ATI 5000HD details leak -UPDATED

hopefully there will be a sager NP9280 with core i9 6 core processor and ATI HD 5870 GDDR5... anyways why is it so difficult to put GGDR5 in mobile cards? Is it because it runs faster and produces more heat?

 

Can you link to your ATI 5000 review please? My polish isn't up to where it should be and google translator isn't very nice when you want to browse through a translated website.

 

Perhaps rschauby, but...... since Kevin_Jack2.0 has not replied to this directly, we really have no way of knowing his thought here, do we? As such, I now must conclude he challenges my credibility here. To this , I respond as follows.....

I really can not blame Kevin_Jack2.0 for feeling as he does. Afterall, to an extent he is correct. I mean, haven't we seen what he concludes, many times?

Anyhow Kevin_Jack2.0, not everyone post here with ill-intent, to mislead, or for self-edification. To make my point here, I'll ask, is this what you do? I think not.

I have never claimed to be ' in the industry', but I do try to keep up, and I have many friends who are..... and I have never pretend'ed to be ' under some ironclad NDA', but what is more ' ironclad' to me, is my ' word' to a friend.

I assure you, I am NOT pretend'ing anything, and the new unannounced ATI 5870 model I have alluded to, is very real. But how could you know I speak the truth? To this, I have a very simple solution.....

To Kevin_Jack2.0, all 'nay-sayers', doubting Thomas's, or simply the curious...... PM me, and I will fore-tell this unannounced new model. I will provide specific manufacturer, model name and number, and all known configuration info..... but not in a way as to betray my source. The information will be provided to you in the form of an encrypted text. As soon as this model is formally announced (should be at CES), I will then provide decryption instructions. This should remove all doubts, and prove I have not been disingenuous in my claims here.

Since I will be providing this information 'in advance', it certainly will not be as Kevin_Jack2.0 suggest; "..... everyone can make these predictions, then when random model comes out, they'll point to it and pretend that's what they were talking about all along".

Incidentially, I know of others here who know of this new model. One of which, is well known and respected here. Thanks for the back-up, Bro! LOL!

Anyhow, I feel I have been challenged here..... and I'll meet that challenge. Let's see those PM's!

 

Come now, Striker, is this necessary?

No need to feel impugned upon.

 

No, not entirely, but I would like others to know I do not resort to what is implied.

At the same time dtwn, it's meant to share a bit of fun and humor. Note my three 'big grin' faces...... no harm intended here whatsoever. PM me and I'll send you my encrypted text.

 

Fair enough.

Just remember to play nice, the Mutaween are always watching.

ALWAYS.

 

LOL! I know! I'll be good, I promise (I'll try anyhow)!

 

Well, I not in the mood to read posts, but I saw my name bolded a couple times. Just know that I was making a semi-general comment about what I see go on in tech threads and news posts across the internet. I've been encountering this "insider trend" across multiple forums for the last few weeks, and I feel the need to point it out because there's too many phantom carrots dangling in my face.

Anyone who's disseminating credible information best ignore it.

Check my résumé; I name names when I have something to say.

In closing, carry on and enjoy your holidays.

 

Thanks for the reply!

You have a wonderful holiday season, as well.

 

Mobility 5730 and Core i5 430M in the ASUS X77.

http://www.fudzilla.com/content/view/16954/1/

Two CES releases down one to go?

 

LOL! Nice! Thanks Phinagle!

Mobility 5870 and Core i7 720QM (but customizable) in the XXXXXXXXXXXXX.

That's the heretofore unannounced 17.3" model I am anxiously awaiting, and expecting to be announced at CES.

That must be the third! Honestly though, in spite of that article, I think there will be more than three ATI Mobility 5xxx notebooks appearing at CES.

Come on CES!

 

I'm pretty sure the mobile memory chips are smaller, and thus cost more to produce. GDDR5 does not really run faster than GDDR3, 1200mhz (2400 effective) on desktops, so pretty much similar, but as double the bandwidth per mhz.

