Well *someone* needs to get one to experiment with!
Well *someone* needs to get one to experiment with!
Sad part is, I don't have any spare RAM and at current prices even just 4GB starts putting it back out of impulse buy.
I'm guessing these are single channel RAM which would make it a bit cheaper.
Edit to add: Another article from Fudzilla, http://www.fudzilla.com/home/item/34...sted-in-europe, seems a lot of coverage for not a lot of information. Perhaps they want to play with one as well
Last edited by DanceswithUnix; 11-03-2014 at 10:09 PM.
That was an interesting read... I've heard about the Anandtech forums but wow, just wow. The sad part is that the moderating team are incapable of recognising inconsistencies. It seems group think has taken hold of the moderating on Anandtech, outside information seems worthless to them.
Just occurred to me that you could be given an infraction for you post CAT, and so could I for thread crapping. I got so lucky finding Hexus first.
Well, I am more worried about some of bias I detect in the AT editorial line. As in buying Intel and Nvidia PR wholesale but taking a far more sceptical stance with AMD. I actually would prefer a far more sceptical stance all around. In fact a lot of their mobile coverage is rather poor as is their Apple coverage. I guess AT do know that if PR takes a dislike to them they might not get as much stuff but the Apple Pro coverage I found hilarious (Apple PR obviously want to push it's aesthetics and gloss over the lack of expansion) and the reviews of the Nexus too.
I know, Anand likes cloud storage (even praises the demise of the SD card) and isn't a Apple Pro power user but those are his personal usage patters and a review should cover more than that (but what does a journalist who travels internationally do when their data is in a cloud and their data plan only applies to the US?)
On the AT forum you have a lot of silly fans (you know the ones who think 'higher price=better') who seem to worship Intel and Nvidia (strangely enough that Nvidia were responsible for the biggest and most widespread silicon failure in history seems to get glanced over). And those who come there to push a financial agenda (shareholders or wanna-be company directors). Both of them are easily ignored I guess but the mods to seem biased. But the Intel/Nvidia posters also know how to play the game better.
Actually, the techsoda poster does or did used to overdo it. Not any worse than the other side but he actually doesn't help his own cause sometimes. But there really are people who spend a lot of time and energy putting AMD while praising Intel and Nvidia or glancing over any of their faults. Strange hobby!
EDIT: while I do like reading Charlie's rants, that kind of stuff obviously doesn't pay the bills. Hard to get advertising revenue when you keep criticizing so many companies. I do think the biggest reason Charlie is so hated by club Nvidia is because he was the only who exposed Bumbgate. But, it seems there are far too many lackeys out there afraid of companies PR departments. I think the likes of Guy Keyney (PCW) would not get published in these days of editors in thrall to company PR departments. Ok, it's only tech journalism but the same sycophantic behaviour seems to be becoming more common in news sources in general.
Last edited by kompukare; 11-03-2014 at 11:34 PM.
To be fair, said lack of rabid fanboyism, and a tendency towards self moderation, make our job very easy at Hexus - I rarely see any thread I feel the need to step in to in my role as a moderator. We moderate with a light touch here because that's generally all that the forums need. OTOH, there's a few mods - myself included - who can come down like a ton of bricks when the need arises.
I'm pretty sure they are, yeah. In fact, one of the articles linked earlier makes the point that while AMD are trumpeting their DDR3-1600 support vs. slower speed on many Atom SKUs, those same Atom's are often dual channel vs. single channel for AMD. OTOH, Intel are using HD2000 based graphics on the Atoms, so it's not like they'll get any real world benefit from dual channel RAM, and if AMD can genuinely turn round an upgradeable] SoC + Mobo for ~ £50 that's pretty impressive. Plus 64-bit DDR3-1600 is exactly what AMD have been using on discrete cards with similar performance/specs for a while now, so I don't think single channel RAM is going to hold AM1 back particularly.
