate 2000."
October 7, 2004 10:14:47 PM

Thanks for the links. That's a little disapointing. I'm not upgrading my 248s until they have dual core opterons operating at 2.4ghz or higher. :smile:

Found this link through the anandtech link(in the comments section):
<A HREF="http://www.eet.com/semi/news/showArticle.jhtml?articleI..." target="_new">http://www.eet.com/semi/news/showArticle.jhtml?articleI...;/A>

Thought it was worth a read for anyone interested in AMDs dual cores...

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
October 7, 2004 10:41:21 PM

I find the whole story very, very disappointing. I mean, they're talking about a very hefty underclock there, and even then, the best synthetic benchmark they can churn (and synthetic speaks lies in dual-CPU territory) is actually only 30% faster? My dual P3 933Mhz is like 197% of the speed of a single 933Mhz in those benchmarks!

The only possible case in which dual core CPUs might become interesting is if (a) their multipliers come unlocked and (b) you buy yourself a nice water cooling system. And even then, the usefulness of actually having a dual-core system clocked equal to a top-of-the-line single-core system is doubtful, because of unoptimised programs. If you're still interested in multiple CPUs, you could just go the Opteron way anyway. And because dual-core will initially carry FX price tags, they're bound to be a niche product for quite a while before dual-core is of any importance...

So bottom line: Dual core is way, and I mean waaaaay, overhyped. Sad but true. Dual-core will only matter for the most of us in 2006, and that's being optimistic.

<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 10/07/04 09:44 PM.</EM></FONT></P>
October 7, 2004 10:52:15 PM

On second thought, I think there'd be only one possible dual-core processor which would actually live up to all hype: a "desktoped" (nice word, huh?) dual-core Dothan or similar. Such a CPU would have a much better architecture overall than Netburst, and would NOT require any underclock; actually, they might even try to overclock current 2Ghz models. Current dothan architecture is good for at least 2.13Ghz, possibly more; a desktop/workstation/server version of it at 2.13Ghz or more would probably easily beat this heavily underclocked Opteron to the punch. (unless AMD actually cranked up the clock a bit on Toledo, but I seriously doubt that they'll manage that)

Given proper x86-64 extensions, such low-power, dual-core Dothans would kick ass in servers and particularly high-density blades. Why not make such a product compete directly with Toledo? It looks so much more easier than try to force a half-baked dual-core netburst-based processor down the market's throat.

This is even more indication that Intel should ditch current Xeon plans and dual xeon plans for a smarter, dual-core dothan approach.

<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 10/07/04 09:54 PM.</EM></FONT></P>
October 8, 2004 12:05:47 AM

DualDothans would be nice.

But there is something that makes me believe that while Dothan work well without being power hungry and dispensing much heat. I doubt you can just crank up the GHZ on these things and turn them into small heat plates like you can with the Prescott. If you could, why hasn't Intel just dumped the Prescott and released souped up DesktopDothans at 2.5 GHZ or greater.

While using a DualDothans in a desktop might be feasible at stock speeds. The reality is that mass migration of desktop apps from a single core enviroment into a multicore enviroment is a long way off. Also, any dualcore would be priced right at or above any highend single-core solutions. Since a dualcore anything is still off by 8 to 10 months away. A dualcore desktop will be competing with both Intel's and AMD's single core offering at around 4.2 or 4.4 GHZ (4200+ or 4400+ for AMD). I doubt we will even see any Dothan at 2.4 speeds, since they aren't meant to ramp as fast as either the desktop architechture of Intel itself or AMD. Which means any single core, in the same price range of dualcores, will smoke the dualcores in any singlethreaded apps, given the assumptions that the individual cpu of a Prescott/A64 dualcore will be gimped or that the Dothans can't ramp as fast as a Prescott/A64.
October 8, 2004 12:33:07 AM

<i>I find the whole story very, very disappointing. I mean, they're talking about a very hefty underclock there, and even then, the best synthetic benchmark they can churn (and synthetic speaks lies in dual-CPU territory) is actually only 30% faster? My dual P3 933Mhz is like 197% of the speed of a single 933Mhz in those benchmarks!</i>

30% faster than a dual processor unicore system running at higher clock speeds. I'm not sure how that's bad at all?

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
October 8, 2004 1:21:50 AM

Another thing that people seem to be forgetting is that windows already can monitor cpu usage. This is the first step to having the OS designate core usage. For most situations, this would only help while multi-threading, but it is a step in the right direction. Many of the available cpu intensive progs are already set up for dual or multi cpu usage, others are HT enabled, and smaller progs could be handled by the OS. Games are the biggest problem, but if we get something like 3Dnow, with DX7, even that problem pales.
October 8, 2004 1:41:07 AM

Yeah, games are going to be an issue. A lot of gamers are going to say this sucks. Oh well. Even an AMD spokesman said in an interview of ExtremeTech that this isn't going to be for gamers at first. For the non-gamers, this is largely going to be good tech. Especially in servers/workstations.

I'm still disapointed in the clock speeds, and I share the skepticism of the synthetic benchmarks, but if they are able to produce dual core chips that run at lower clock speeds, yet just as fast as the latest unicore dual processor systems, this is going to be a winner. Depending on the price that is.

I just keep thinking about being able to upgrade my system to a 4 way, without having to plunk down the money on a 4 way motherboard, and Opteron 8 series processors. Though I suspect the price of these things are going to push my upgrade plans a year down the road. I'm not spending more than 800 bucks a chip. That'd be worth it for my needs. But I want faster clock speeds.