32nm GDDR5 chips are only due for Q2-Q3 2010 IIRC, so we may start seeing more of them at the end of next year.

 

OT but some cards are hitting 1450mhz+, and getting close to 1500mhz. gddr3 can't do that.

 

Yeah sorry. 1200mhz is just the defaut speed on most cards. I think on GDDR3, 1200 is near the max possible. I don't remember seeing any card with higher than GDDR3 2400, and it was the 8800 ultra

 

Uh, finally got that NBLB2 sample, too bad there's no program that would recognize the MR HD and the latest drivers are 9.9 beta If anyone's interested:
http://www.techpowerup.com/gpuz/9354u/ GPU-Z Validation

 

apparently gddr3 can go pretty far, although i personally wouldn't push it like that.
http://www.fudzilla.com/content/view/11683/40/ gddr3 at 2800mhz

 

Hynix had an announcement here a few days ago:

It's more the voltage - would want 1.35V GDDR5 rather than the older 1.5V memory makes the difference. Think samsung has 50nm definitely 1.5V cant remember offhand about a 1.35V variant. Cannot recall any 32nm GDDR5 announcements.

Note in the above the memory speed - 7Gbp/s or 1750Mhz will likely not show up in desktop products let alone mobile till they slightly modify the GDDR5 interface, too many errors and EM interference in surrounding components presently. 1500Mhz is more realistic general standard for next year.

 

not looking good if i want a GDDR5 card... so far only NVDIA has a GDDR5 card in a laptop.... GTS250M in Toshiba X500..... sad...

 

i wouldn't worry as much about the g5 vs g3..

from my testing...the only way g5 is really doing better is if it's on a 256 bit bus...on a 128 bit bus...it's slower than g3 on a 256 bit bus. (with respect to each manufacture (nvidia/ati)

example:
i ran a gt 240 gddr5 card against a 9800 gt....
the 9800 won...had the g5 card been on a 256 bit bus...it would have probably smoked the 9800.

so i wouldn't get to happy about having g5/128. for it to matter..it has to be g5/256...(speculations of course)

 

against a mobile 9800gt you mean? gt 240 only has 96 sp.

 

no. a DT 9800 vs a DT 240
g5 should have made up for the 16 shut down shaders
on a 128 bit bus..it doesn't. on a 256 bit bus..it would have

edit:
in the end...it wont matter if nvidia doesn't come with anything...because then you don't know if your really missing anything or not...

and again....(it's still all speculations)

 

Thats in no way a definitive comparison. The 9800gt has faster core, faster shaders, and more shaders.
Further, the 9800gt has 900mhz GDDR3 on a 256bit interface vs the GT 240 at 1700mhz on a 128bit interface. The GT 240 would need to have memory running at 1800mhz to compare fairly.

 

lol my short term memory sucks so much. That's exactly the article I had read. 40nm just appeared, mass production next year.

 

9800 gt


gt 240


and this guy running a 9800 with far less specs than mine



sorry...the only fair comparison your gonna get is 1 240 vs another 240. period.
needless to say...that 240 went right back to the store. a waste of money that i was lucky enough to get back!

 

I am just trying to get the point across that you can't attribute the performance difference between the 9800gt vs gt 240 to the difference in memory. The GT 240 is slower in every way. In fact I'd say the other variables have more to do with it than the memory.
Look at the clock differences you were running. You were running the 9800GT over 100mhz faster on the core and over 300mhz faster on the shaders.
In summary, you are comparing the 9800gt with it's
higher core clock,
higher shader clock,
higher shader count,
double the ROP,
24 additional texture units
and then you attribute the performance difference to memory?

 

i am not an expert in these things but my understanding is that a gddr5 running at 128 bit, "x"mhz core, "y"mhz shader will be similar/near to a grrd3 running at 256 bit, "x"mhz core, "y"mhz shader in terms of performance. but the gddr3 will consume more power than gddr5....
correct me if i am wrong.