Interesting experiment from Tom's Hardware - using a single slot, low profile (AMD) cooler on the 750Ti;
http://www.tomshardware.com/reviews/...0-ti,3761.html
The GPU throttles its boost clocks a little, but the clock and performance hit is in the region of 5%. A made for market low-profile-single-slot 750Ti could be fitted with a slightly more appropriate cooler, which would help with that.
I've been banging the drum for low profile/single slot 750Ti's since they arrived, so pleased to see it works. They could be amazing for very small form factors. Some of those after-market coolers are just ridiculous on a 65w card. It's strange that no board partner has tried it.
Tom's did a similar experiment with a passive cooler on a 750Ti a week or two ago too (again, having to use an AMD heatsink)
There seems to be abnormally high coverage of the GTX750TI on some websites. It makes me wonder whether Nvidia PR is involved.
We never saw anywhere as much coverage with AMD cards like the HD6850 or the HD7750.
BTW,some major thottling problems with the GTX750TI with a single slot cooler can be seen.
Even the passive card is starting to throttle more and more over time.
Every single test Toms has done is in an open air bench setup or a full sized case.
In a low profile case like those Wesena ones,there is very little ventilation.
The hot air gets recirculated - I should know I used a single slot HD5670 GDDR5 in a modified Shuttle case and plenty of higher TDP single slot cards.
BTW,there is a low profile GTX750TI:
http://www.guru3d.com/news_story/gal...file_card.html
It can be spun as much as anything but the GTX750TI consumes more power than an HD7750 and uses boosting which the HD7750 does not. Hence it will always need better cooling.
Moreover,every Nvidia Kepler mobile SKU never ever used clock boosts,as it aggressive.
I stated this two years ago at the GK104 launch and I was correct.
Last edited by CAT-THE-FIFTH; 12-03-2014 at 10:42 AM.
I think it's just a really interesting unit. The first and only showcase of Maxwell (albeit on a different process), and pretty remarkable performance/watt. There's also been no other new GPUs (other than rebadges) or CPUs since it appeared.
That's a 6% clock throttle (it's a misleading graph - doesn't start at zero)
It uses more power than lots of cards, but none of them perform anywhere close. Even a low-clocked non-boosting 750Ti would outperform anything else on the market in such a potentially small package. Reference 750Ti gives almost twice the performance of a 7750.
Last edited by cheesyboy; 12-03-2014 at 10:47 AM.
It isn't really when you consider what they have done with the GM108 when compared to the GK106 and GK104:
1.)There is only ONE GPC,meaning no interconnects at all. Higher end Kepler GPUs have more than one GPC. Interconnect problems were one of the main issues the GF100(the GPU in the GTX480) had power consumption problems.
2.)DP compute is even further degraded.
3.)Each pair of units in an SMX has half the texture units,and they are shared.
4.)Added more low power states.
5.)128 bit memory controller.
6.)A dedicated die - many of the competing GPUs are die salvaged parts.
7.)Lacks dedicated functionality of certain competing products.
8.)Relies on aggressive clock boosting tech.
9.)Added loads of cache to make up for bandwith issues.
If a bigger Maxwell was made on 28NM,it would not be as big an improvement over Kepler as people think it would be,as a lot of the kibbled functionality in the GM108 will be present. This is why Nvidia is probably waiting for 20NM.
Even the R7 260X is an anomaly - it consumes around 20% more power than an HD7790,which means TrueAudio functionality comes at a great price. That card was just a waste of time IMHO.
Which means in a mini-ITX case,its going to be worse. Mobile Kepler parts don't use boost and it is very telling.
Not even the GK208 based parts which consume less power than the GM108.
I have owned plenty of single slot cards in small cases,ventilation is always the problem in these situations.
There was a single slot bus powered HD7850 2GB which was released last year to OEMs.
Yes,an HD7850 2GB. A die salvaged part with a greater TDP.
That is based on a first generation 28NM GPU released before the GK104.
As usual the Nvidia hype machine works wonders. The problem is that cards like the GTX660,R9 265 and R9 270 work fine even in an SG05 or a Shuttle and are generally faster than a GTX750TI in most cases,and are in a similar price range.
My Xeon E3 1220 and GTX660 based SFF PC is lucky to consume 200W at the wall when gaming for example.