I want my cake and I want to eat it too. :lol: 

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
October 8, 2004 1:48:31 AM

i really wonder what you were expecting fomr the first dual cores anyway? you expected a instant doubling of performance? i dont know where you got those rose colored glasses lol, but really 30-50% is in line with what i expected. at least with this number, we can tell amd is at least trying to be real here. id rather them admit this up front then try to hype something that doesnt materialize.

and we have no idea about pricing yet, since the first dual cores will be opterons, we may see it simply take the spot of the high priced single cores now, while the single cores drop in price. thats very possible in that area.

now for dothans, where is the evidence these things do well in server /workstation situations? i havent seen any such benchmarks to confirm this. you say they would rule there, but there is no such evidence. dothans are excellent chips in the mobile area, but utnil i see some new tests done, its nto much more. it has alot of potential and if intel is smart it will be planning more wiht the architecture. but making them out to be the second coming for intel is a bit much at this point.

so now about the underclcoked ahtlons, 3-5 times is what was said. so that means that at intro there could be 2ghz per core opterons, while there were 2.6ghz opterons. tahts possible i would say. really, if thats true, then 2ghz per core isnt so bad is it? this is all relative to waht amd has single core wise for opteron by that time. so we will see.
October 8, 2004 10:10:30 AM

We really need to look what dual core is aimed at.
We are talking opteron and not Athlon64 at the moment, hence the server space.
Frequency is generally not the driver from business apps, however parallel processing is. I.e. servers will have a high concurrent user base.
Doubling the number of cores will (in user terms) generally double the number of users, assuming we are not running at 100% CPU usage.
Maximum number of users will be less than a dual CPU setup, clocked higher.


Look at traditional servers (i.e. SUN/HP/IBM), in the past the CPUs were not clocked very fast, but had a large number of CPUs to support a large user base, but were still able to support more concurrent users than a system with half the CPUs and double the frequency.

For a customer running at 80% CPU utilisation with a fully loaded 2 or 4 way system, just changing the CPUs to dual core even if the frequency is 20% less will give them a 100% uplift in the number of users supported.
If it were running at 100% CPU they would obviously only get an 80% uplift.
October 8, 2004 3:18:09 PM

Well i think it will be a great revolution on the market. I do not know the architecture of the dual core procs, but see it this way: it should imitate dual processors in any way... maybe memory management or other stuph will not be duplicated. anyway supposing it imitates dual proc systems, in most cases of 'regular' cpu usage it should somewhat almost double the performance, since if u are runnin several apps, they can be running on separated cores because they are dependant. If u run a server like apache, or IIS, this is extremly usefull since there are always lots of service runnin and requiring processor. When double core applications enter the market, this will also run at somewhat the double of speed, depending on the apps. Now for the commom gamer, it all depends on what the programmer can do. When there was no GPU, most of graphics calculation was handled within the proc; the proc also handled physics engine and AI. now graphics calcs are runnin in GPU. did it improve the quality of games? yes. did it improve quality of apps. no. It all depends on how programmers can (re)design their apps to make thins run faster. in these days the bottleneck of graphics is the graphics card speed. So i think for gamers it is much more interesting SLi than dual cores, for the coming year. I think the real revolution will occur in the end of 2005 when serious dual core apps will begin to enter the market
October 8, 2004 6:39:46 PM

http://www.extremetech.com/article2/0,1558,1666805,00.a...
This is from the same article

"However, Kevin Krewell, editor of The Microprocessor Report and an analyst for In-Stat/MDR, said that most games will still favor single-core chips for the time being.
"Everybody's going multicore, but the best single-threaded performance will still be on a single-core processor," Krewell said. "Don't expect a [dual-core chip] inside gaming PCs; most gaming apps are still single-threaded, and those things will come later," he said of multithreaded games. Intel's hyperthreaded Pentium 4 architecture was designed for multithreaded applications."



I use Maya 6.0 alot and it would probably take advantage of a dualcore system. However, I could tolerate lower performance in Maya alot more than I could in any new game, especially considering A64 dualcores will be in a FX price range. Consider if a AMD dualcore was released on the Oct18. Would you rather have a FX55 or a dualcore 4000+.

I picked a dualcore 4000+ instead of FX55 dualcore as an examples because the price of a dualcore FX55 would probably be priced just about out of anyone hands. A 4000+ since its performance gains are related to more cache, higher Hypertransport speed and/or different chipset then the other 2.4GHz chips.

Now here is where it gets hairy. Now if these things follow the opteron and are 3 to 5 speed clocks slower. A speed clock looking at AMD's recent clock increases means 200mhz. That means a chip that operates from anywhere from 1.8Ghz to 1.4Ghz with 1MB cache and 1GHz HT. Doesn't look to appealing against a FX55 using a singlethreaded app does it.

Now purchasing a dualcore in about 3 years, after a majority of games have being released optimized for dual core and dualcores themselves operated at speeds closer to their single core cousins, will be more appealing.
October 8, 2004 8:30:22 PM

Wow... Great debate here. ( if it is that)

Only time will tell, we can only forcast devolopers and how the economy will adapt. I'm realy excited to see how these turn out!


PS. Not one troll or flame in here. WOW!!!!

"If youre paddling upstream in a canoe and a
wheel falls off, how many pancakes fit in a doghouse? None! Icecream doesn't have bones!!!"

"Battling Gimps and Dimbulbs HERE at THGC"
!