 

sorry, but the gt 240 has far more bandwidth with a higher memory clock. if that we're on a 256 bit bus..the gt240 card would have won. im not going to look over every single card in exitence in the this category, because it's not worth my time. either you understand it or you don't. it's not rocket science for you to try and dissect. it was just to show how it fairs against the 9800. which it lost too. period!

lets pay attention first, before we go trying to take apart a S P E C U L A T I O N post. your opinion is duly noted in this, but if your trying to pass it off as facts...not going to work. sorry.

and for the record... a 4870m/280m/260m/4850m/9800mgtx/9800mgt all beat the gt 240..(test in question..vantage). and that's a single less powerful mobile card against a DT card. so what's your rational for that?

yep, that is pretty damn close.
they told us that they couldn't run gddr5 in the past because of the heat it was kicking out. the 4870m's should have had gddr5 memory...but test we're ran and they couldn't keep them cool enough to be benefit the end user...so they scrapped it and ran g3 4870m 1g cards instead....which by the way...are kicking nvidias butt all over the place..*sniff* sniff* so much so i ditched them cards and bought ati's

 

This GT240 is running 128bits GDDR5 at 2100mhz (4200 effective). That gives him 134GB/s. That's almost as much as 512bits GDDR3 2400 (almost like the GTX 285). That is just too much bandwidth for what the core is capable of. I doubt the GT 240 would see a performance drop with half the bandwidth (134GB/s is what the Radeon 5870 has!)

 

Let's take a look at two examples:

That's GDDR3, so:
256 (bits) * 2 (Double Data Rate) * 993 (Clock) = 508416 bits / second = 63,3GB/s - so it's fine. Now another one:

That's GDDR5, so:
128 (bits) * 4 ( Quad Data Rate) * 1500 (Clock) = 768000 bits / second = 96,0GB/s
The GDDR5 memory behaves like Quad Data Rate, compared to Double Data Rate that every other DDR type memory gives.

 

and here is my 5970 over clocked.

 

show off

 

But the specs given by GPU-Z all have to be doubled right? I mean, they are only for 1 Crypress

That makes 64 rops, 320GB/S, 160B/s texture filtering, 3200 SPs

@Baginenny: My 5770 crashes after the slightest overclock, could it be that the defaut Vcore is too low? I'm pretty sure GDDR4 also acts like quad data rate. By why wasn't it more used? Because GDDR5 was ready and had better performance/price ratio?

 

are you serious? everyone knows the 5970 is a dual gpu *facepalm*
the part where it says..2 gpus enabled should have told you that.
think your looking for a 5870
oh wait...yes...those are the specs for 1 gpu showing in the example...hummm...since that's what it was posted for in the first place...
and that would be the gpuz example and not the vantage run..just in case you wanted to nitpick some more..

side note...
5770 vs 260...is this a fair fight?

 

That is exactly why. Qimonda (memory manufacturer) and other memory makers were ready with GDDR5 (which had double the I/O of GDDR4) soon after GDDR4 was available. I only remember a few ATI cards coming out with GDDR4, and they quickly and aggressively transitioned to GDDR5 as soon as they could possibly get their hands on it. It allowed them to decrease their bit width (from things like 512 or larger on Nvidia cards) down to 256 bit even on high-end cards due to the massively increased I/O made possible by GDDR5. This allowed for thinner PCB and greatly decreased cost to manufacture for both the chip maker (TSMC) and the board makers (ASUS, Sapphire, HIS, etc). GDDR5 was one of the key reasons the high-end AMD/ATI 4000 series prevailed over Nvidia's equivalent offering.

Now that it's been about a year, GDDR5 is finally being manufactured at a small enough process and low enough voltage to eventually be put into notebooks, though as others have mentioned, it will take at least another two quarters before that happens.