If Maxwell was so brilliant then Nvidia would be refreshing their cards on 28NM NOW.
They are not. They are waiting until 20NM as the power consumption drop and density increase will mask any efficiency losses when scaling up.
Edit!!
The GK106 for example has 3 GPCs.
Last edited by CAT-THE-FIFTH; 12-03-2014 at 11:20 AM.
Your passion on this exceeds mine, CAT.
I take more interest in the low to mid end of the GPU market than I do the behemoths at the top end and, as such,I'm impressed with the 750Ti. What you say about maxwell scaling and power on 28NM is almost undoubtedly true, but wasn't Maxwell always intended for a process shrink? 750Ti is an anomally in that sense, almost a tech demo, and that's why it has attracted interest.
As for your other points;
260x: might the 10% clock speed increase and 2Gb vs 1Gb memory make up at least some of the power difference?
M-ITX case: 750Ti's boost setup limits itself to 80 degrees. If it can't keep below that it will revert to reference speeds with no boost; 1020mhz. That's another 5% or so below Tom's benchmark, probably translating to a 4% or so further performance dip. It's still great.
7850 bus powered: Whilst that's impressive, I would presume it couldn't be replicated in large numbers which is why we never really saw one. 7850 reference is well over the bus limit for power draw (whilst the 750Ti isn't).
The thing is with AMD parts,is that don't appear to bin them as much as Nvidia do as touched on earlier in the thread by another poster. The HD7850 is rated for a TDP of 130W yet is nowhere near that,and some of the undervolting tests on the AMD Pitcairn based cards have been very interesting. Its the same with the CPUs,the Phenom II X4 CPUs could be undervolted a decent amount.
Would the GTX750TI throttle under base clockspeeds?? Maybe we need to see a test in a case like the following over an hour:
http://techreport.com/r.x/thermaltak...ntq/front2.jpg
That is an example of a small media case,ie,the Thermaltake Element Q.
The Galaxy LP card is 1.5 slot,so I think a single slot card would need a modification of this:
http://img.hexus.net/v2/graphics_car...50/7750-01.jpg
A bit more beefier than that probably due to the boost clocks.
However,this is the Wesena/Streacom case I am talking about:
http://www.streacom.com/products/f7c-evo-chassis/
I am not sure how well the GTX750TI would do in such a case with a single slot LP cooler.
Shuttles and cases likes the SG05 seem fine with higher TDP cards.
Am amused that you single that card out. My son was having problems playing Portal which smacked of video driver problems. Thinking time was finally up for his 8800GT I was about 5 mins off hitting the buy button on a 260X when I got Steam to check the game binaries and it found and replaced a corrupted one. Why did I look at that card? I wanted 2GB of VRAM, about £100, and knowing how AMD abandon older tech in their drivers I wanted the latest GCN revision. It would have gone in my wife's PC, she has to be AMD.
As for refreshing on 28nm, given how many months is takes to get a chip into production and how imminent 20nm is, it would seem a good idea to wait.
Spent last night reading through that CAT-THE-FIFTH (thanks for the link), that's pretty shocking i didn't realise quite how biased some of these places were (although some places can be ofc). I tried to read a fair few forums, toms, ocuk etc... i found even toms can be a bit biased but that moderating is insane, and completely unfair. I feel sorry/bad for people talking about AMD on anandtech now :s Shocking.
Thank the lords i was guided here, to this great forum!
Last edited by Jellyfish; 12-03-2014 at 04:58 PM.
The thing is the HD7850/R9 265 are around the same price,or only slightly more and are much faster,plus for around £120 to £130 you can get GTX660 and R9 270 cards too.
There is this R9 260X with BF4 for around £80:
http://www.scan.co.uk/products/1gb-x...dvi-pci-e-card
It is only a 1GB card,but if the 2GB cards were that price things would look a bit better.
£80 for a 260 and bf4 is a pretty sweet deal for the money. Not bad at all, but yea with 2gb would be even better.
There are currently 7 users browsing this thread. (0 members and 7 guests)