 

there is also stated on record why they didn't use g5 ram in the earier notebooks...and it was as mentioned earlier..heat. temps at idle we're like 60 to 70 degrees.(speculations because i dont have the direct quotes at hand) and someone explained that the gddr3 memory was performing better than the gddr5..(go figure. i was looking mighty surprised at this one)

edit:
i think it's on dells site somewhere.

 

could be higher latency at the same speed.

 

From what I recall the reason they didn't put GDDR5 in mobile cards was that GDDR5 had some problems with working at low clock speeds resulting in some image distorsion - so they had to use higher clocks in idle mode ergo the power consumption would be higher than with the GDDR3. The power consumption in stress is lower with GDDR5 though - they require lower voltages.

Could be - you can look for some Voltage Tweaks that will allow you to change the Vcore and Vmem for your card. If I remember there was some tool called MSI Afterburner which managed to do the trick, but I'm not sure it works with 5700 series.

 

260 a bit better for $40 more expensive.

 

see...no fair fight anywhere amongst cards....thanks for the reply!

and at frys...they are about the same price. back and forth between each one costing 10 bucks more. 179.00 to 199.00. super clocked being 224.00

that goes for microcenter as well. they are the same price....189.00

 

Newegg.com has the 5770 for as low as $144
http://www.newegg.com/Product/Produ...der=BESTMATCH&Description=Radeon+5770&x=0&y=0

The 2nd gen cards are about $165 and the 1st gen between $170 and $200 (1st gen= reference ATI fan, which is better, and most of them come with Dirt 2). Prices have gone up because of the 40nm chip shortages, but they should go down in some times. I paid mine $160CA at launch.

If they are the same price, it's like you want. The 5770 has DX11 + lower power consumption but performances are a bit slower.

 

ok, ya killing me now...rotflmao! im still not going to buy the cheapy version of the card. (144.00 sounds like rma/returned/open box) so out of that bunch...they would get 169.99 to 200 bucks out of me for the card.
at a store. where i can take it right back and get my money back or a better card on the spot. the egg is cool and all, but im not a firm believer in wait..wait..wait...lol...so that extra 10 or 20 bucks in the store im going to pay.

great find by the way! nice prices. i almost got a 5770 for testing till they could find my 5970...but lucky for me..they found it after an hour.haha

and last but not least...that's what brought me around to trying out the gt 240..i was going to use it as a physx card. i was thinking like you with the 5770..
new gen card
new tech
g5 memory
hdmi
dual monitor and possibly 3 (speculation on that)
faster than an a 9800 for physx (is what i was thinking-bahaha)

boy was i wrong..

 

Maybe I'm misunderstanding you. The GT240 obviously has slightly lower memory bandwidth than the 9800GT at default clocks(1700mhz vs 900mhz). The GT240 has ~55 GB/sec vs the 9800GT with ~57 GB/sec bandwidth. I'm not disputing that fact. I am disputing your conclusion that the GT240 is slower because of it's memory and bus width combination. I am saying that the reason your benchmarks suggest the GT240 is slower is because it is slower in every other respect vs the 9800GT. In fact, the memory most likely has the least to do with the GT240's performance deficit.
Would the GT240 w/ 256bit bus have more memory bandwidth than the 9800GT? Yes of course, but I doubt that would make the GT240 faster than the 9800GT except in very select circumstances.

I'm not sure there is any reason to get half nasty about this.

I am looking at the last example. Again, I see a comparison between the 9800GT which has every reason to be faster than the GT240. The 9800GT in this example has a faster clock, faster shaders, more shaders, double the ROP, 24 more texture units and approximately the same memory bandwidth that the GT240 has.
Then I see you making the claim that the GT240 is slower because of memory bandwidth. If the GT240 had a 256bit bus it still most likely wouldn't be faster than the 9800GT

This goes to my point. There is no reason the GT240 would be faster than a 9800GT for physx or anything else, regardless of GDDR5. I think it is you that has misunderstood "the numbers," not me.

 

nice try, but no.
1: my post started as just an informational thing on my findings about two cards. nothing more.
2: you thought you would try to take the post apart and correct it.
3: you should have taken it as is. plain and simple.
4: this is very important...when the phrase SPECULATION is inserted. it's for people who think we are speaking facts, when in turn...we are just voicing an opinion.

so in closing...it's you who are misunderstanding things...

ps: i think you better go relook at them screen shots..because your numbers are off yet again!..*facepalm* the gt 240 is at 134 gb

 

My friend wanted a 5870 to match is i7 920 but because of the higher prices/hard to find, he opted for a dual 5770 set up.

And now you just remind me to set up my 8600M GT as a physics accelerator.

 

1: it doesn't work on vista
2: it works on windows 7
3: you have to install the patch
4: you need certain drivers as well..it doesn't work with them all.
5: my 9800 edged out the 240 by 10 on physx (155 to 165 - vantage)

edit:
how does he like it?
tell him to run the heaven demo as well..

 

Woops I made a typo. Not 8600M GT but my desktop 8600GT XXX eating dust in the basement.
Yeah he likes it. Crysis maxed @1920x1200 with 4xAA, and he's in heaven

 

already knew what cha meant about the 8600. (not the unnecessary correcter here..lol)
heck, i might have bought that card from your for benching purposes..im trying to get my points up over at hwbot.

nice, glad he likes it.

all this talk is making me a bit antsi about the up and coming models..very interested in this 5800 model..it may be the answer to not running crossfire, but with substantial gains over the 4870 crossfire.....(SPECULATION - for the post correcting bunch..*snickering*)

 

Maybe part of your confusion steams from GPU-Z quoting memory bandwidth to be 134GB/s. That is not reporting correctly. At default clocks the GT240 has (1700mhz*2)*128bit ~ 440,000Mbits/s = 55,000MBytes/s ~ 54GBytes/s. So, at the clocks shown in GPU-Z: (2100mhz*2)*128bit ~ 540,000Mbits/s = 67,500MBytes/s ~ 66GBytes/s, not 134GBytes/s.

 

until they correct it..that's what im going by
anything else?
Product Specifications

Core Clock (MHz): 550 MHz
Processor Cores: 96
Processor Clock (MHz): 1340 MHz
Fill Rate (billion pixels/sec.): 26.4
Memory Amount: 512MB GDDR5 On-board Frame Buffer
Memory Frequency (effective): 3400 MHz
Memory Interface: 128-bit
Memory Bandwidth (GB/sec): 54.4 <--ps thanks for the correction here!
On-board Outputs: DVI-I and HDMI and VGA
BUS Technology: PCI Express 2.0

Key Features

• NVIDIA® Unified Architecture
• Full Microsoft DirectX 10.1 support
• NVIDIA CUDA™ technology support
• Microsoft Windows 7 support
• DirectCompute Support
• OpenCL Support
• NVIDIA PhysX™ technology support
• NVIDIA PureVideo® HD technology support
• Dual-Link HDCP-Capable
• OpenGL 3.1 support
• Max resolution DVI 2560 x 1600 @ 60Hz
• Max resolution VGA 2048 x 1536 @ 80Hz
• Max resolution HDMI 1900 x 1200 @ 60Hz

so im going to say you are correct in this aspect.

which brings me right back to the beginning...i WOULD HAVE THOUGHT this card would be better than a 5 year old card. i was wrong. . exactly why it got returned.

also:
what makes it worse..it didn't beat a 9800M gtx card either...damn shame....
http://service.futuremark.com/compare?3dmv=563166
which has far lower clocks, but riding in at 60 on bandwidth.
can't get any fairer than that. considering the mobiles use far less power.

 

Yeah the GT 240 is definitely overpriced.

As you can see with the 9800M GTX the GT 240 is getting beat not because of the memory but because it has half the ROP and less texture address/filter units.

That was what my whole disagreement was about. Either way, I'm glad you returned the GT 240